Site Loader

1. INTRODUCTION
The development of technology, especially in the last decades, allowed the possibility of huge changes in the businesses’ way of creating new products, demanding higher quality requirements, prospect financial growth, mass produce and communicate internally and especially with their customers. Nowadays, it is unquestionable the presence and importance of technology in the modern and dynamic world that we live in, as it has created unlimited possibilities. Unlimited possibilities as the progress of the Web 2.0 that made possible for online users to choose new word-of-mouth alternatives as online reviews. Those are defined as ways of consumers expressing their thoughts and opinions towards any type of product or service in an online community (Kem et al., 2014).
Online reviewing became one of the key factors on the pre-purchase decision, as for instance, nowadays 63.6% of consumers check online reviews on Google before deciding whether to buy/visit a product/destination. More, other platforms as Yelp and TripAdvisor became multimillion dollar companies, due to this need for consumers to obtain information before taking any decision. The first, for instance, has more than 186 million users posting 150 million reviews online monthly, about the most varied subjects (ReviewTrackers, 2018). The TripAdvisor platform itself has 455 million users, which have written more than 661 million reviews since it’s’ foundation – it accounts with 7.5 million accommodations, restaurants and attractions. From those, over 1.2 million are hotels, 4.7 million are restaurants, and 975 thousand are attractions. The extension of these firms is worldwide, and the numbers show the importance of this platform to the online community (Smith, 2018).
These opinions have gained such importance that every industry is dependent on it, and the tourism one is no exception. According to Schuckert et al. (2015:608) “71% of independent travel related bookings are done online, while 36% of all package tours are booked online”. Moreover, 65% of travellers looking for leisure activities or vacations will pursuit for information online before deciding to take the travel, and “69% of their plans are determined by online travel reviews” (Fang et al., 2016:498).
Having cleared out the importance of online reviewing, and online communities, globally and particularly in the tourism industry, it is time to state what this thesis is aiming at. The literature has studied online reviews and the tourism industry in multiple and different ways, since the effects of the first in the hotel booking intentions (Cheng et al., 2017), the effect of the user’s membership level ranking on the consumer’s decision making (Chen, 2015) or even trying to predict future trends based on past stats (Schuckert et al., 2015). A gap in the literature was found, when using online reviews to compare how tourists perceive two different destinations, and this is the theme that this thesis proposes to study.
The first destination was chosen based on its importance to the country that means the world to me: Portugal. Madeira has a GDP of 5,244 million Euros (data for 2017), and its significance to the world is gradually being recognized – it was awarded as Best Island Destination by the World Travel Awards in 2017 and Mankind Heritage by UNESCO, in the same year. The second one was chosen based on the first, meaning that it had to be an island destination with cultural similarities so that the purpose of this study made sense. After an extensive research, Bermuda was selected to be the second destination to be studied.
Furthermore, having defined the gap in the literature that this thesis proposes to study, as well as which destinations it will be aimed at, a clear and explicit research question had to be formulated: “How tourists perceive two island destinations with identical culture, but different demographic characteristics, through Social Networking Sites?”; To answer to the research question it will be used a mixed approach combining two different study methods: Netnography and Text Mining. It is believed that such combination is capable of providing suitable and relevant information to deliver a relevant and appropriate answer to the announced question.
In order to do what was mentioned before, the project must be structured. This dissertation will be monitored under the following structure (see Figure 1):
1. Introduction – it is in this chapter that the theme being studied is introduced, as well as its relevance to the existing literature and the structure that this thesis will follow;
2. Literature Review – it will be analysed pertinent themes to the final goal (Customer Experience, Online Engagement, Loyalty and Satisfaction, Social Media and Online Reviews, Tourism Experience and Sentiment Analysis);
3. Methodology – it will be presented the methodological process of the two types of studies conducted in this thesis: Study 1 – Netnography and Study 2 – Text Mining;
4. Results – chapter intending to share the results obtained regarding both studies;
5. Conclusions and Implications – Based on the Literature Review and the Results obtained, this chapter will include a Discussion with conclusions, Theoretical and Managerial Implications and Limitations and Future Research;
6. References – Presentation of the references used across the entire dissertation;
7. Annexes – Chapter containing the annexes from the entire thesis.

2. LITERATURE REVIEW
The following chapter intends to give a clear theoretical background of the constructs that are important to be studied to the research field, such as Customer Experience, Tourism Experience Model, Online Engagement, Online Reviews and Sentiment Analysis. The content hereby mentioned results of an extensive research and aims at being a preponderant help in obtaining new insights and breakthroughs in the Marketing field.

2.1. CUSTOMER EXPERIENCE
The concept of Customer Experience has been considered in multiple and different ways (Hirschman and Holbrook, 1982; Gilmore and Pine, 2002; Thompson et al., 1989).
According to Lemon and Verhoef (2016) there are seven relevant theoretical literature findings across the years that led to the progress of the concept of Customer Experience.
The first one – “Customer buying behavioural process models” (Lemon and Verhoef, 2016), goes back to the 1960s-1970s when for example, Howard and Sheth (1969) created the “Theory of Buyer Behaviour” that consisted in a diagram flow where the buyer dependent on significate, symbolic and social environment inputs, along with exogenous variables decides on a certain purchase behaviour, and correspondent evaluation. Moreover, recent studies have indicated that those inputs and exogenous variables can be classified according five factors: Cultural, Social, Personal, Psychological and Economical (Ramya and Ali, 2016). It was also during the 60s that the famous AIDA model regarding advertising was created (Lemon and Verhoef, 2016), and nowadays continues to be a tool used by marketers, and keeps getting improved along the recent years – see Figure 2 below.

The theories concerning the buying process are crucial to understand the efforts that the firms need to have considering the customer experience (Pucinelli et al., 2009; Verhoef et al., 2009), and most of the early ones are still considered as the basis for the true construct of the customer experience (Lemon and Verhoef, 2016).
The second one – and one of the most important aspects for this research, is a topic that continues, nowadays, to be studied and it respects to the “Customer satisfaction and loyalty” (Lemon and Verhoef, 2016). There are multiple studies and researches towards this theme, and one of the most important aspects respecting this, is studies identifying the drivers and outcomes of the customer’s satisfaction (Bolton and Drew, 1991; Bolton, 1998). “Customer satisfaction measurement has become a rather standard practice within marketing, although other assessments and metrics have gained traction over time” (Lemon and Verhoef, 2016:72).
The third relevant finding is “Service Quality” and it brought “the focus on (1) the context in which experiences arise and (2) the journey mapping and measurement/assessment aspects of customer experience” (Lemon and Verhoef, 2016:72). It was through this subject that the worldwide known SERVQUAL model was created (see Figure 3) – it calculates the gap between customer’s expectations and perceptions – it can be regarding products or services (Mont and Plepys, 2003).

The fourth literature finding that was important for the development of Customer Experience concerns “Relationship Marketing”. It was during this period that most of the research regarding B2B channels was designed (Geyskens et al., 1998), but also B2C (Sheth and Parvatlyar, 1995; Perterson, 1995). Relationship Marketing has allowed to understand in a more meaningful way the relationships with the customer, increasing the range of the previous concepts analysed, by including emotions and cognitive perceptions (Lemon and Verhoef, 2016).
The fifth one concerns “Customer Relationship Management” (Lemon and Verhoef, 2016) , also known as CRM – “is the activity which is interested in the main customers of the organization, in the efficiency of the organization and in the customer knowledge management, with the aim of enhancing the effectiveness of the organization decisions related to customers, leading, therefore, to the improvement of the marketing performance in particular and the organizational performance in general” (Soliman, 2011:167), and its contributions to the customer experience literature, concentrates on the relation of determined elements with each other and the firm’s results (Lemon and Verhoef, 2016).
The sixth one refers to “Customer Centricity and Customer Focus” (Lemon and Verhoef, 2016)., and it concerns on strategic attitudes being projected, applied and discussed during the 2000s, focusing on the customer as the centre of the marketing strategies of the firms (Sheth et al., 2000). It was during this time that the so called ideal customer was created, based on benchmarking and the firm’s data on real clients, which enhanced the customer experience design (Lemon and Verhoef, 2016; Herskovitz and Crystal, 2010). All in all, it can be said that this centricity in the customer allowed firms to processing all the customer experience in their viewpoint, instead of the firm’s angle.
The seventh and final one, is “Customer Engagement” (Lemon and Verhoef, 2016), which is also quite important for future notes in this dissertation. “Customer engagement is connected to customer value management through its objective, namely, to maximize the value of firm’s customer base” (Bjimolt et al., 2010: 341). A definition of this concept was given, stating that customer engagement manifests itself through complex behaviours fitting the brand, going outside the spectre of the purchase itself (Verhoef et al., 2010; Doom et al., 2010).
Having concluded which findings served as background and that led to the appearance of the Customer Experience concept, it is now time to study this subject.
Some researchers see this concept as autonomous of its form (Schmitt et al., 2015), which ultimately means that it is a holistic concept, believing that it englobes five different customer’s parts: cognition, emotions, socialization, sensations and spiritualism, which all connected form a response according to the different interfaces with a brand/firm (Lemon and Verhoef, 2016). Other definition states Customer Experience as “encompassing every aspect of a company’s offering – the quality of customer care, of course, but also advertising, packaging, product and service features, ease of use, and reliability. It is the internal and subjective response customers have to any direct or indirect contact with a company” (Meyer and Schwager, 2007:118).
As it is possible to see there are several definitions, but some of them are more accepted than others. For instance, Schmitt (1999) identified in his studies five different varieties of experiences: “sensory (sense), affective (feel), cognitive (think), physical (act) and social-identify (relate) experiences”, a definition also congruent with Verhoef et al. (2009). A different, yet similar, conceptualization was created by Brakus et al. (2009) – their work englobes two main domains: internal responses – such the ones already mentioned: sensations, feelings and cognitions; and behavioural responses induced by stimulus related with the brand’s design.
Globally academics have come to agree that this definition involves the following variety of dimensions: cognition, emotions, behaviour, senses and social constructs (Schmitt, 1999; Verhoef et al., 2009), that respond “to a firm’s offerings during the customer’s entire purchase journey” (Lemon and Verhoef, 2016:74).
On a particular universe, the retailing one, Grewall et al. (2009) stated that Customer Experience is able to be categorized along concepts related with the retail mix (e.g. price experience). The Online Customer Experience has also gained a tremendous importance along the recent years, Rose et al. (2012) stated that the feeling of having full authority needs to be full comprehended by e-retailers, because it influences the emotions present in an online transaction. The same was concluded by Berry et al. (2002) saying that firms need to understand that the emotional component of the customer experience needs to be managed within the same seriousness they carry regarding their products and services functionalities.
On a tourism level, authors have shown that there are travellers who keep progressively researching for vacations, accommodations (such as hotels), restaurants, or just recreation activities/experiences online (Loureiro, 2015; Sreejesh and Ponnam, 2017). The choices of these travellers often pass through platforms such as TripAdvisor or Booking, which aim at providing as much information as possible regarding travel logistics matters (Bilro, 2017). As seen before, the experience that the traveller sets in refers to the holistic view of an experience (Bilro, 2017; Lemon and Verhoef, 2016; Schmitt, 1999), but the online world “offers different stimuli in some features, such as aesthetics/design, information and interaction” (Bilro, 2017:83), involving mostly the cognition and emotions fields (Sreejesh and Ponnam, 2017). If the experience motivates the customers through those stimuli, they will have tendency to become more involved with the brand and the experience itself. “Indeed, a strong sense of motivation, involvement and a positive response to online stimuli may generate online engagement” (Bilro, 2017:83). In the case of positive engagement, it “can promote the creation of an emotional tie between a firm’s brand and its customers which in turn enhance customer loyalty” (Gentile et al., 2007:404).

2.2. ONLINE ENGAGEMENT, LOYALTY AND SATISFACTION
Starting from the conventional definition of Engagement, already seen in the previous subchapter – Doom et al. (2010) stated engagement as a behavioural process in form of a customer attitude towards a firm, going beyond a simple purchase, driven from motivation variables (Bilro, 2017). According to Bowden (2009) and Fernandes and Esteves (2016), customer engagement starts with the customer satisfaction and prior attitudes, like cognitive and emotional ties with the brand, and involvement. If these factors are considered positives, it leads at the end in engaged and loyal customers.
One aspect that seems to be fundamental to several authors is the role that consumers must have in order to feel fully engaged, concluding that it must be active and most of all co-creational (Brodie et al., 2011; Kumar et al., 2010). This interaction between the firm and the customer is according to Brodie et al. (2011:14) crucial and the basis of the definition of this concept – “a psychological state that occurs by virtue of interactive, co-creative customer experiences with a focal agent/object (e.g., a brand) in focal service relationships”.
This concept is nowadays seen as a way of firms taking benefit from, and along the years literature has brought different branches of this tree, as for example Customer Engagement Measurement (Hollebeek et al, 2014) and the study of its value (Kumar et al., 2010), among others.
Moving to the online world, the internet allowed firms to develop their own communities online, both through their website and SNS (Zheng et al., 2015) – SNS (Social Networking Sites), which allowed the creation of another branch of the tree, with the development of Online Engagement.
Making the passage from the conventional definition to the online environment, Bilro (2017:84) states that “the online engagement construct goes far beyond purchasing behaviour, is stimulated by motivational drivers (such as involvement, external stimuli operationalized as the atmospheric cues in stores, customer-generated media or other online contexts), and is supported by interaction, the exchange of information and messages”. Moreover, and following the same logic, previous studies by Doom et al. (2010) found that in a society that is clearly more interactive nowadays (with more firms-customers and customers-customers relationships) – through SNS and others, the non-purchasing behaviour generated through the previous referred stimuli, is now a critical point.

The Figure 4 presented above shows a diagram that represents the process of engagement in the consumer and firm’s point of view, through Social Networking Sites. As it is possible to see engagement on consumers may occur due to the online brand’s content created, but also through other consumers. The diagram shows therefore the use importance of the participation of brands in this SNS (Barger et al., 2016).
It is undeniable that Engagement and Loyalty are interconnected and only through the first it is possible to reach the second. Making a parallel with the online world, Online Engagement is the bridge to achieve e-Loyalty. According to Zheng et al. (2015) findings, brand loyalty is truly directly influenced by user’s engagement, and indirectly through their communities. Adding to that, e-marketers should pay close attention to their online strategies in order to achieve engagement.
According to Srinivasan et al. (2002:41) “customization, contact interactivity, care, community, convenience, cultivation, choice, and character” are significant components of e-loyalty, and thus firms should pay close attention when structuring the consumer buying experience process. Also e-loyalty is defined as the positive attitudes towards the e-retailer shown by the customer that will ultimately result in a repeated purchase (Srinivasan et al., 2002; Islam et al., 2012).
Moreover, e-loyalty can be perceived as customers being interested in recommending their experiences to others, and spreading positive information regarding the firm. The higher the positivity expressed, more expected the customers are to re-purchase/re-experience and spread positive e-WOM (Word-of-Mouth) (Islam et al., 2012; Zeithaml et al., 1996).
Finally, one important aspect regarding e-loyalty concerns e-satisfaction and e-trust as drivers of it. According to Ribbink et al. (2004:453) “a significant positive effect of e-trust on loyalty was demonstrated, while assurance was shown to affect loyalty positively both via customer satisfaction and via e-trust”, which translates into, the higher the satisfaction, the higher the probability of a customer becoming loyalty to the firm, thus increasing the re-purchases (Chang et al., 2009; Ribbink et al., 2004).
This last paragraph brings us another concept – Customer Satisfaction, which was described by Chang et al. (2009) as the capacity of the experience creating a positive feeling in the customer. More, the same study refers that a positive satisfaction may increase the re-purchase factor. Additionally, other positive outputs might come with positive satisfaction, such as diminishing switching opportunities and positive e-WOM (Anderson and Srinivasan, 2003).

2.3. SOCIAL MEDIA AND ONLINE REVIEWS
As it is possible to see from the previous subchapter the online world assumes a preponderant factor in both firms and customers strategies. Former articles showed that the customer online engagement will ultimately generate e-WOM (Islam et al., 2012; Zeithaml et al., 1996; Anderson and Srinivasan, 2003), and it is this way of spreading information in the world that this chapter will deal with.
The rapid development of technology allowed users to have the possibilities of opting around different word-of-mouth alternatives, “such as online user reviews and professional reviews, during online decision making” (Zhou and Duan, 2016:202). These user reviews contain the opinions, thoughts and the significance attributed regarding certain services or products to the ones who wrote them (Zhang et al., 2014), and can be found in media, firm’s websites, forums, blogs, social networking sites and specific online reviews sites (Cheung et al., 2008; Lee and Youn, 2009). From those it is possible to find two different e-WOM types: e-commerce WOM and social media e-WOM (Yan et al., 2016).
Online reviews belong to social media e-WOM, and assume preponderant helpfulness for fellow users in two main features: stand out the most important information regarding a product/service and the possibility of evaluation of those same products/services (Zhou and Duan, 2016). The same should happen for experiences and activities, once according to Pan et al. (2007) e-WOM also influences the tourism industry, and for that reason it makes online reviews a key source of data for travellers.
Moreover, “often consumers do not follow the expert reviews and look for user reviews that provide view of people who experienced the products in their day to day life” (Gobinath and Vidyapeetham, 2016:412), the same happening in the tourism industry as travel reviews are more fun, and provide more reliable and useful information compared with the one posted by travel agencies (Ye et al., 2009).
Another factor that brings possible customers to use online reviewing as a source of information is the cost of search – “the advantage of online reviews is that they reduce this search cost and provide variety of information for the consumers to aid them in their decision making” (Gobinath and Vidyapeetham, 2016:413).
The decision making is therefore, influenced in large scale by the information that it is available online written by reviewers (e-WOM), and it makes the purchase decision making sturdily influenced by it (Goldenberg et al., 2001).
Some researchers have studied the reasons that bring users to go online and present their opinions towards their customer experience. Zhou and Duan (2016:208) showed that the one thing that drives reviewers to share opinions online is due to the concept of self-enhancement – “defined as user’s emotional desires to gain attention and enhance their images among others”. Other authors refute that, stating that it is not the only driver, as social benefit and economic incentives are also reasons that might explain the motivation behind reviewer’s behaviour (Yoo et al., 2013).
The previous mentioned authors also reveal that “customer’s reviews participation has a significant impact on site identification building” (Yoo et al., 2013:676). It can be assumed that online reviewing is, therefore, a co-creational experience, and for that reason one driver of users feeling engaged with the firm/product/service (Brodie et al., 2011; Kumar et al., 2010; Yoo et al., 2013), and potentially leading to customer retention due to e-loyalty creation (Zheng et al., 2015).
To understand the importance of online reviews in the tourism sector, since 2015 data have shown that “71% of independent travel related bookings are done online, while 36% of all package tours are booked online” (Schuckert et al., 2015:608), also Fang et al. (2016:498) got to some additional results “65% of leisure travellers will search online before deciding on a travel destination, and 69% of their plans are determined by online travel reviews”. Peng and Chen (2013:282) have no doubts “as a source of travel information, the impact of social media on online booking is obvious”.
Another booking experiences or services, using as base trustworthy platforms – like TripAdvisor, turns customers less worried with fraud schemes, the motive being reviewers posting good but also bad experiences (Schuckert et al., 2015; Zhong and Leung, 2013). This was perceived by (Kusumasondjaja et al., 2012) as negative reviews being observed as more credible, nonetheless positive satisfaction reviews increase customer’s booking intention, while the contrary might reduce it (Cheng et al., 2017). Liu and Park (2014) found that a positive review achieves a higher usefulness on customers than a negative one. Other studies confirm that – such as the conclusions presented by Schuckert et al. (2015), stating that a different valence of reviews, have a profound impact on readers and their purchase decision. Moreover, consumers are believed to read a set of positive and negative reviews before taking any decision (Purnawirawan et al., 2012).
Moreover on the same subject, Chen (2015:1259) found in his studies that “positive information encourages consumers to buy products and negative information presents the problems of products, which will reduce the credibility”. Nevertheless, a negative review can be a good source of information for a firm, which may use it to spot product or service malfunctions, and be that way able to change the strategies (Chen and Tseng, 2011; Cheng et al., 2017). The literature also indicates that partial negative information inside a review does not mean that the reviewers are not disposed on recommending, and promoting, the firm (Bilro, 2017).
Moving on to text characteristics of online reviews, according to Fang et al. (2016) reviews that have a clear and explicit writing style are perceived as more useful, as well as reviews expressing uttermost sentiments. Perceived usefulness was observed by academics as a cognition process dependent on accepting or not the existent technology (Mou et al., 2017), being reflect in consumer’s behaviours (intention) regarding purchasing from an online source (Ashraf et al., 2016).
The writing style can be perceived as the language style, and according to Wu et al. (2017) it is a very important step in the pre-purchase evaluation, and sometimes it is dependent on the reviewer’s expertise level. In their studies, they mention “recent marketing research demonstrates that figurative language is more appropriate to express hedonic (vs utilitarian) consumption experiences” (Wu et al., 2017: 591), and Moore (2015) adds that for hedonic products (such as travels) explaining reactions is preferable, rather than explanations.
In certain platforms such TripAdvisor – already mentioned, it is possible to differentiate the different levels of membership of reviewers. In this case this platform makes it on a scale from 0 to 6, and according to Chen (2015:1258) “different membership levels often have different attitudes towards online reviews” and according to him, the ones considered as high ranked reviewers will normally receive, in a higher probability, the information, but will also reject it with a higher chance.
Again, making the parallel with the tourism world, findings indicate that “consumers actually exhibit lower levels of attitude and reservation intention if the review is written in figurative (vs literal) language” (Wu et al. 2017:590). The importance of the language is reduced when the review is written by a high expertise reviewer. This information is quite useful, because firms may use it as a company’s strategy to encourage low expertise into becoming high ones, and promote the already high ranked reviewers (Chen, 2015).

2.4. TOURISM EXPERIENCE
This chapter is dedicated on literature regarding the Tourism Experience and defining a model already created by other authors that will serve as background for future studies in this dissertation.
According to Otto and Ritchie (1996) the tourism experience may be defined as mental state defined by the participants feelings, while products and services are tangible and intangible – respectively, “experiences represent events that commit people in a particular manner and, as such, are memorable” (Mendes et al., 2010:112). Moreover, Pine and Gilmore (1999) mentioned that the tourism experience starts previously of the arrival to the endpoint and finishes with memories and intentions for visiting it in the future.
As it was seen during the Customer Experience chapter, the experience itself is a holistic concept (Schmitt et al., 2015; Schmitt, 1999; Verhoef et al., 2009), the same happens in the Tourism Experience, where “regardless of specific quality assessments and perceptions, tourists evaluate the tourism experience as a whole” (Mendes et al., 2010:112). This experience process is quite complex and involves a certain number of players in the act (Uriely, 2005), it consists “of a continuous flux of related and integrated services which are acquired during a limited period of time, often in different geographical areas” (Mendes et al., 2010:112).
The tourists’ satisfaction factors are more of a “combination of inherent factors and associated satisfaction in terms of acquired and consumed services during the holistic tourism experience” (Mendes et al., 2010:113) rather than a direct relation with acquiring a specific product or service. The term Service Quality by itself it is quite subjective, but when applying the Tourism concept, it gets even more subjective. The context outlines the hedonic characteristics of travels, and the tourism itself (Ritchie and Crouch, 1997).
Tourists’ expectations to acquire benefits – satisfaction included, are dependent on operation, symbolic and life experiences gathered on the activities done or services procured during the tourism experience (Vega et al., 1995). The concept of satisfaction, on a tourism level, has been defined several times. For example, the broader definition sees it as a cognitive view of the reaction (satisfied/not satisfied) of a customer on the post-purchase moment (Mendes et al., 2010). It was also defined as the fulfilment obtained by the pleasure provided by a product/service, or one of its’ features (Oliver, 2010). According to Bosque and Martín (2008:553) satisfaction may be perceived as “an individual’s cognitive-affective state derived from a tourist experience”, which ultimately means that it results from both decisions and emotions occurred during the touristic experience (Bigné and Gnoth, 2005).
Having seen a brief past of the literature towards the Tourism Experience, including its definition and the relation with satisfaction, it is now time to explore a model that will be used further on in this dissertation – the Tourism Experience Model (Gnoth and Matteucci, 2014).
According to Deans and Gnoth (2012) “the Tourism Experience Model (TEM) pulls together our insights into experiences and experiencing in tourism so as to be able to locate tourists in a commonly understood grid of references”, meaning that it allows destinations to understand what their position is, in the tourist’s experiences. Positive memories are fundamental being linked to the destinations otherwise tourists may easily switch destinies for any other amusing experiences.

The model – that can be seen in Figure 5 above, consists in an axial representation regarding two different sides. The vertical axis corresponds to the Consciousness – “relates to the style of how tourists receive their experience of the destination” Deans and Gnoth (2012:n.d).
According to Gnoth and Matteucci (2014:6) awareness has been processed that it leads to the experience of the object as its outcome”. The vertical axis comprises two different domains; the upper side comprehends “Human Being” and it relates with “finding ourselves as human beings while stripping ourselves of the socially induced values, habits and stereotypes to get close to our existential being as felt in moments of or similar to flow and peak experiences” (Deans and Gnoth, 2012:n.d), meaning that it is relating with experiences that go beyond the stereotypes and dogmas of society, where tourists are able to be themselves. The lower size relates to the “Person” and to roles constructed by society as socially accepted, or in other words “consciousness here receive experiences as guided by role-expectations” (Deans and Gnoth, 2012:n.d), meaning that the higher the tourist’s behaviour according the usual standards expected by society, the higher the role-authenticity is.
The horizontal axis – Activity one, comprehends two different kinds of experiences that tourists usually engage in, and it is on the principle that in order to experience tourists have to automatically move from one place to a destination, basically they are actives. The left side of the axis relates with Recreational Activities, while the right one Exploratory Activities (Deans and Gnoth, 2012; Gnoth and Matteucci, 2014).
Starting with the left side of the axis, that comprehends Recreational Activities: according to Deans and Gnoth, (2012:n.d) it deals with activities that are performed and a result of “habit, training and repetition (…) have their place in helping people regain their balance, their strength, or their self-esteem, or all of these together”. It involves the type of activities that are connected with entertainment, rather than changing the inner-self of the person, not going beyond of where they are right now. Some examples are enjoying a good hotel, its pools and decoration (Deans and Gnoth, 2012; Gnoth and Matteucci, 2014). The right side is quite different, as it involves the Exploratory Activities. In this type of activities tourists are found to explore new visions, understandings, body experiences and feelings, opposite to the Recreational ones that were a result of repetition and routine activities. The new insights are categorized by being challenging and a new way perspective of obtaining knowledge and skills. The tourists living such experiences are not thinking about the fun, but on “finding explanations for the phenomenon within its own” (Deans and Gnoth, 2012:n.d).
As it is possible to see in Figure 5, the authors have identified four different modes of experiencing, in the two main dynamics: Experiencing as Being and Experiencing as Becoming. Experiencing as Being includes two experiences (left side of the axial representation): Pure Pleasure and Re-Discover; and Experiencing as Becoming includes also two other experiences: Knowledge Seeker and Holist (Gnoth and Matteucci, 2014).
Starting with the first one – Pure Pleasure, the authors defined it as “predominantly role-authentic and expectations marked by the rituals and customs, fads and fashions of every-day life” (Gnoth and Matteucci, 2014:9). Being connected with the left side of the axis it involves activities that are considered useful, but already performed in another situations – and therefore, a result of repetition. Tourists experiencing this do it so in activities that involve no hard challenges, and the benefits are entirely to themselves. Due to being so stereotyped activities, the tourists tend to pay slight or no consideration to the environment that surrounds them, by the way their behaviour also indicates that the experience itself is full of allusions to the tourists’ country (e.g.: eating food they would normally eat at home), which ultimately indicates that the culture of the destination country is seen as an outsider (Gnoth and Matteucci, 2014). Another characteristic of such experience is looking for activities that involve “lazing by the poolside, casual sports, or socialising and catching up with friends while enjoying culinary delights” (Gnoth and Matteucci, 2014:10). Re-visitation of a favourite spot of holidays is also another trait of this experience, due to tourists seeking to repeat previous senses in order to “recover from the stresses and strains of every-day-life” (Gnoth and Matteucci, 2014:10).
The second experience, also on the left side of the axial representation is Re-Discover. In this type of experience tourists are looking for “goal-oriented activities that require some focus and possibly effort. After a period of relaxation has taken hold in the holiday-maker he/she now begins to seek to re-invent herself, to recreate” (Gnoth and Matteucci, 2014:10). Tourists seek to remember their past through the activities they are doing, but also hoping for better and brighter ones in the future. Repeating activities is linked with the past memories but also are able to make tourists explore new circumstances, some activities more than others (Russell and Levy, 2012). Literature also points out that the hunt for the existential being is characterized by activities that involve families (Wang, 1999) and family reunions (Gnoth and Matteucci, 2014). Often these groups of people travelling are looking to improve their socially learned traditions and behaviours facing the fresh journeys they are living. For many tourists getting closer with their existential being involves sports activities, and in that case travelling might be useful for rebuilding their skills and gaining new strengths. In such experience, leisure activities are commonly categorized by tourists spending time, money or skills – or all of those together (Gnoth and Matteucci, 2014; Stebbins, 2007).
The third experience – and already in the right side of the axial representation in Figure 5, relates with Experiencing as Becoming and is named Holist – being defined as “experiencing therefore relates to a form of exploratory activity of the authentic self that is marked by an experience of ongoing self-change” (Gnoth and Matteucci, 2014:13). Traits of such experience involve tourists being connected intrinsically with the environment that surrounds them, as well as living each moment as unique. Cohen (1979) stated that tourists seeking for Existence – as this ones (see Figure 5), often believe that living somewhere abroad of their home nations would be more significant and happy for them, inclusive they think about moving to those places in a long-lasting base. Moreover, the same author reveals that experiential activities often come with emotional ties to romance and nostalgia connected with the destination’s culture – also reinforcing the relationships with others. Other characteristics found in the literature, across the years, says that a “deeper self-transformation gave rise to new life trajectories; whereby adopting alternative lifestyles or undertaking new careers” (Gnoth and Matteucci, 2014:13). There are some types of experiences that are common in experiencing Holist: spiritual tourism, authentic tourism, transformative travel, volunteer experiences, adventure tourism, nature-based sports, independent travels and religious tourism (Gnoth and Matteucci, 2014) – authors that based this inference with other literature papers (Cohen, 1979; Norman, 2012; Turner, 1973; Lean, 2012; Zahra and McIntosh, 2007; Duffy and Overholt, 2013; Humberstone, 2011; Wilson and Harris, 2006).
The fourth, and last, experience of the TEM is Knowledge Seeker – defined as “pre-meditated outcomes often again determined by society, and fulfilling such socially derived needs as esteem, authority, influence, or power, that come with the acquisition of new knowledge, financial wealth, and the image of having been to certain places, actually seen certain things, and learned from them in some form or another” (Gnoth and Matteucci, 2014:14). This experience is highly connected with some type of tourism categories, such as: museums tourism (McIntosh, 1999), cities tourism (Maitland and Ritchie, 2009), dark tourism (Biran and Hyde, 2013), rubbish-dump and slum tourism (Biran et al., 2011). Furthermore, other characteristics of this experience are the fact that tourists seek to visit galleries and shows, but never forgetting sports appearances. The emotional component also has an important role here as some tourists, dependent on the type of tourism chosen, seek for “a desire for emotional involvement and the construction of identities” (Gnoth and Matteucci, 2014:14).

2.5. SENTIMENT ANALYSIS
The advancements in the technological world have created the possibility to produce big flows of information (Big Data) and consume it at a speed higher than ever before. Nowadays, there is an interested in running these data through “machine learning methods in natural language processing” (Bilro, 2017:104). The same happens in the tourism industry – seen “as an industry where customer experience is crucial for its growth and reputation, has mainly adapted to the evolving technology and the availability of new data sources” (Alaei et al., 2017:2), meaning that tourists are nowadays able to access information from multiple sources, create their own content in multiple ways and spread it with the entire world at any time (e.g. social media), making them one of the most influencing sources to other travellers.
As we have seen earlier in Chapter 1.2, positive satisfaction is a variable half way through in converting possible travellers in real customers of a destination (Chang et al., 2009; Ribbink et al., 2004). Wang (2017) goes deeper, and explores the positive emotions demonstrated in user-content generated and states that in the tourism industry such emotions are crucial and destinations depend on it. Collecting the data of such vast travels thoughts and opinions was mostly done, in the past, through the form of interviews and surveys – which would most of the times, collect answers with positive bias (Dodds et al., 2015), and was very dispendious in time and budget.
To fight such expenditure of resources, researchers and firms started to develop new ways of storing data and analyse it, referring to it as Big Data analytics (Kirilenko et al, 2017).
One of those methods is called Sentiment Analysis, which consists in a “computational study of people’s opinions, appraisals, attitudes, and emotions toward entities, individuals, issues, events, topics and their attributes” (Liu and Zhang, 2012:1), and it aims at studying the polarity of a certain text script (e.g. review, blog text or any user-generated content), being able to characterize if such excerpt is positive, neutral or negative. It is quite useful in the tourism industry, because polarity may influence further travellers (Alaei et al., 2017). Sentiment analysis will also, through the selected text, be able to analyse if the excerpt is based on subjective or objective premises. For instance, subjective reviews have as ground opinions and stereotype bias – reviewers let themselves be influenced by their feelings, beliefs and opinions given by others, while objective ones are accurate and based on concrete proofs and determinate interpretations (Feldman, 2013).
The Figure 6, shown below, schematizes the, so-called, general process of a Text Mining operation. As it is possible to see, the process begins with the selection of data that will be used in the study, as well as its collection – generally, this information is provided from unstructured text sources (like reviews, blogs, forums, etc.). The process then continues with the processing of the data, through a certain algorithm, which will culminate in the final output with the pretended results (Shaikh and Deshpande, 2016).

As it was mentioned before company’s strategies can often be influenced by the data that they retrieve from user-content generated sources, and in this case Sentiment Analysis technique is capable of providing information “of their products or services influence on customers’ satisfaction, and which can help to create or change the focus of companies/brands’ strategies” (Bilro, 2017:104). Furthermore, according to Valdivia et al. (2017) TripAdvisor is a suitable source for such type of study, as it is a text source with a high number of reviews, covering firms for all over the globe. The reviews of such platforms, especially relating travels, are written with a good grammar (Kirilenko et al, 2017) and are usually tremendously positive – such specifications along with full sentences rather than short ones and use of slang or abbreviations are more suitable for this type of methodology (Lu and Stepchenkova, 2012).

3. METHODOLOGY
On this chapter there will be a full description of the methodology process that best responds to the research question. To do so it is intended to use a mixed approach composed by two different studies.
The first study (Study 1) will be performed under a netnographic perspective which intends to answer my questions merely on a qualitative level. The conducting process will be done through the scientific method of Robert V. Kozinets – considered the father of Netnography – and the process described in his book. Netnography is an already known and well-established approach to qualitative research, which combines together Internet and Ethnography, and can be used by multiple researchers and scholars, through online marketing research techniques, to best explain the free behaviour of individuals on diverse cultural worlds (Kozinets et al., 2014).
The second one (Study 2) will be done through the form of Text Mining – which is a process used by researchers and professionals to analyse big volume of stored information (Big Data) through “machine learning methods in natural language processing” (Bilro, 2017:104). There are multiple platforms that are able to do such analysis, but the one selected to this study is Meaning Cloud. A platform that can be used in Social Networking Sites, through Social Media Analytics – “leverages semantic technology to automatically “understand” the structure and meaning of news and social conversations (far beyond a simple aggregation of mentions) and extracts the most actionable bits of information” (LLC, 2018).
Through the analysis of both studies it is believed that it is possible to have a significant and meaningful result to answer to the research question – established in the upcoming chapters. Along with both techniques it will be used important scientific articles to compare with the information gathered from study 1 and study 2.
3.1. STUDY 1 – NETNOGRAPHY
Netnography as previously mentioned, results from a combination between Internet and Ethnography and its’ process results of a combination between the methodologies used in the study of cultural anthropology, applied to the online communities (Eastin et al., 2011).
According to Bowler (2010) stating Kozinets (2010) it follows six steps of ethnography: research planning, entrée, data collection, interpretation, ensuring ethical standards, and research representation.
Having in mind the previously mentioned book, it is possible to obtain a simplified flow of a netnographic research project that includes all of those six steps, which will be the procedure used from now on.

As it is possible to see in Figure 10, the netnographic research project used for Study 1 involves five different steps.
In Step 1 it will be defined the Research Question – as in the purpose of the project research, as well as which Social Sites are more suitable for conducting the experiment and a small amount of Topics that serve as guidance to the final objective. In Step 2 it will be done the identification of the communities that were chosen to be relevant to answer to the Research Question, along with a description of those communities. Regarding Step 3 it will be explained if in Study 1 it is required a Participation-Observation phase, and the different stages of the Data Collection. Furthermore, in Step 4 it will be explained how the information and results gathered will be represented (articles, books, reports, presentations, etc.), as well as explaining some details about the Data Analysis. Finally, in Step 5 it will refer the implications of this project as well as where the Report Findings will be presented.
There is in fact, another process that is described in Kozinets (2015) – see Annex 1 – but it is only meaningful for experiences that require a participant-observational phase, which means that the correct methodology for my research, which does not involve any participants is the one presented in Kozinets (2010).

3.1.1. STEP 1 – DEFINITION OF THE RESEARCH QUESTION, SOCIAL SITES OR TOPICS TO INVESTIGATE
According to Kozinets et al. (2014) research questions that are going to be studied through netnographic procedures may be related with a phenomenon that occurs both in offline and online worlds; that can be experienced solely virtually; or concerning the online world.
Furthermore, to identify online communities the researcher may address the investigation through online forums that will help answering the research question. It can be done through bulletin boards, chat rooms, play-spaces, virtual worlds, blogs, wikis, audio-visual sites, social content aggregators, and social networking sites (Kozinets et al. 2014).
In study 1 the best online forum to conduct the experience is social networking sites due to the huge variety of interactions and available communities it possesses. Having in mind Kozinets (2010), social networking sites are a very good example of combining web-pages, private emails, blogs, forums and chat rooms. Moreover, it allows “the access to unbelievable number of consumers, at low costs, high speed and ease of applicability” (Zaglia, 2013:222).
Additionally, Kozinets (2010) says that in order for a site to be considered suitable for a netnographic study, it must be relevant, active, interactive, substantial, heterogeneous and data-rich.
Since my research is focused in the tourism industry and the analysis of how tourists perceive two island destinations with identical culture but different demography (such as island dimension, world region and population size) the best suitable social networking site is TripAdvisor.
In TripAdvisor’s webpage it is possible to gather all the information that shows us why it is suitable for a study like this. It has more than 600 million reviews and opinions, and more than 7.5 million different accommodations, airlines, attractions and restaurants. Moreover, it is said to have more than 455 million monthly average unique visitors (TripAdvisor, 2017). According to Chua and Banerjee (2013) the majority of TripAdvisor reviews can be considered largely reliable.
As mentioned before since this dissertation’s goal is to study how tourists perceive two island destinations with identical culture but different demographic characteristics through social networking sites, the following topics will be addressed:
A. Overall analysis of the tourists perception of Island 1 experience according to TripAdvisor’s reviews;
B. Overall analysis of the tourists perception of Island 2 experience according to TripAdvisor’s reviews;
C. Comparison of both results;
3.1.2. STEP 2 – COMMUNITY IDENTIFICATION AND SELECTION
In order to choose communities to be studied, Kozinets et al. (2014) states that scholars and researchers must take into consideration communities that are more relevant to answer to the research question; have a higher number of user’s interactions; have a bigger number of reviews posted; more rich and useful data; large number of interactions between the users.
After researching island destinations across TripAdvisor’s website, having in mind the previously mentioned characteristics for communities, it was chosen the Madeira Island and the Bermuda Island to apply this netnographic study.
In order to best compare the information displayed in the next two sub-chapters – a brief characterization of both islands – a summary comparison table has been created (see Annex 2).
3.1.2.1. MADEIRA’S DESCRIPTION
The archipelago of Madeira is a Portuguese region located in the Atlantic Ocean with a total area of near 800 square kilometres, composed by Madeira Island, Porto Santo, “Ilhas Desertas” and “Selvagens”. Funchal is the capital of this archipelago where only the first two mentioned islands are habitable.
This region accounts with a population of 267 thousand people and has a GDP of 5.224 million Euros (Madeira, 2018).
It was an extraordinary year for the island’s tourism industry since it reached an historical maximum of 1.4 million visitors (RTP, 2018). In 2017 the number of overnight stays reached 7.5 million performing a total of 407.4 million Euros and lodging revenues of 263.6 million Euros. Moreover, the annual rate of bed occupancy stood at 69.7% – more 0.5% than the previous year. Another interesting data is the fact that the number of cruise transit passengers increased by 3.4% – totalizing 537 535 transit passengers (Direção Regional de Estatística da Madeira, 2017).
The position that it occupies geographically makes the region present a tropical climate, which can be seen by the average temperature of 19º Celsius and 25º Celsius during the summer months – June, July and August (Madeira, 2018).
Finally, it has gained a lot of awards in the former years, being the two most important ones: Best Island Destination by the renowned World Travel Awards and Mankind Heritage by UNESCO (Madeira, 2018).
Regarding TripAdvisor, Madeira’s presence in this platform is quite expressive – with 364 178 evaluations and opinions spread across five different categories: hotels, flights, activities, restaurants and forum. Additionally, visitors have placed more than one thousand photos in this social networking site, regarding their experiences (TripAdvisor, 2018).

3.1.2.2. BERMUDA’S DESCRIPTION
Bermuda is an archipelago located in the North Atlantic Ocean composed by 138 islands, where eight of the main ones are connected through bridges. From all of these islands, only a dozen of them are inhabitable. It is a self-governing British territory that has as capital the land of Hamilton.
This region has a total area of 54 square kilometres and accounts with a population of near 70 thousand people. It has a GDP of $5.20 billion (which exchanges to 4.231 million Euros) (WordAtlas, 2018).
According to the Bermuda Tourism Authority (2018), in their annual arrivals report, Bermuda’s economy was injected with $431 million through their visitors in 2017. It represents a 20% increase regarding the previous year. Additionally another important fact occurred once there was an increase by 9% in the hotel occupancy – reaching 63.1% – in comparison with 2016.
Moreover, 2017 was a remarkable year for the Bermuda’s tourism industry once it reached an historical record of 693 thousand visitors to the island. Furthermore, the Caribbean Tourism Organization (2018) revealed a total of 231 thousand overnight visitor arrivals – this data refers to the months of January until November (World Tourism Organization, 2017).
Another important aspect is the fact that there was an increase in the number of cruise passengers’ arrivals totalizing 418 thousand visitors – 20 thousand more than the previous year. By cruise is, since 2006, the preferable transportation for visitors to get to the island (Bermuda Tourism Authority, 2018).
The climate of Bermuda is very humid, influenced by the Gulf Stream but not actually a tropical paradise once in the winter the precipitation can be very high. The temperature varies from 16º Celsius in January and 30º Celsius during the summer months – June, July and especially August (Forbes, 2018).
Regarding TripAdvisor, Bermuda has also a considerable presence in the platform with 86 765 reviews spread across six categories: hotels, holiday rentals, flights, things to do, restaurants and forum. There are nearly 1900 photos displayed in these commentaries uploaded by the reviewers (TripAdvisor, 2018).

3.1.3. STEP 3 – COMMUNITY PARTICIPANT-OBSERVATION AND DATA COLLECTION
In this stage Kozinets (2010) adds a participation-observation phase to evaluate and takes notes of the community being studied online. In study 1 this stage is not required once there is enough information publicly exposed to answer my research question.
In consequence, this sub-chapter will only take into consideration the formulation of a strategy to collect the data.
In first place is necessary to mention that all the reviews being collected are withdrawn from the TripAdvisor website, and are aimed at responding to the research question. Secondly, due to the huge amount of reviews it is necessary to define the period-range of reviews being extracted – as seen before adding each island there are more than 450 thousand reviews.
Thirdly, identify the categories being studied as well as the number of elements in each category. Finally, and perhaps the most important, it is critical to establish the parameters in which the reviews are being studied.

3.1.3.1. DATA COLLECTION – PERIOD-RANGE OF REVIEWS
As seen in the chapter 1.1.2, there are nearly half a million opinions and reviews in the TripAdvisor platform regarding the islands of Madeira and Bermuda. In that sense, it is necessary to reduce this to a number in which the study can be carried on. To do so the information about the climate was considered – since both regions have the same summer months, it was chosen August of 2017 as the reference for the extraction of commentaries.

3.1.3.2. DATA COLLECTION – REVIEW EXTRACTION SPECIFICATION
Also seen in the chapter 1.1.2 there are different categories in the TripAdvisor website for reviewers to place their opinions, photos and comments. To identify the total amount of reviews being collected it was chosen the categories that the platform has in common for both destinations: hotels, restaurants and activities.
In this way, in order to extract a considerable number of reviews, it will be selected the 10 elements of each category with the most comments available in English (note that only comments with a correct spelling are selected) during the month of August of 2017.
The category of activities has a special character, since it is possible to divide it into two different types – Leisure (such as rock climbing and walking) and Cultural (such as visits to museums, churches, monuments and others). There are also Hotel and Restaurant.

In Annex 3 and 4 it is possible to find 8 schemes that represent the number of reviews to be taken from each hotel, restaurant, leisure activity and cultural activity. In the four categories there are a total of 117 724 reviews written in English – but only 1783 of those wore written during August of 2017 and in proper English – as shown in Figure 9.

3.1.3.3. DATA COLLECTION – CATEGORIZATION/CODING OF REVIEWS
Having established and explained the time frame of the extraction, as well as the number of categories and total number of reviews being analysed, the only step missing is the formulation of a strategy to study those reviews. According to previous research during the literature review, it was studied several authors and papers which are now important to the categorization, or coding, for the Kozinets (2010) process of studying the information.
In order to answer to the research question it will be assumed seven different codings in which the reviews will be analysed:
1. Rating: studying the rating of the reviews will allow the understanding of the overall satisfaction of each destination in the four mentioned categories, as well as serve as a comparison method between other strategies;
2. Categories: representing the place from where the review was extracted, in order to be able to understand overall which categories are most attractive for the reviewers;
3. High/Low membership level of reviewers: it is a factor that was already studied by other authors, such as Chen (2015) that concluded that low membership levels will accept the information with a higher probability, while high membership levels will promote the growth of low membership levels. Hereupon, it is important to study the number of high vs low membership reviewers. TripAdvisor platform uses scale from 0 to 6 to classify the reviewers, and in this case the following scale will be used: 0-3 Low Ranked Reviewer and 4-6 High Ranked Reviewer;
4. Language Type: according to Wu et al. (2017:590) “consumers actually exhibit less favourable attitudes and lower reservation intention after reading a figurative (vs literal) review posted by a low expertise level” while being high expertise reviewer attenuates this situation; Therefore, it is important to compare information between language type vs high/low membership level in each destination; For this matter if the review contains slang, any kind of Symbology (see in point 6), wrong use of Punctuation, or any other aspect that it is considered relevant, the language will be considered as Figurative. In any other case, it will be classified as Literal;
5. Tourism Experience Model (TEM): this framework was created by Juergen Gnoth and that can be found in Gnoth and Matteucci (2014) paper. It aims at providing future directions in the holiday tourism research. The model allows destinations to position themselves as to how they seek to serve the tourist – it permits visualizing where the destination fits according to the tourist’ experiences. It consists in two axes (Consciousness and Activity). The first one is related to how the tourists perceive their experience. In one end, the model presents the Person – it is based on the authenticity reflected in socially accepted role-performances. The other end of the dimension is related to the Human Being – finding ourselves as a human being in order to get close to our existential being. The second one works on the premise that tourists leave their homes in order to experience, which means that they are always active. On one end, there is the Recreational Activities, which are entertaining experiences merely done from the habit, training and repetition of the tourists’ lives. The other end of the axis brings us Exploratory Activities, where tourists seek new insights, understandings, social and bodily feelings, mainly characterized by challenges, learning and knowledge. There are 4 different tourism experience perspectives: Re-Discover, Pure Pleasure, Holist and Knowledge Seeker. The first two are connected with Experiencing as being, while the other two are related with Experiencing as becoming. In Annex 5, it is possible to find the axial figure representing the model as well as the parameters which allow distinguishing the tourists’ experiences;
6. Content: it will be used to be able to understand what type of subjects and words are relevant for the reviewers, in a global way – to do so the platform Wordle will be used as an output of the final results;
7. Symbology: in here it will be studied the emojis and emoticons used by the reviewers – in each case trying to understand if it has a positive or negative connotation;
8. Positive/Negative information: another important aspect considering the perception of each tourist in each destination is the positive/negative information the reviewers express in TripAdvisor – as well as compare this information with others already mentioned. According to Chen (2015), “positive information encourages consumers to buy products and negative information presents the problem of products, which will reduce its’ credibility” – it is necessary to have in consideration that travels are considered hedonic products. As well as the Content code, the platform Wordle will be used as an output of the final results;
These are the seven strategies established in order to understand what is the perception of tourists regarding the experiences lived in Madeira and Bermuda.

3.1.4. STEP 4 – DATA ANALYSIS AND INTERACTIVE INTERPRETATION OF FINDINGS
According to Kozinets (2010) this phase is the one where the data collected is transformed into a research representation such as, articles, books, presentations or reports.
Since this is a master thesis dissertation, the representation of the findings will be done through the form of a written report and an oral presentation afterwards.
The data for the overall analysis of both islands was collected from November of 2017 until February of 2018, and as it was mentioned before, the reviews are evaluated according to their rating, reviewer’s membership level, language type, tourism experienced, content, symbology and positive/negative information.
The comparison between each island is done through the form of a qualitative interpretation of the findings, having in consideration the results presented for each destination.

3.1.5. STEP 5 – WRITE, PRESENT AND REPORT RESEARCH FINDINGS AND/OR THEORETICAL AND POLICY IMPLICATIONS
As it was just mentioned in the previous chapter, this dissertation is a partial requirement for the conferral of the Master in Marketing at ISCTE Business School, and in this sense it will be conducted a written dissertation thesis, as well as an oral presentation where the results are explained.
Both theoretical and policy/managerial implications will be exposed in the chapter of Conclusion – and are a consequence of an interpretation along with the results shown in the respective chapter.

3.2. STUDY 2 – TEXT MINING
The conventional definition of Text Mining states that it “is the discovery by computer of new, previously unknown information, by automatically extracting information from different written resources. A key element is the linking together of the extracted information together to form new facts or new hypotheses to be explored further by more conventional means of experimentation” (Hearst, 2003). Its appearance was due to the increase in large-scale of information being stored in a given database (such as online information), given birth to the so called Big Data.
The Text Mining process results of a technologic development that is capable of analysing big sequences of unstructured text, such as the ones found stored online, and recognize important phrases and the relationships between them (Fan et al., 2006). This system allows manipulating those sequences of text through categorization, clustering and studying the sentiment attributed by the ones who wrote it (Srivastava et al., 2009).
Internet technology has given power to people that did not feel they had that before (Alaei et al., 2017), which is reflected in the travel tourism industry as travellers commonly create e-WOM through social networking sites as Facebook, Instagram, TripAdvisor and others. These users are not just looking to improve their travelling wisdom, but also to collect experiences, feedbacks and recommendations from fellow travellers (Ye et al., 2009). It leads to a large volume of information being available and stored online at an outstanding speed that can be used by scholars and researchers to study the attitudes and experiences of tourists (Alaei et al., 2017).
As it was mentioned before in subchapter 1.1.1, the main goal of this dissertation is to understand how tourists perceive two island destinations with identical culture but different demographic characteristics through social networking sites, and it is believed that a Text Mining processes might be useful to answer to it, as it allows reading and examining unstructured text – as the ones presented in online reviews.

3.2.1. STEP 1 – DEFINITION OF A TEXT MINING TOOL
According to Spinakis and Chatzimakri (2005) there was an increase of the diversity of text mining tools and techniques that researchers and scholars now possess in order to perform text mining analysis.
The literature was able to identify three categories of text mining tools (Figure 10): Proprietary Text Mining Tools – which are tools that companies or firms own, and it is mandatory a certain fee in order to use them; Open Source Text Mining Tools – tools that can be obtained at zero cost and used at free will; and Online Text Mining Tools – which are programmes that can be managed from their own website, but commonly their functionality is partial (Kaur and Chopra, 2016).

Along with two experienced specialists in the field of the Marketing area, it was decided that the best tool to obtain real results towards the research objective is Meaning Cloud (a Proprietary Text Mining Tool) – which is a Text Analytics platform that can be downloaded and added to the Excel.
Meaning Cloud is described by the company as the “easiest, most powerful and most affordable way to extract the meaning of all kind of unstructured content: social conversation, articles and documents” (LLC, 2018). It is a tool commonly used by scholars and academics that allows extracting the texts sentimental polarity, select a particular set of text and sentimentally study it, text clustering and theme and topics identification (Bilro, 2017).
The main features of the software outcome from a complex coding that allows, for example, users to use their own linguistic dictionary, easily incorporate the software into Excel with a simple installation, and availability to work in numerous languages (Portuguese, Italian, English, French or Spanish). It also allows the analysis of unstructured material from almost every source of information (Emails, Inquiries and Surveys, Social Networking Sites, and others) (LLC, 2018).
It is this last piece of information that makes the software suitable to help answering to the research question, once Social Networking Sites are the key solution to the main objective in this dissertation. These social sites (such as blogs, online forums or even newspapers) do not present a clear text structure, most of the data is dispersed or the text is not orderly. Meaning Cloud is capable of using Social Media Analytics, a tool able to “understand the structure and meaning of news and social conversations (far beyond a simple aggregation of mentions) and extracts the most actionable bits of information. All this returns more consistent results than human analysis at infinitely greater speed and volumes” (LLC, 2018).
According to its’ website, Social Media Analytics is suitable for finding discussions between customers in social media, supervise those same conversations by discovering threats to the company, information to do the most varied analysis and trace users’ behaviour (LLC, 2018).

3.2.2. STEP 2 – RESEARCH SOCIAL SITES, COMMUNITY SELECTION AND SAMPLE IDENTIFICATION
In this chapter it will be identified the Social Networking Sites to be used in the experience, as well as the type of analysis that it will be used with the Meaning Cloud software.
In chapter 1.1.1 (Step 1) it was identified the characteristics of TripAdvisor on being the best suitable social networking site focused in the tourism industry capable of providing data to, together with literature and the analysis performed, answer to the research question. As Study 2 serves as a complement to Study 1, it only makes sense that in order to provide the most complete and plausible response towards understanding how tourists perceive two island destinations with identical culture but different demographic characteristics through social networking sites, that not only Study 2 is performed under reviews extracted from TripAdvisor, but also maintain the same sample extracted in Study 1 (Chapter 1.1.3.1 and 1.1.3.2) so that the analysis performed are done under the same patterns and conditions of text.
Therefore, the only aspect worth investigating in this chapter is, understanding if TripAdvisor is a suitable platform to work under the Meaning Cloud software. As it was already seen, in the Literature Review Chapter and in the previous Chapter 1.2, a Text Mining process is able to analyse big sequences of text classified as unstructured, such as the ones found online (Fan et al., 2006), also “millions of visitors exchange content on popular platforms for mutual benefit, such as social networking (e.g., Facebook), content sharing (e.g., Reddit), blogging (e.g., LiveJournal), micro-blogging (e.g., Twitter), multimedia sharing (e.g., YouTube), location sharing (e.g., FourSquare), review forums (e.g., TripAdvisor), and other sites” (Kirilenko et al., 2017:1), which enlarges the creation of e-WOM (Electronic Word-of-Mouth) (Confente, 2015), mainly in an unstructured way.
In order to analyse such big volume of information in the tourism industry, on the platforms mentioned before, new techniques had to be developed and literature shows the importance of text mining tools in studying social networking sites data, defining that such applications resulted in huge developments in collecting, cleaning, processing and evaluating data (Alaei et al., 2017) of the e-WOM Social Media data gathered (Hippner and Rentzmann, 2006; Schmunk et al., 2014).
Finally, having in mind the literature already mentioned that defines TripAdvisor has one of the platforms where it is possible to find e-WOM, as well as a large amount of data volume to extract, being that data considered as unstructured and therefore suitable for Text Mining tools (as Meaning Cloud), it is possible to conclude that the sample extracted to perform the Netnographic study (Study 1) – Chapter 1.1.3.1 and 1.1.3.2, can also be used to perform a Text Mining analysis (Study 2).

3.2.3. STEP 3 – DATA COLLECTION, PROCESS AND ANALYSIS
In this chapter it will be mentioned the characteristics of the sample collected – something already done during Chapter 1.1.3.1 and 1.1.3.2, but this time on a briefer way. As well as, the type of analysis that will be performed under the Meaning Cloud scope, together with how the results will be worked on, and displayed to this dissertation.
Regarding the sample, as seen before, during the months of November, December and January, it was collected 1783 reviews from TripAdvisor in order to analyse them with the ultimate objective of collecting enough information to properly answer to the research question.
The reviews that were extracted relate to the month of August of 2017 and were extracted from the ten most reviewed elements in TripAdvisor, according to four different categories – Hotels, Leisure Activities, Cultural Activities and Restaurants. Based on those premises it was possible to collect 1148 reviews to analyse the Madeira Island – 217 from Hotels, 309 from Restaurants, 163 from Leisure Activities and 459 from Cultural Activities, and 635 reviews to study the Bermuda Island – 127 from Hotels, 136 from Restaurants, 214 from Leisure Activities and 158 from Cultural Activities. Moreover, in Chapter 1.1.3.3 the it is explained the creation of the strategies and the coding in which the reviews would be studied – Rating, Categories, High/Low Membership Level of Reviewers, Language Type, Tourism Experience Model (TEM), Content, Symbology and Positive/Negative Information.
For the Text Mining process only three of the codings were selected to be studied: Categories, Tourism Experience Model and High/Low Membership Level of Reviewers – see Figure 11 shown below.

The three codings, selected to be studied, outcome from a discussion of ideas with the thesis supervisor, since studying all the possible codings created in the database through a Text Mining process would be a very long process, and only those three were considered. The rest of the codings may be studied in future research.
The next step was to define the type of tests and analysis that would be done with the Meaning Cloud software. Along with a specialist in the field it was decided to perform three types of tests: Text Classification, Topics Extraction and Sentiment Analysis. All of these processes were performed via Excel and done for both island’s reviews. A new and better framework can now be developed, and it will be on those premises that the Text Mining analysis will be focused on – see Figure 12, below.

The first one – Text Classification, is capable of analysing big segments of text from social media reviews, creating several categories or groups, to more easily understand which are the themes the reviewers are mentioning the most. Those groups are built based on the content of each review and are a result of a complex algorithmic analysis (LLC, 2018). This study is performed both for Madeira and Bermuda, and the results will be analysed in four different chapters, according to two different perspectives:
1. Global Text Classification Results: from all the text groups presented as an output according to the Meaning Cloud, the results will be worked with the help of a specialist in the field, and select those that have the highest/most significant presence in relation to all the others;
2. Text Classification applied to TEM/High Low Membership Level of Reviewers/Categories: this type of analysis will be applied to each of the three coding selected, and through the Excel the text groups will be worked on so that they can be presented in the dissertation, again according to its’ significance in comparison with the others.
It is important to refer that each coding is different, and there is not a percentage that defines the ideal significance, which means that it will result of an opinion-based analysis. In case of existing reviews that the software is not able of identifying the correspondent text groups or labels, the same shall not be considered but still worth mentioning in the results section.
The second analysis is a Topics Extraction one. This type of analysis performed by Meaning Cloud goes through each sentence of each segment of text presented – in this case goes through each sentence in each review, and extracts element by element its meaning. The elements can be: concepts, entities, time or money expressions, quantifiable expressions, relations and quotes (LLC, 2018). Again, this process comes from a complex algorithmic analysis, and the same procedure regarding the previous analysis will be applied:
1. Global Topics Extraction Results: in this analysis, due to the amount of reviews analysed, there will be an extensive list of topics being extracted by Meaning Cloud, meaning that it will be impractical to study them. In this sense, it will only be presented the 20 concepts or entities with the higher significance (the ones that the software identified more times);
2. Topics Extraction applied to TEM/High Low Membership Level of Reviewers/Categories: again, this type of analysis will be applied to each of the three coding selected, and through Excel the results will be worked on to be presented. Due to the high amount of elements being displayed by the Meaning Cloud outputs, the amount of elements being mention will differ from coding to coding, to a maximum of 20 elements.
The third and final analysis is named Sentiment Analysis, and as seen in the Literature chapter it is a “computational study of people’s opinions, appraisals, attitudes, and emotions toward entities, individuals, issues, events, topics and their attributes” (Liu and Zhang, 2012:1). It performs a “detailed multilingual sentiment analysis for texts from different sources (…) identifies the positive, negative, neutral polarity in any text, including comments in surveys and social media” (LLC, 2018). The Sentiment Analysis results in two different studies: a Global Sentiment Analysis and a Topics Sentiment Analysis.
The first one – Global Sentiment Analysis aims at studying several aspects:
1. Polarity: the first step is to transform the outputs in a quantitative level: 1 (N+) – strong negative; 2 (N) – negative; 3 (Neu or NONE) – neutral; 4 (P) – positive; 5 (P+) – strong positive; according to the review that it refers to. Also, it will be studied the number of reviews according to each polarity and its’ average;
2. Agreement: it analyses if the sentiment between the different elements in each review is in accordance (agreement) or not (disagreement). It will be counted the number of reviews in agreement/disagreement, as well as the average polarity of each type of reviews;
3. Subjectivity: studies if the reviews were influenced by prior bias or stereotypes based on the reviewer’s personal feelings, taste or opinions. Again, it will be analysed the number of reviews subjective/objective, as well as the average polarity of each type of reviews;
4. Irony: it analyses the degree in which reviewers were being ironic or not with their opinions. It shall be accounted the number of ironic/non-ironic reviews, as well as the average polarity of each of those types of reviews.
The four enlisted steps above mentioned shall be studied in a global and general way, and applied to each of the three coding selected.
The second one – Topics Sentiment Analysis studies the polarity of the elements (concepts or entities) that it refers to, grouping those forms of text according to their types. Due to the high amount of types that may appear, because of the very high amount of concepts being analysed, with the help of a specialist in the field, it will be developed a strategy to decrease them, and only 21 text groups will be analysed in the following way:
1. Analyse the average polarity of the final 21 text groups;
2. Analyse the average polarity of each coding according to the 21 text groups existent;
3. Analyse the average polarity of each text group according to each coding.
4. RESULTS
In this chapter it is located all the compilation of data collected as well as the data interpretation of both island destinations: Madeira and Bermuda. The collection of the data was based on the utility that it could have to answer to the research question. Study 1 is based on a netnographic process by Kozinets (2010) and Study 2 is based on a text mining process, using Meaning Cloud, in order to obtain a more rigorous analysis – Text and Topics Extraction, as well as Sentiment Analysis.
The results will be displayed and analysed in the following order: 1) Analysis of both netnographic studies regarding the islands of Bermuda and Madeira; 2) Analysis of the text mining results regarding both destinations, as well.
• Note 1: for rounded values, two decimal places will be used throughout the chapter.
• Note 2: due to having limited amount of pages to consider in this dissertation, the results regarding Study 1 (Symbology, Content, Positive/Negative Information) – Annexes 6.11, 6.12, 6.13, 6.14, 7.11, 7.12, 7.13 and 7.14 and Study 2 (Topics Extraction and Topics Sentiment Analysis) – Annexes 8.5, 8.6, 8.7, 8.8, 8.28, 8.29, 8.30, 8.31, 9.5, 9.6, 9.7, 9.8, 9.28, 9.29, .30 and 9.30 will only be presented in the annexes, not in this chapter.

4.1. RESULTS STUDY 1 – NETNOGRAPHY
As mentioned before, the data collection of the netnographic process was made having in consideration 8 different codes: rating, categories, high/low membership level of reviewers, language type, Tourism Experience Model, content, symbology and positive/negative information.
The following results were extracted from the database of reviews created through the elaboration of pivot tables.

4.1.1. RESULTS STUDY 1 – MADEIRA
An overview over the dataset shows us that 1148 reviews were collected in the TripAdvisor platform regarding the opinions and evaluations of tourists in August of 2017 in Madeira. From those, 217 were taken from hotels, 309 from restaurants, 163 from leisure activities and 459 from cultural activities – see Annex 6.1. Through this information it is immediately possible to draw a reading that the tourism of Madeira has a higher preponderance of Cultural Activities than Leisure Activities – since there is a considerable discrepancy between those two numbers of reviews.

4.1.1.1. GENERAL RATING AND CATEGORIES
Keeping on a general level, the overall average rating of reviews was 4.59 (from 0 to 5), which is a very good rating level considering the total amount of reviews, and shows us that overall the reviewers are extremely satisfied regarding the activities and establishments that they attended during their stay. Moreover, in Annex 6.2, it is shown the average rating of each hotel, restaurant, leisure activity and cultural activity extracted to this part of the research.
Only 3 items have achieved the maximum rating average – Madeira Fauna & Flora, True Spirit and Venture Nature Emotions – Day Tour, and curiously all of them belong to the category of leisure activities. It is also easily seen that the two lowest average ratings belong to Mercado dos Lavradores (3.53) and Palheiro Gardens (3.57), both of them Cultural Activities.
This information takes the research to study the average rating per categories in general, as can be seen in Annex 6.3. Although all the categories have a very high average rating (for example, none of them have an average rating inferior to 4), which shows us the already mentioned general satisfaction of the reviewers, it can be concluded that there are two categories in which the reviewers present a higher satisfaction: leisure activities (4.81) and restaurants (4.77).

4.1.1.2. MEMBERSHIP LEVEL OF REVIEWERS
Regarding the Membership Level of Reviewers there is a majority of Low Ranked Reviewers (LRR) versus High Ranked Reviewers (HRR) – 61.15% LRR (702 reviews) and 38.85% HRR (446 reviews). This information can be consulted in Annex 6.4.
To further compare the differences and similarities between LRR and HRR, it is important to study how the two types of reviewers rate their experiences in the island – Annex 6.5. It shows the average rating of reviews according to the membership level of reviewers – HRR (4.53) and LRR (4.63), which shows a slight more satisfaction on the LRR experiences in the island, in comparison with the HRR.

4.1.1.3. LANGUAGE TYPE
Another aspect agreed to study how tourists perceived the touristic island of Madeira was the type of language used by reviewers to express their opinions and commentaries regarding their experiences – in this case literal language vs figurative language. In Annex 6.6, it is shown that there is a clear difference between the numbers of reviews written in literal language versus figurative language. The literal language is the most used by the reviewers with 910 reviews (representing 79.27% of the total sample) vis-a-vis the figurative one with only 238 reviews (representing the other 20.73% of the total sample).
The average rating of people with the different written language type is also not significant, since both have very similar values – 4.56 for the figurative language and 4.60 for the literal one – see Annex 6.7.
The dataset also allows to combine these two last codes, that can be perceived as important to answer to the research question, and obtain the number of reviews using literal/figurative language according to the membership level of the reviewers – see Annex 6.8. In this case the results show that 59.78% of the literal reviews are expressed by the LRR, while the other 40.22% by the HRR. The case is slightly different in the figurative language type, since 33.61% of these were wrote by HRR, while the rest (66.39%) were made by LRR.
4.1.1.4. TOURISM EXPERIENCE MODEL
Once studied the average general rating, the categories, as well as the membership level and the language type, it is now time to understand what experiences were lived by the tourists according to the chosen model – Tourism Experience Model (TEM).
As explained during the Methodology, the TEM is a model that allows destinations to position themselves as to how they seek to serve the tourist – or in another words, it permits a graphic representation of the destination according to the tourist’s experiences.
In Annex 6.9, it is shown the number of experiences lived by tourists during their stays in the island. It is clear that there are two experiences that are undoubtedly more present in their stays: Re-Discover and Pure Pleasure. From the 1148 reviews extracted, 595 are identified as experiencing Re-Discover, 379 perceived as experiencing Pure Pleasure, 138 experiencing Knowledge Seeker and only 36 perceived as experiencing Becoming. In terms of percentage, the results show that experiencing Re-Discover represents more than half of the reviews (51.83%), Pure Pleasure nearly 33.01%, Knowledge Seeker representing 12.02% and finally Holist indicating 3.14%.
According to Annex 6.10, the Holist experience is the one which the reviewers consider as the most satisfying (4.78 out of 5), followed by Knowledge Seeker (4.62) and Re-Discover and Pure Pleasure (both with 4.58). All in all, only the Holist experience is evaluated a little higher than the others, while the rest does not show a considerable difference – all of them are very close and represent the general satisfaction seen previously by the reviewers.

4.1.2. RESULTS STUDY 1 – BERMUDA
Studying the dataset obtained from the TripAdvisor platform it is possible to see that it was obtained 635 reviews regarding the opinions and evaluations of tourists in August of 2017 in Bermuda. From those, 214 were taken from leisure activities, 158 from cultural activities, 127 from hotels and 136 from restaurants – see Annex 7.1.
Contrary to what was concluded in the previous chapter, the results in the Bermuda Island show that there is slight preponderance for tourists to attend more leisure activities than cultural ones – although the gap between them is smaller than in Madeira.

4.1.2.1. GENERAL RATING AND CATEGORIES
The overall average rating of reviews was 4.43 (from 0 to 5), which is ultimately a very good rating level for such significant amount of reviews, and tells us that most part of the reviewers consider as very satisfying their experience in this destination.
In Annex 7.2 it is shown the average rating of each category item, meaning the average rating of every hotel, restaurant, leisure activity and cultural activity analysed. It is possible to see that there are several items with the maximum average rating – which can be explained due to the high satisfaction towards that item or the low number of reviews extracted from it. For example, Fourways Inn presents an average rate of 5 but only one review was considered suitable in August of 2017 to be extracted – being too skewed, it is not possible to draw any conclusions. Additionally, Bermuda Fun Golf, Hartley’s Undersea Walk, KS Watersports Tours & Rentals and Mad Hatters also present an average rating of 5 – and here it is possible to say that there is a very high satisfaction regarding leisure activities (since the first three correspond to this type of activity).
On the opposite way, Barracuda Grill, Hog Penny, Pickled Onion, Portofino Restaurant, Elbow Beach Bermuda and Royal Naval Dockyard are the items with the lowest average rating – considered the ones with an average rating below 4. From these, four of them correspond to restaurants, one to hotel and one to cultural activity.
Moreover, in Annex 7.3 it is shown that even though all the categories have an average rating above 4 (out of 5), which is extremely positive and demonstrates the general satisfaction felt by the tourists in this destination, the leisure activities are the ones that satisfy the reviewers the most – with an average rating of 4.71. The second category with the highest average rating is hotels (4.52), followed by the cultural activities (4.23) and finally restaurants (4.13).
From this information it is possible to outline that the leisure activities and the hotels are considered the most satisfying and relevant categories in the tourists’ holidays.

4.1.2.2. MEMBERSHIP LEVEL OF REVIEWERS
Concerning the Membership Level of Reviewers, in the database set to study the island of Bermuda, there is a majority of LRR (360 reviews) compared to HRR (275) – it leads to a percentage of 56.69% and 43.31% respectively – Annex 7.4.
Moreover, in Annex 7.5 it is shown the average rating of reviews per Membership Level, or in other words, the average rating given by the two types of reviewers in the experiences/activities they had in the island.
In view of such results, it can be seen that there is a small difference between how both membership level reviewers evaluate their experiences in Bermuda. LRR evaluate slightly higher (4.56) when compared to HRR (4.24) – a difference of 0.32 amongst each other. It is fair to say then, that LRR tend to higher evaluate their experiences in the island – having in consideration the reasonable amount of reviews extracted.
4.1.2.3. LANGUAGE TYPE
As seen before, another aspect that was studied was the type of language used by the different reviewers in order to express their feelings and opinions – literal vs figurative one.
In Annex 7.6, it can be seen that there is a massive difference between the numbers of reviewers using literal vs figurative language type. 80.31% (510 reviews) of reviewers used literal language to write their commentaries and opinions regarding Bermuda, while only 19.69% (125 reviews) were written using a figurative context.
The average rating of both language types is quite similar – and can be found in Annex 7.7. The reviewers using literal language type gave an average rating of 4.41 and the ones using figurative language were slightly more satisfied (4.50). Nonetheless, the results are very close and show the reviewers’ satisfaction already mentioned.
As seen in the previous subchapter, the dataset permits to combine the two last mentioned codes: membership level of reviewers and literal/figurative language used by reviewers. In Annex 7.8, it is shown a graphic representation of the number of reviews using a specific language type according to the membership level of the reviewers.
The results show that from the 275 HRR – 82.18% are using literal language to express their opinions regarding Bermuda (corresponding to 226 reviews), while only 17.82% (49 reviews) were written in a figurative type of way. Regarding the LRR the case is quite similar – 78.89% of the reviewers (284 reviews) used literal language, while only 21.11% (76 reviews) wrote their commentaries using figurative language.

4.1.2.4. TOURISM EXPERIENCE MODEL
After studying the number of reviews, rating, membership level of reviewers and the language used by the reviewers, it is time to understand what experiences were lived by the tourists that visited Bermuda and expressed their thoughts and feelings in TripAdvisor about it – using the Tourism Experience Model (TEM).
The number of experiences, according to the TEM is shown in Annex 7.9. From the graphic representation it is clear that the Re-Discover experience is undoubtedly the one most handled by the reviewers – corresponding to 392 reviews (61.73%) of the total number of experiences. The rest was identified as: 135 perceived as experiencing Knowledge Seeker (21.26%), 90 perceived as experiencing Pure Pleasure (14.17%) and 18 as experiencing Holist (2.84%).
Furthermore, in Annex 7.10, it is possible to see that the Holist experience is the one that makes reviewers evaluate with the higher rating (4.72 out of 5), while the three categories remaining present a very close average satisfaction rating: Pure Pleasure (4.49), Re-Discover (4.41) and Knowledge Seeker (4.40). Being noticed that, the Holist experience is completely detached from the rest.

4.2. RESULTS STUDY 2 – TEXT MINING
As previously mentioned in the methodology chapter, a text mining study was conducted in order to have a clear and more objective analysis of what themes, concepts and words are being used by the reviewers in their opinions. The text mining is performed by software that can be added to Excel, called Meaning Cloud.
The text mining process, in this study, consists in 3 different types of analysis:
1. Text Classification analysis – which will give us a certain number of labels (Themes) that are being exposed online by the TripAdvisor users in their reviews;
2. Topics Extraction analysis – which is more detailed than the previous study, and results in collecting a series of entities and concepts being mentioned by the reviewers (presented in Annexes);
3. Sentiment Analysis – which results in two separate studies: a Global Sentiment Analysis and a Topics Sentiment Analysis (presented in Annexes); the first one allows us to study the polarity, agreement, subjectivity, confidence and irony of the reviews; the second one is more related to being able to study the average polarity according to the entities and concepts mentioned by the reviewers.

4.2.1. RESULTS STUDY 2 – MADEIRA

4.2.1.1. RESULTS STUDY 2 – TEXT CLASSIFICATION

4.2.1.1.1. GENERAL TEXT CLASSIFICATION RESULTS
The results of the general data regarding the analysis of the Text Classification on Madeira’s extracted reviews indicate a wide variety of themes being mentioned: Lifestyle and leisure, Tourism, travel and commuting, Art and culture, Environment, weather and energy, Social issue, Economy and Finances, Sport, Science and Technology, Crime, law and justice, Disaster and accident, Labour, Health, Politics, Greeting and Thanks, Education, Unrest, conflict and war and Religion and belief.
Some of these themes have no significant expression and for that reason, only a few of them were considered as critical for the results. The most important ones were gathered and can be found in Annex 8.1.
As it is easily seen, according to the different amount of reviews in the themes that were identified, it can be created three separate groups.
The first group contains the themes with the highest amount of reviews, and it includes Lifestyle and Leisure present in 379 reviews and Tourism, travel and commuting counting 285. Together this group is responsible for 664 reviews (57.84% of the total amount of reviews).
The second group of themes comprises also two subjects: Art and Culture accounting 91 reviews and Environment, weather and energy found in 88 reviews. Combined both themes are responsible for 179 reviews (15.60% of the total amount of reviews).
The third, and final group, is composed by the themes with the smallest amount of reviews and it embraces three distinct subjects: Social that was found in 46 reviews, Finances totalling 46 reviews as well and Sports with 40 reviews. This last group, summing all the reviews, gathers 132 reviews (11.50%).
As it can be summed up, the total amount of percentage of the three groups formed to simplify the analysis is 84.94%. It is also important to refer that 119 out of the 1148 reviews extracted were unable to be analysed by the software, which ultimately results in 10.37% of the reviews not being properly studied. This information leads to a total of 95.31%, being that the remaining percentage is distributed across all the other themes mentioned in the first paragraph.

4.2.1.1.2. TEXT CLASSIFICATION APPLIED TO TOURISM EXPERIENCE MODEL
It is possible to distinguish the themes most mentioned by the reviewers according to the experiences that they lived during their stays in the island. The results can be found in Annex 8.2
In that table are shown the themes that have a bigger significance in each experience. For instance the Holist experience contains two major themes: Environment, weather and energy counting 13 reviews and Tourism, travel and commuting with 11 reviews. Together they are responsible for 66.70% of the total amount of reviews, being the rest spread across several other themes.
The Knowledge Seeker experience presents four major themes on the final results. Lifestyle and leisure with 35 reviews, as being the theme most significant, followed by Tourism, travel and commuting counting 33 reviews. Art and culture comes next summing 29 reviews, and finally Environment, weather and energy totalling 15 reviews. Summing it all up, it leads to a total of 81.16% of the total amount of reviews.
The Pure Pleasure identified reviews comprise 5 major themes: Lifestyle and leisure being by far, the most important one with 195 reviews, followed by Tourism, travel and commuting with 54 and Art and culture with 26. Finances and Social come next with 15 and 13 reviews, respectively. Together it sums up to 80.00% of the total amount of reviews.
The Re-Discover experience is the one that comprises the most variety of themes with significant relevance. Tourism, travel and commuting is the highest one with 187 reviews, followed by Lifestyle and leisure totalizing 144 reviews. After those appears Environment, weather and energy with 51 reviews and Art and culture counting 34. Finances and Social both show the same amount of reviews – 30, and lastly Sports totalling 25 reviews. These themes are responsible for 84.20% of the total amount of reviews.

4.2.1.1.3. TEXT CLASSIFICATION APPLIED TO CATEGORIES
It is possible to distinguish the themes most mentioned by the reviewers according to the categories of each review when extracted. The results can be found in Annex 8.3.
The reviews extracted from the Cultural Activities comprise six major themes: Tourism, travel and commuting as the one with the highest amount of reviews (158), Environment, weather and energy with 64 reviews, Art and culture counting 45 reviews, Lifestyle and leisure summing 39 reviews, Sports totalizing 26 reviews and finally, Finances with 23 reviews. Summing it all up it corresponds to 77.34% of the total amount of reviews.
The Hotel category shows a far less variety of themes, with only two subjects being significant to be mentioned: Lifestyle and leisure with 159 reviews and Tourism, travel and commuting counting 37 reviews. It leads to a total of 90.32% of the total amount of reviews.
The Leisure Activities appear in this table with three major themes. Tourism, travel and commuting as being the most important one, with 71 reviews, followed by Lifestyle and leisure with 48 reviews and finally, Environment, weather and energy totalling 15 reviews. Together these themes are responsible for 82.21% of the total amount of reviews.
The Restaurant extracted reviews include as most important themes: Lifestyle and leisure with 133 reviews, Art and culture accounting 34, Finances totalizing 22, Tourism, travel and commuting summing 19 reviews and finally Social including 18 reviews. Those are responsible for 73.14% of the total amount of reviews.

4.2.1.1.4. TEXT CLASSIFICATION APPLIED TO MEMBERSHIP LEVEL OF REVIEWERS
In Annex 8.4 it is possible to find the results of the Text Classification analysis according to the Membership Level of Reviewers code.
The High Ranked Reviewers in their opinions and thoughts regarding Madeira are mentioning four big themes, which can be separate into two different groups. The first group comprising the two themes with the highest amounts of reviews, being: Lifestyle and leisure with 141 reviews, followed by Tourism, travel and commuting counting 111 reviews. The second group containing: Environment, weather and energy with 40 reviews and Art and culture totalling 36 reviews. These themes lead to 73.54% of the total amount of reviews.
The Low Ranked Reviewers include six major themes that can also be separate according to the number of reviews, but this time in three separate groups. The first group, being the one with the themes most mentioned, includes Lifestyle and leisure with 238 reviews and Tourism, travel and commuting with 174 reviews. The second group comprises Art and culture with 55 reviews followed by Environment, weather and energy with 48 reviews. Finally, the third group embraces Social with 32 reviews and Finances totalling 31 reviews. Together, these are responsible for 82.34% of the total amount of reviews.
4.2.1.3. RESULTS STUDY 2 – GLOBAL SENTIMENT ANALYSIS
The Sentiment Analysis results in two different studies: a Global Sentiment Analysis and a Topics Sentiment Analysis. As mentioned before the first one studies subjects like the polarity, agreement, subjectivity, confidence and irony of the reviews, while the second one is more related with studying what topics were mentioned by the reviewers, along with understanding the polarity of the topics.

4.2.1.3.1. GLOBAL SENTIMENT ANALYSIS GENERAL RESULTS
The first thing that it was chosen to study, is the polarity – this corresponds to a positive/negative sentiment of the element it refers to, in this case it is referring to the amount of reviews itself and can be found in Annex 8.9. The scale was constructed in the following way: 1 – strong negative; 2 – negative; 3 – neutral; 4 – positive; 5 – strong positive.
The results show that there is clearly more reviews containing positive opinions and thoughts regarding Madeira, than the opposite. The scale positive is the one most found in the reviews, counting 780 reviews. Then, the one with the highest number of reviews is strong positive with 254 reviews. Only in the two mentioned scales, there are a total of 1043 reviews, nearly 90.07% of the total number of reviews. This shows the huge positive polarity in which the reviewers expressed their outbursts in TripAdvisor. Moreover, 74 reviews were considered as having a neutral polarity, 39 reviews were accounted to negative polarity and only 1 review was considered as strong negative.
The average polarity according to the reviews that were extracted was 4.86 – which is very good and shows the positive sentiment, happiness and satisfaction in the elements that the reviewers debated their thoughts.
The next thing being studied is the agreement – this one corresponds to the agreement between the sentiments detected in the text, having two possible meanings: agreement – same polarity between the different elements in the review; and disagreement – different polarity between the different elements in the review.
As it is possible to see in Annex 8.10, there is only a small difference between the number of reviews that were considered as having all the elements in agreement versus the ones in disagreement. For instance, 611 reviews were classified as agreement – corresponding to 53.22% of the total reviews, while 537 were classified as in disagreement – corresponding to the remaining 46.78%.
It might also be interesting to analyse the average of polarity according to the reviews in agreement/disagreement. In Annex 8.11 those results are shown. As it is possible to see, from the reviews that were considered in agreement the average of polarity is 4.36, while the ones in disagreement is 3.77. This means that the reviews in which the elements have the same polarity are more positive, than the ones in which the elements do not.
Another aspect to be studied is the Subjectivity. It is the degree in which the reviews when written were influenced by the reviewer’s personal feelings, taste or opinions. If it was influenced by any of that, it is considered subjective, if it was not it is objective.
The results in Annex 8.12 show that there is a vast majority of reviews that were considered subjective – 1089 reviews, corresponding to 94.86% of the 1148 reviews extracted. The number of objective reviews stood in 59, corresponding only to 5.14%. From this it can be taken that the vast majority of reviewers uses their personal feelings, taste or opinions to influence their commentaries online.
Again, it might be interesting to explore the positive/negative sentiments of the reviews that were considered objective and subjective. In Annex 8.13 it is possible to find that, and the results show that there is a slight difference in the two types of reviews. For instance, the reviews considered subjective have a higher average of polarity than the objective ones – 4.10 and 3.81 (out of 5) respectively. It can be said that the reviewers when being subjective show more positive sentiments, than when being objective.
As previously mentioned, the software allows the results to be studied according to the degree of irony of the reviews. In Annex 8.14 it is possible to analyse that there is a clear preponderance of non-ironic reviews, in comparison with ironic ones – 1101 reviews versus 47. These outcomes lead to 95.91% of the reviews being considered as not having ironic marks, and only 4.09% having.
The average Polarity according to the Irony can be found in Annex 8.15 – and from there, it is possible to see that there is only a really small difference between the average polarities of both elements. The non-ironic reviews present an average of 4.09, which is higher when compared to the ironic reviews – 3.96. Nonetheless, it means that the non-ironic reviewers show a highest positive sentiment in their reviews, in comparison with the ironic ones.
In the following sub-chapters it will be presented the results regarding the same codes explained in this chapter but according to the TEM, Category and Membership Level.

4.2.1.3.2. GLOBAL SENTIMENT ANALYSIS APPLIED TO TOURISM EXPERIENCE MODEL
As it was already done using previous techniques (example: Text Classification and Topics Extraction) the study is applied to the different codes. In this sub-chapter it will be analysed the results referring to the application of the Global Sentiment Analysis to the TEM – Tourism Experience Model.
In Annexes 8.16, 8.17, 8.18 and 8.19, it is possible to find the results of the software analysis, hereby expressed in this chapter.
In Annex 8.16 it is presented the results regarding the average polarity according to TEM. As it is possible to see, the activities and places connected with the Knowledge Seeker experience (4.18) were the ones, from the reviews extracted, that created in the reviewers the most positive sentiment, followed by Pure Pleasure (4.16) and Holist (4.08). The Re-Discover experience was the one where the reviewers felt the least positive sentiments.
Moreover, in Annex 8.17 it can be seen the number of reviews according to TEM and Agreement, and it can be said that from the reviews in Agreement, the Pure Pleasure reviews were the ones detected as having the most positive sentiments (4.49), while the Holist one was the one containing the least positive sentiments (4.28). In relation to the reviews classified as in Disagreement, the Knowledge Seeker experience was the one that conducted reviewers to be more satisfied and having positive feelings (3.92), and on the opposite side can be found Re-Discover (3.69).
Furthermore, in Annex 8.18 can be seen the results regarding the same aspects but in accordance with the Subjectivity, instead of Agreement. From the reviews considered as Objective, it can be stated that the reviews containing Pure Pleasure experiences were the ones motivating reviewers to better positive sentiments (4.18), while the Re-Discover one the opposite (3.55). In relation with the reviews considered Subjective, the most positive sentiments came from the reviewers experiencing Knowledge Seeker (4.19), and the opposite from Re-Discover (4.04).
Finally, in Annex 8.19 the results regarding the Global Sentiment Analysis according to TEM and Irony are presented. From the reviews classified by the software as Ironic, the experience that made the reviewers feel the most positive sentiments was Re-Discover (4.05), and the least positive sentiments was Knowledge Seeker (3.60). Regarding the Non-ironic reviews, the most positive sentiments were found in the reviews containing Knowledge Seeker experiences (4.20), and the opposite in reviews containing Re-Discover experiences (4.02).
4.2.1.3.3. GLOBAL SENTIMENT ANALYSIS APPLIED TO MEMBERSHIP LEVEL OF REVIEWERS
The same aspects shall now be analysed according to the Membership Level of Reviewers. In Annexes 8.20, 8.21, 8.22 and 8.23 can be found all the results regarding this subject. In Annex 8.20 it is presented a graphic representation of the average polarity according to high/low ranked reviewers. The results are very close, with the low ranked reviewers being just a little bit more positive in the elements that constitute their reviews in comparison with the high ranked ones – 4.12 versus 4.03, respectively. This means that the LRR showed more positive sentiments regarding the island than the HRR.
Moreover, in Annex 8.21 it can be found the results of the Global Sentiment Analysis according to the Membership Level of Reviewers and Agreement. From the reviews in Agreement, it can be said that the LRR were the ones that experienced the most positive sentiments in comparison with the HRR (4.38 vs 4.33, respectively). In relation with the reviews in Disagreement, the results are the same, the LRR showed a better sentiment towards their stays in the island than the HRR (3.79 and 3.75, respectively).
Furthermore, in Annex 8.22 it can be seen the same results as before but according to the Subjectivity. From the reviews considered Objective, it can be stated that the LRR were the ones that felt a better positive sentiment (3.92), in comparison with the HRR (3.64). Additionally, from the Subjective reviews, it can be stated the same – LRR 4.13 vs HRR 4.05.
At last, in Annex 8.23 can be analysed the results according to the Irony. From the reviews classified as Ironic, again the LRR were the ones that had the best sentiments towards their stays (4.04), when compared with the HRR (3.88). In relation with the Non-ironic the results are about the same – LRR 4.12 and 4.04 HRR.

4.2.1.3.4. GLOBAL SENTIMENT ANALYSIS APPLIED TO CATEGORIES
In Annexes 8.24, 8.25, 8.26 and 8.27 it can be found the average polarity according to the different categories which the TripAdvisor allows to segment and the codes of the Global Sentiment Analysis.
In Annex 8.24 it is presented the results regarding the average polarity according to the different types of Categories. It can be said that the Restaurant category, from the reviews extracted, is the one that aroused the best positive sentiments (4.30) in the reviewer’s experiences. Moreover, it can be found Leisure Activities (4.15), Hotel (4.04), and finally the category that aroused the most negative sentiments in the reviewers were Cultural Activities (3.94).
Moreover, in Annex 8.25 it can be seen the results regarding the Global Sentiment Analysis according to Categories and Agreement. From the reviews considered as in Agreement, it can be said that the Restaurant category (4.52) is the one that manifested the most positive sentiments in the reviewers stays, and the most negative sentiments were felt by the reviewers in Cultural Activities (4.21). In relation with the reviews considered in Disagreement, the results are the same – Restaurant being the category with the most positive sentiments (3.94) and Cultural Activity with the most negative ones (3.58).
Additionally, in Annex 8.26 the same results can be analysed but according to the Subjectivity instead of Agreement. From the reviews classified as Objective it can be stated that the most positive sentiments aroused from Leisure Activities (4.57), and the most negative ones from Cultural Activities (3.66). Furthermore, from the reviews analysed as Subjective, the most positive sentiments aroused from Restaurants (4.31), while the most negative ones from Cultural Activities (3.97).
At last, the same results but according to Irony can be found in Annex 8.27. From the Non-ironic reviews, the Restaurant category (4.30) was the one arousing reviewers the best sentiments, while the Cultural Activity (3.95) the worsts. In relation to the Ironic reviews, the same conclusions are applied, Restaurant category (4.23) as being responsible for the most positive sentiments, while Cultural Activities (3.68) responsible for the most negative ones.

4.2.2. RESULTS STUDY 2 – BERMUDA

4.2.2.1 RESULTS STUDY 2 – TEXT CLASSIFICATION

4.2.2.1.1 GENERAL TEXT CLASSIFICATION RESULTS

The results of the general information regarding the analysis of the Text Classification on the Bermuda’s extracted reviews include a variety of themes being mentioned by the reviewers, particularly 16 different ones. The most important ones were gathered in Annex 9.1, and culminate in 5 major themes that are really preponderant in the total number of reviews extracted – corresponding to 87.60%.
It is possible to create two separate groups according to the number of reviews where each theme was found, by looking at the graphic representation
The first group includes two themes: Lifestyle and Leisure (176 reviews) and Tourism, Travel and Commuting (165 reviews). These themes are responsible, together, for 341 reviews.
The second group of themes comprises three subjects: Art and culture (55 reviews), Social issue (48 reviews) and Sports (41 reviews). These themes are responsible for 144 reviews.
It is also important to refer that 81 out of the 635 reviews extracted were unable to be analysed by the software, which ultimately results in 12.76% of the reviews not being properly studied.

4.2.2.1.2 TEXT CLASSIFICATION APPLIED TO TOURISM EXPERIENCE MODEL
The results of the Text Classification according to the Tourism Experience Model regarding Bermuda can be found in Annex 9.2.
The Holist experience includes only two major themes preponderant: Lifestyle and Leisure (5 reviews) and Art and Culture (3 reviews).
The Knowledge Seeker comprises three major themes being mentioned. Lifestyle and Leisure with 35 reviews, Tourism, travel and commuting counting 29 reviews and Art and Culture summing up to 24 reviews.
The Pure Pleasure comprises two most important themes: Lifestyle and Leisure (37 reviews) and Tourism, travel and Commuting (21 reviews).
Finally, the Re-Discover experience includes four major themes: Tourism, travel and commuting with 113 reviews, Lifestyle and Leisure counting 99 reviews, Social Issue with 33 reviews and Sports accounting 31 reviews.

4.2.2.1.3 TEXT CLASSIFICATION APPLIED TO CATEGORIES
It is possible to differentiate the themes most mentioned by the reviewers according to the categories of each review when extracted, using a Text Classification analysis. The results can be found in Annex 9.3.
The Cultural Activity category englobes three major themes: Tourism, travel and commuting (49 reviews), Art and culture (27 reviews) and Lifestyle and leisure (24 reviews).
The Hotel category includes Lifestyle and leisure (48 reviews) and Tourism, travel and commuting (45 reviews).
The Leisure Activity comprises Tourism, travel and commuting (60 reviews), Lifestyle and leisure (31 reviews) and Sport (26 reviews).
Finally, the Restaurant category comprises only one major theme – being that, Lifestyle and leisure (73 reviews).

4.2.2.1.4 TEXT CLASSIFICATION APPLIED TO MEMBERSHIP LEVEL OF REVIEWERS
In Annex 9.4 it is possible to find the results of the Text Classification analysis according to the Membership Level of Reviewers code.
The High Ranked Reviewers in their reviews were found to be mentioning three major themes: Lifestyle and Leisure (80 reviews), Tourism, travel and commuting (79 reviews) and Art and Culture (27 reviews).
The Low Ranked Reviewers present a little more diversity and show five themes in the most important ones: Lifestyle and Leisure (96 reviews), Tourism, travel and commuting (86 reviews), Social issue (34 reviews), Art and culture (28 reviews) and Sport (28 reviews).

4.2.2.3 RESULTS STUDY 2 – GLOBAL SENTIMENT ANALYSIS

4.2.2.3.1 GLOBAL SENTIMENT ANALYSIS GENERAL RESULTS
The first thing being studied is the polarity. From the results achieved it is possible to see that the average polarity is 4.10, which means that globally the sentiment of the reviewer’s towards their experiences and activities in the island was positive. Moreover, the information above can be proved in Annex 9.9, where it can be seen clearly that the result “positive” – 4, was the one most experienced (406 reviews).
The second thing being studied is the agreement (Annex 9.10) – which, as seen before, studies the polarity between the different elements in the same review. From the 635 reviews, it is possible to see that 355 of the total reviews were considered in agreement, while 280 were classified as in disagreement. Moreover, and in a more important matter, the average polarity of each type of review can be found in Annex 9.11. The sentiment of the reviews in agreement is fairly more positive than the ones in disagreement – 4.35 and 3.79, respectively.
Furthermore in Annex 9.12, the number of reviews according to the Subjectivity can be found – also as seen before, the subjectivity is the degree in which the reviews when written were influenced by the reviewer’s personal feelings, taste or opinions. It is possible to see that there is clearly more reviews considerate subjective (575), while only 60 considered objective. Moreover, in Annex 9.13 the results show that the reviews considered subjective are a little more “sentiment positive” than the ones classified as objective (4.11 and 4.05, respectively).
At last, the results of the Irony can be found in Annexes 9.14 and 9.15. The first one states that from the 635 reviews, almost the total was categorized as non-Ironic (621 – corresponding to 97.80%), while only 14 were considered ironic (2.20%). The second one refers to the sentiment of the reviewers when using irony or not. It can be seen that when the reviewers used irony to express their reviews the sentiment that they were feeling was more positive than when not using irony.
In the following sub-chapters it will be presented the results regarding the same codes but applied to TEM, Category and Membership Level of Reviewers.

4.2.2.3.2 GLOBAL SENTIMENT ANALYSIS APPLIED TO TOURISM EXPERIENCE MODEL
In Annex 9.16 it can be seen the average polarity regarding each type of experience felt by the reviewers in their stays. The Re-Discover experience was the one that created a better positive sentiment in the reviewers (4.12), followed by the Knowledge Seeker experience (4.10) and Pure Pleasure one (4.09). Clearly, in a not so positive sentiment the Holist experience with 3.83, was the one that the reviewers were not so satisfied about.
Moreover, in Annex 9.17 it is possible to see the Average Polarity according to both TEM and Agreement, and understand that from the reviews in Agreement, the ones gathered from reviewers who were experiencing Holist are the ones that the sentiment was more positive (4.60), but from the ones in Disagreement this experience was the one that manifested the most negative sentiment among the reviewers. Furthermore, from the reviews in Disagreement, the reviewers who showed a more positive sentiment were the ones experiencing Knowledge Seeker, while on the opposite side this experience showed to be the most negative sentiment-related in the reviews in Agreement.
In Annex 9.18 it is possible to do the same analysis but regarding the Subjectivity, instead of Agreement. It can be said that from the Objective reviews, the Holist experience was the one that had the most positive sentiment (5.00) in the reviewer’s stays and on the opposite side can be found the Knowledge Seeker experience (3.94). Furthermore, from the reviews considered Subjectivity reviews the Re-Discover experience was the one where the reviewers had a bigger positive sentiment (4.13), contrary to the Holist (3.76).
Finally, in Annex 9.19 the same analysis can be done according to TEM and Irony. From the reviews classified as ironic, it is possible to state that the reviewers experiencing Pure Pleasure were the ones with the most positive sentiment (4.50), the most unhappy were the ones experiencing Holist and Knowledge Seeker (4.00, both of them). Moreover, from the non-ironic reviews, the ones with a most positive sentiment were the ones experiencing Re-Discover (4.12) as well. On the opposite side, it can be found the reviewers experiencing Holist (3.82).

4.2.2.3.3 GLOBAL SENTIMENT ANALYSIS APPLIED TO CATEGORIES
In Annex 9.20 the results show that the category which created the most positive sentiment in the reviewer’s experiences in the island was the Hotel category (4.18), and the one that was the most negative experience was the Restaurant category (4.01).
Moreover, in the Annex 9.21 it is possible to see the average polarity according to Categories and Agreement. The results show that from the reviews in agreement the ones that urged a more positive sentiment in reviewers were the ones connected to Hotel category (4.39), while the opposite can be found in the Restaurant category (4.28). In the Disagreement classified reviews, the most positive was the ones from Restaurant category (3.80), and the most negative sentiment was seen in both Hotel and Cultural Activity (3.77, respectively).
Furthermore, in the Annex 9.22 it is shown the results according to Categories and Subjectivity regarding Bermuda. From the results gathered in the reviews considered Objective the category that urged the most positive sentiments in the reviewers was the Hotel (4.21), and the opposite was the Leisure Activity ones (3.89). From the reviews considered Subjective, again the Hotel category was the one in which the reviewers felt the most positive sentiments (4.18), on the other end the Restaurant category (4.02) was the one where the reviewers experienced the most negative sentiments.
Finally, in Annex 9.23 it can be seen the results according to Categories and Irony regarding Bermuda. It can be said that, from the reviews considered Ironic, the category that urged the most positive sentiment was Hotel (4.33), and the most negative ones were Cultural Activity and Restaurant (4.00). From the reviews classified as Non-Ironic, again the Hotel category was the one in which the reviewers felt the most positive sentiments (4.18), and the Restaurant (4.01) was the one in which the reviewers felt the most negative sentiments.
4.2.2.3.4 GLOBAL SENTIMENT ANALYSIS APPLIED TO MEMBERSHIP LEVEL OF REVIEWERS
In Annex 9.24 can be seen that the Low Ranked Reviewers showed a more positive sentiment in their opinions and thoughts regarding the activities done and places visited in the reviews extracted (4.12), in comparison with the High Ranked Reviewers (4.08).
Moreover, in Annex 9.25 it is shown the results of the Global Sentiment Analysis according to the Membership Level of Reviewers and the Agreement. It can be stated that from the reviews classified as in Agreement and Disagreement, it was the Low Ranked Reviewers (versus the High Ranked ones) who experienced the island with a best positive sentiment (Agreement – 4.36 LRR vs. 4.34 HRR and Disagreement – 3.81 LRR vs 3.75 HRR).
Furthermore, in Annex 9.26 it can be seen the same results but in accordance with the Subjectivity, instead of the Agreement. The results show that again the LRR were the ones experiencing the most positive sentiments in the island, in comparison with the HRR (Objective – 4.06 LRR vs 4.03 HRR and Subjective – 4.12 LRR and 4.09 HRR).
Finally, Annex 9.27 shows the same results but in accordance with the Irony instead of Subjectivity or Agreement. It can be said that the same results are applied, both in Ironic and Non-Ironic reviews, it was the LRR that showed the most positive sentiments (Ironic – 4.22 LRR vs 4.20 HRR and Non-Ironic – 4.11 LRR vs 4.08 HRR).
5. FINDINGS/NOTES
This chapter shows some conclusions that were taken after the Results, indicated in the last chapter, and the Annexes, and it will serve as a support for the Discussion chapter coming next.

5.1 MADEIRA (Study 1)
Based on the results of Study1 – the first aspect to be mentioned is that the amount of Cultural Activities regarding Madeira, is far superior than the Leisure Activities one (459 and 163 reviews, respectively), which indicate that the activities that are done by the tourists in the island are more related with culture than leisure.
Moreover, the average rating (4.59) calculated based on the reviews extracted, indicates that generally the reviewers are truly happy and satisfied with the activities, accommodations and places attended during their stays. Although this fact, from the four categories in which the reviews were extracted, it is possible to say that there are two of them that got an average rating slightly superior, being Leisure Activities and Restaurants the favourite ones according to the reviewers.
Another interesting fact is that although reviewers tend to do more Cultural Activities than Leisure ones, the average rating of the second is higher than the first (4.81 – Leisure Activities and 4.45 – Cultural Activities).
In relation to the Membership Level of Reviewers code, it is possible to say that since the vast majority of reviewers are Low Ranked, according to Chen (2015) different membership levels have different attitude towards the information, and LRR will accept the information in a higher probability while HRR have higher change of receiving the information but less probability to accept it, it is fair to say that the higher the number of LRR the best the information exposed publicly will be received and accepted. Following that same logic, it can be said that since the majority of reviewers of the Island of Madeira throughout the TripAdvisor platform are Low Ranked Reviewers, they are more likely to accept the information online than the opposite.
Another conclusion towards the Membership Level of Reviewers is that there was no significant difference between how the two types of reviewers rate their experiences in the island.
As far as it concerns the Language type code, it was obtained that the reviewers mostly used Literal language in detriment of the Figurative one. The difference of the average rating between reviews expressed in the two different language types was also not significant.
What may be interesting to conclude regarding this coding, is that from the information gathered there is a significant percentage of LRR using Figurative language type, and according to Wu et al. (2017:590) “consumers exhibit less favourable attitudes and lower reservation intention after reading a figurative (vs literal) review posted by a low expertise level”. The 22.51% (almost a quarter) of Low Ranked Reviewers using Figurative language may be therefore a threat to the Tourism of Madeira.
Regarding the results of the application of the Tourism Experience Model according to the reviews that were extracted, it is possible to conclude that more than half of the reviews (51.83%) were found to have reviewers experiencing Re-Discover, which is quite dominant. Nonetheless, that one was the experience with the lowest average rating given by the reviewers, along with Pure Pleasure. The Holist experience was the one best rated according to the reviews extracted.
To take any conclusions regarding this topic it is necessary to have in mind, the framework, and how the reviews are spread across the TEM graphic representation. The next figure – Figure 13, analyses the figure in a horizontal way and expresses the total percentage of reviews per experience that were extracted to study Madeira.

As it is possible to see there is a clear discrepancy in the results obtained. 84.84% of the reviews are presented with tourists perceiving their experiences as Pure Pleasure or Re-Discover.
This last data is quite important to understand that there is a side of the graphic that is way more important than the other. As mentioned in the Methodology chapter, the horizontal axis – Activity axis – comprises two domains. The left side corresponds to the Recreational Activities, and the other one Exploratory Activities – “here tourists seek new insights, understandings, social and bodily feelings (…) create new knowledge and skills as much as they induce the tourist to change and transform through learning” (Gnoth & Deans, 2012). By the perception of the graph, it is easily recognizable that the Madeira Island is experienced, through the four categories studied, by the reviewers as a destiny of Recreational Activities rather than Exploratory.
Recreational activities are experiences that the visitors engaged before and are a result of habit, training and repetition. Those activities are held as “helping people regain their balance, their strength, or their self-esteem, or all of these together” (Gnoth & Deans, 2012). These activities are normally considered entertaining and as shown in Figure 13 are related with Being rather than Becoming – modifying the inner self of the person.
After analysing the horizontal axis it is necessary to study the vertical one as well – Consciousness Axis – “relates to the style of how tourists receive their experience of the destination” (Gnoth & Deans, 2012).
As it was seen before this axis comprises two domains. The upper side corresponds to the Human Being – as in “finding ourselves as human being while stripping ourselves of the induced values, habits and stereotypes to get close to our existential being” (Gnoth & Deans, 2012). The lower side corresponds to Person – as receiving the experience as guided by role-expectations. The higher the authenticity the more “role-authentic” the person is.

In Figure 14, it is possible to see the percentage of experiences according to the vertical axis. The upper side of the axis, which wraps the Human Being, comprises 54.97% of the experiences lived by the tourists, while the Person one comprises the remaining 45.03%.
According to Gnoth and Deans (2012) Madeira is therefore perceived consciously more as a place where people are able to forget, or move away, from the entrenched dogmas, stereotypes and values of society and get closer to what they really are – their inner self.
Combing the two results obtained, from the last two figures and the studies scientifically accepted that were already mentioned, Madeira can be seen as a destiny where people especially through Recreational Activities are able to get closer with their existential being.
In relation with the Symbology code, it can be said that there is a minimum percentage of reviewers using any kind of emojis/emoticons (only 3.48%). The positive thing that can be outlined is the fact that almost all those reviewers used symbols in a positive way (39 out of 40).
Moving on to the Content code, it can be said that there are some themes that are more important for the reviewers to mention in their reviews than others. For example, it was very important for reviewers to mention certain words related with gastronomy, accommodation and activities frequented/experienced in the island. There are just a few words (and in a small scale) that appear as a negative connotation, which highlights that generally the content referred by the tourists writing reviews is quite positive. For the Tourism of Madeira it is also important to refer that reviewers are inviting other possible tourists to visit this destination – that can be explained by the appearance of “recommendation” in so many reviews and in the final Wordle output.
The Positive aspects mentioned by the reviewers showed a load of adjectives, which is normal once the reviewers used them to classify whatever they were referring to. Other important aspect is the appearance of words connected with all of the four categories, of the extracted reviews, in the final Wordle. This means that globally there was a dispersion of the positive content across all the accommodations, restaurants, leisure and cultural activities experienced by the tourists. This information also significate that there words linked with all of the TEM experiences, which serves as a complement to the already mentioned satisfaction shown by the reviewers on all categories/experiences. Again it appears to be important for reviewers to “recommend” other possible tourists to have the same experiences that they had.
Still about the same subject, according to Chen and Tseng (2010) and Cheng, et al. (2017) “positive opinions may persuade users to purchase a product”, which shows the importance of the positive content hereby analysed. Adding this, to the fact that from the Wordle results of the Content code there is a vast majority of words with a positive connotation and the fact that in the final database 98.78% of the total amount of reviews contain positive information, it is therefore believed that overall, the opinions expressed online regarding the island of Madeira, as a destination, are positively persuading potential tourists to choose to visit the island.
At last, the Negative aspects also show a bunch of adjectives in the final output, which again is normal, as the reviewers used them to express and classify the subjects that they were referring to. From the results it is possible to say that the majority of words with a negative connotation are linked with the left side of the TEM (Re-Discover and Pure Pleasure), meaning that there were more negative things to say regarding Recreational Activities than Exploratory ones. It is also important to refer that 33.45% of the total amount of reviews contain at least a piece of negative information – more than one third of the reviews extracted, which is quite significant.
According to Chen (2015) negative information is capable of presenting the problems of products (such as hedonic ones – like travel destinations), which will ultimately decrease the trustworthiness of the same. This is fact that the Tourism of Madeira has to deal with, due to the significant amount of negative information circulating online. Nonetheless, this data is important because the responsible entities are now aware of the fact and can modify their strategies in order to correct the situation.

5.2 MADEIRA (Study 2)
Starting by the first analysis that was performed under the Meaning Cloud scope – Text Classification, it is possible to say that from the 17 different themes that it was able to identify, there are two that are undoubtedly more presents in the reviews than the others: Lifestyle and Leisure and Tourism, Travel and Commuting. This means that the reviewers are interested in expressing their thoughts towards the life they had during their stays in the island, but also locations and transports they caught while they were experiencing the Island. This information can be confirmed by the Topics Extraction analysis that was done, in which from the 20 major concepts extracted; almost all of them are linked with the two themes above mentioned:
1. Lifestyle and Leisure: “Hotel”, “Room”, “Staff”, “Pool”, “Snooker”, “Restaurant”, “Food”, “Wine”, “Breakfast”;
2. Tourism, Travel and Commuting: “Madeira”, “Location”, “Funchal”, “Car”, “Driver”, “Trip”.
From the 20 concepts extracted as the ones most mentioned by the reviewers, 15 can be included in both of themes, showing the importance of these subjects for the reviewers.
Getting back to the Text Classification results, it can be seen that in all of the four categories TripAdvisor allows to segment the reviews, it can be found the same two themes already mentioned. There are afterwards themes that appear in two of those categories, and are also important to characterize the perception of the reviewers towards the island of Madeira, such as: Environment, Weather and Energy, Art and Culture and Finances. Through the Topics Extractions results it is possible to see that it complements this information, as the main concepts obtained for each category can be included in the themes analysed by the Text Classification. For instance:
1. Cultural Activities appear to be more connected with Tourism, Travel and Commuting – “Car”, “Madeira”, “Funchal”, “Location”, “Monte”, “Trip” and more;
2. Leisure Activities involve Environment, Weather and Energy, Travel, Tourism and Commuting, but also Lifestyle and Leisure – “Madeira”, “driver”, “trip”, “road” , “dolphin”, “location”, etc.;
3. Hotel respects more Lifestyle and Leisure and Tourism, Travel and Commuting – “Room”, “breakfast”, “pool”, “hotel”, “location”, “snooker”, “restaurant”, etc.;
4. Restaurant category involves more Lifestyle and Leisure, Art and Culture and Tourism, Travel and Commuting – “Food”, “Restaurant”, “Service”, “Steak”, “Wine, “Atmosphere, “Location” and “Madeira”.
On the Tourism Experience Model application, the results of the Text Classification obtained something close to what was already mentioned. Tourism, Travel and Commuting is the theme that the reviewers mention in all of the four experiences, followed by Lifestyle and Leisure, Environment, Weather and Energy, Art and Culture – appearing in three of the four experiences. On the Topics Extractions results it is possible to find:
1. Holist appears to more connected with Tourism, Travel and Commuting and Environment, Weather and Energy – “garden”, “Funchal”, “Madeira” and “Location”;
2. Knowledge Seeker more involved with Tourism, Travel and Commuting, Lifestyle and Leisure and Art and Culture – “Madeira”, “driver”, “Wine”, “guide”, “trip”, etc.;
3. Pure Pleasure indicates clearly more words related with Lifestyle and Leisure and Art and Culture – “Room”, “Pool”, “Hotel”, “Food”, “Breakfast” and “Restaurant”;
4. Re-Discover one is linked in a higher level with Tourism, Travel and Commuting, Lifestyle and Leisure and Environment, Weather and Energy – “Madeira”, “Funchal”, “Location”, “Car”, “garden”, “trip”, “staff”, “room” and so on.
Finally, regarding the two types of reviewers based on their Membership Level, there is not a huge difference, except the fact that low ranked users seem to give more importance to Social Issues and Finances than the high ranked users. On a Topics Extraction level it can be seen exactly that, by the results showing almost the same exact concepts – see attachment X.
All in all, the results of the two analysis hereby mentioned – Text Classification and Topics Extraction, serve as a complement to each other, and what it is important to refer is that going coding by coding, or in a general level, the themes that matter the most, and for that reason the ones most mentioned by the reviewers in their reviews regarding Madeira are Lifestyle and Leisure, Tourism, Travel and Commuting, Environment, Weather and Energy and Art and Culture. These four themes are the base of the e-WOM that it is being created online in TripAdvisor regarding Madeira.
The next step is to discuss the results obtained in the Sentiment Analysis performed – Global Sentiment Analysis and Topics Sentiment Analysis.
Starting with the Global Sentiment Analysis, on a general level, the average polarity is a very important indicator, because it indicates the average general sentiment attributed by the reviewers in the e-WOM. A score of 4.86 is extremely positive, and it shows that the globally the reviewers felt profound positive sentiments in their stays and experiences in the island. Such result can be perceived as online reviewers being positively engaged with the island spreading positive information about it, thus increasing the possibility of re-experience or new customers to experience (Islam, et al., 2012; Zeithaml, et al., 1996). Other studies corroborate the same conclusion, as a score close to the maximum scale (4.6 out of 5) may also outcome in those reviewers becoming loyal to the destiny (Chang, et al., 2009; Ribbink, et al., 2004), which will ultimately enhance re-visiting it (Chang, et al., 2009).
Keeping on a general level, the reviewers when in agreement with everything they are mentioned tend to be more positive, than when in disagreement – Agreement (4.36) and Disagreement (3.77). Moreover, when the reviewers write their reviews based on their personal feelings and dogmas they show more positive sentiments, than the opposite – Subjective (4.10) and Objective (3.81). Finally, the reviewers when not being ironic showed more positive sentiments in their reviews, than when being ironic – Ironic (3.96) and Non-Ironic (4.09).
Moreover, regarding the Category coding it is important to refer that on a general level, the Restaurant was the one with the higher average polarity (4.30) and for that reason the one which urged in the reviewers the best positive sentiments. On the opposite side it can be found the Cultural Activities (3.94), with the most negative sentiments. The Restaurant was in fact the category with the highest positive sentiments in almost every aspect – Agreement, Disagreement, Subjectivity, Irony and Non-Irony, only Objective got Leisure Activities has having the most positive sentiment reviews. On the opposite side, Cultural Activities was the one found having the reviews where the reviewers showed the least positive sentiments, in all tests. Regarding the Tourism Experience Model, on a global way the Knowledge Seeker experience was the one in which the reviewers felt more positive sentiments – that could also be seen in the reviews in Disagreement, Subjectivity and Non-Ironic. The Re-Discover experience was, globally, the one found creating the most negative impact in the reviewer’s sentiments – also the most negative in reviews in Disagreement, Objective, Subjective and Non-Ironic. Finally, the last coding in the Global Sentiment Analysis tell us that the Low Ranked Reviewers obtain more positive experiences, and more positive sentiments, in every possible angle when compared with the High Ranked Reviewers.
Moving on to the last analysis performed – Topics Sentiment Analysis, indicate that the reviewers selected the topics of Hotel, Services and Gastronomy as the ones most positive. The topics with the most negative sentiments score was Measures – involving distances, prices, etc., Tourism and Weather and Meteorology. Although being the most negative topics, these last three topics are still above the neutral score, meaning that globally it can be found more positive sentiments than negative ones.
Regarding the Category coding, the most important aspect to refer is that Restaurant was the category with the highest positive sentiments and Cultural Activities the one with the most negative ones – as has happened in the Global Sentiment Analysis. Moreover, the Tourism Experience Model coding tells us that the Pure Pleasure experience was the one containing the topics that the reviewers felt more positive sentiments, while the contrary happens in the Knowledge Seeker experience. The last coding – Membership Level of Reviewers, tells us that the Low Ranked reviewers tend to have their best experiences related to Social Events, Services and Hotels, while the High Ranked ones with Hotels, Services and Gastronomy. As it is possible to see the results are close between them.

5.3 BERMUDA (Study 1)
The first thing to be mentioned is that contrary to Madeira, in Bermuda there is a slight preponderance for the tourists to attend more Leisure Activities than Cultural ones – although the gap between them is not as significant as in the Madeira’s case (214 and 158 reviews, respectively).
Furthermore, the average rating (4.43) that was calculated based on the 635 reviews extracted, indicate that globally the reviewers manifest happiness and satisfaction with the accommodations, activities and places attended during their stays. There is one category in particular that is distinguished by having a average rating (4.71) superior than all the others, together with the fact of being the one with the highest amount of reviews, which reinforces the strong happiness towards the Leisure Activities. Moreover, the Restaurant category was considered by the reviewers as the most negative among all the others.
Moving on to the Membership Level of Reviewers it was seen that there was a majority of LRR (56.69%) in comparison with the HRR (43.31%), which according to Chen (2015) means that the majority of reviewers of the TripAdvisor platform regarding Bermuda are more likely to accept the content displayed online by other reviewers.
One thing that has to be taken into account is that contrary to the conclusions taken about Madeira’s case, in Bermuda is possible to say that LRR tend to better evaluate their activities and experiences in the island in comparison with the HRR – 4.56 and 4.24 average rating, respectively.
In relation to the Language type used by the reviewers it is quite evident that there is a massive preference for using Literal language (80.31%), in detriment of the Figurative one (19.69%). Also, as has happened in the previous case, the average rating difference between each Language type is quite similar and therefore not significant to take any major conclusions.
Again, what is important to refer is that both HRR and LRR prefer the use of Literal language which is very positive because according to Wu, Shen, Fan and Mattila (2016), Figurative language can cause less favourable attitudes, and lower booking intent, by potential readers when written by LRR – the same does not happen when you are talking about HRR, since the language effect is attenuated. Although this result there is still an amount of reviewers that should concern the entities of the Tourism of Madeira, by the fact that 21.11% of the LRR are using Figurative language, which may cause potential tourists to drop their intentions to visit the island.
Regarding the results of the Tourism Experience Model, it can, again, be concluded that the vast majority of tourists lives experiences of Re-Discover – 61.73%, which is a percentage quite expressive. Yet again, despite being the most “lived experience” it had one of the lowest average ratings (4.41), only overtaken by the Knowledge Seeker experience, which was the one evaluated as more negative by the reviewers (4.40). The Holist experience was the one with the highest average rating.
As done in the previous chapter, it is now necessary to have in mind the TEM framework, and how the four different experiences are spread across its’ graphic representation. The next figure – Figure X, represents the percentage of the experiences taken from the reviews.

On a horizontal point of view, it is clear that the left side of the axial representation is way more important than the right one. 75.90% of reviewers either experienced Re-Discover or Pure Pleasure, while only 24.10% experienced Holist or Knowledge Seeker.
The left side of the axis, as seen before, comprises the Recreational Activities, which means that reviewer’s activities in Bermuda are often a result of habit, training and repetitive actions. People are helped through those activities to regain their balance, strength and self-esteem.
One thing that can be immediately concluded is that, through the four mentioned experiences, reviewers perceived the island of Bermuda as a destiny of Recreational Activities rather than Exploratory.
The vertical axis (see figure X) shows a higher difference in the percentage of reviews in the upper/lower side of the graphic compared with Madeira. 64.57% of the reviewers either experienced Re-Discover or Holist – as seen also in the Madeira’s results but in a smaller gap.

The conclusions reflected in the previous chapter may also be considered in this case. Bermuda can be seen as an island destination where people are capable of being closer to their existential being – being able to forget about the induced values and stereotypes of society.
Doing a combination between the results taken from figure X and figure X, it is possible to say that Bermuda is perceived as a destiny where people, especially through recreational activities are able to get closer with their existential being.
Advancing to the Symbology code, the results show a minimum use of such emojis/emoticons in their reviews – only 1.10%, which tells us that perhaps such amount of reviews is not sufficient to study this code and more reviews would be needed. Nonetheless 71.43% of the emojis/emoticons found were used in a positive way, which is ultimately a good outcome.
The next conclusions to be taken are related with the Content code. Again, some of the major themes being referred by the reviewers in their reviews are related with words linked to gastronomy, activities done, places visited and certain leisure words that are connected with accommodation services. On a smaller size it is also possible to find words linked with culture. Therefore, it is possible to say that the results show a spread of words across all the Categories, and consequently all the TEM experiences.
Again, there are some words that are mentioned with a negative connotation, one of them in a relatively big size – “price”, meaning that globally the reviewers did not enjoy the prices that were practiced in the different island activities/experiences.
Other thing worth mentioning is again the appearance of words related with reviewers inviting other possible tourists to enjoy the same experiences that they had – words like “recommendation”.
In relation with the Positive aspects being stated by the reviewers in the amount of reviews that were extracted, the results showed once again the appearance of a lot of adjectives that were used by the reviewers to classify whatever they were talking about. As has happened in the Madeira’s case, the Bermuda’s Wordle reflects words linked with accommodations, restaurants, leisure and cultural activities – which imply that all the experiences that the TEM is concerned, are represented as well. It is also important to refer that, as mentioned in the Content considerations, due to the high average satisfaction shown by the reviewers the appearance of the word “recommend” pops up in a relatively big size.
According to Chen and Tseng (2010) and Cheng, et al. (2017) positive opinions from other reviewers may influence others to purchase a product – including hedonic products, such as travel destinations Chen (2015). Since that in 96.70% of the reviews it is possible to find positive information, together with the conclusions taken by other articles like the ones presented, it can be inferred that the reviews written online in the TripAdvisor platform regarding Bermuda, are able to positively persuade potential tourists to choose the island as a travel destination.
To end this chapter, the considerations about the Negative aspects regarding Bermuda shall be pointed out. Again, as normal a load of adjectives were found to be a part of the final output as a way of reviewers to classify whatever concept they were mentioning. From the results it can be stated that the majority of concepts that pops out are related with the left side of the TEM (Re-Discover or Pure Pleasure), standing out the fact that the island is more of a place of Recreational Activities than Exploratory ones. The last information worth mentioning is the 32.28% of negative information being referred across the 635 reviews extracted, which again is quite significant.
According to the same article used in the last chapter towards the Negative aspects code, Chen (2015) states that negative information about hedonic products (such as travel destinations) will shrink the reliability of the same in the reader’s perspective. This is a fact that the responsible entities of Bermuda shall have in consideration, due to the tourism problems that it might create.
5.4 BERMUDA (Study 2)
Doing the same analysis as in the last chapter, but this time regarding the results of Bermuda’s text mining process, it is possible to say that from the 16 themes that the Meaning Cloud under the Text Classification analysis, was able to identify, there are two of them that are far more present in the extracted reviews: Lifestyle and Leisure and Tourism, Travel and Commuting. Doing a parallel connection with the results obtained from the Topics Extraction, it is possible to see that the major 20 concepts can be framed into the two mentioned themes:
1. Lifestyle and Leisure: “Beach”, “Service”, “Store”, “Food”, “Room”, “Restaurant”, “Hotel”, “Staff”, “Family” and “Dinner”;
2. Tourism, Travel and Commuting: “Bermuda”, “Hamilton”, “Place”, “Location”, “Island”, “Area”, “Cave”, “Boat” and “Cruise”.
From the 20 concepts considered as the major ones obtained by the Topics Extraction, all of them can be included as belonging to both themes, showing the importance of those in the total of reviews extracted.
The results of this analysis on the Category coding appear to be on the same page as the information mentioned before. Lifestyle and Leisure appears to be linked with all the categories studied, while Tourism, Travel and Commuting in three of the four categories. Moreover, it can be found Art and Culture linked with Cultural Activities and Sports with Leisure Activities. The information resultant of the Topics Extraction serves as a complement to the Text Classification:
1. Cultural Activities: appear to be connected with Tourism, Travel and Commuting and Lifestyle and Leisure – “Bermuda”, “Beach”, “Place” and “Room”;
2. Leisure Activities: again, the major topics extracted seem to be connected with Tourism, Travel and Commuting and Lifestyle and Leisure – “Bermuda”, “Food”, “Staff” and “Beach”;
3. Hotel: the same analysis as before – “Beach”, “Room”, “Bermuda” and “Place”;
4. Restaurant: besides Lifestyle and Leisure, it appears to be connected with Environment, Weather and Energy – “Beach”, “Cave”, “Hamilton” and “Dolphin”.
On the Tourism Experience Model level, the results of the Text Classification also maintain close results with what was presented before. Lifestyle and Leisure appears to be an important theme for all types of experiences, while Tourism, Travel and Commuting seem to be significant for three of them. Arts and Culture appears in two experiences, and Sports and Social in only one of them. It is possible to see that also with the major concepts found by the software for each experience:
1. Holist: was found to be linked with Lifestyle and Leisure, as well as Art and Culture – “Service”, “Dinner”, “Church” and “Town”;
2. Knowledge Seeker: more involved with Tourism, Travel and Commuting and Lifestyle and Leisure – “Bermuda”, “Place”, “Beach”, “Room” and “Hamilton”;
3. Pure Pleasure: indicates to be more connected with Lifestyle and Leisure, but also with Tourism, Travel and Commuting – “Room”, “Beach”, “Food” and “Bermuda”;
4. Re-Discover: linked with Lifestyle and Leisure, Tourism, Travel and Commuting, but also Social – “Beach”, “Bermuda”, “Staff”, “Food”, “Hamilton”, “Restaurant”, “Friendly”, “Room”, “Service” and “Location”.
At last, through the Membership Level of Reviewers it was possible to see that both types of reviewers tend to write about the same themes – Lifestyle and Leisure, Tourism, Travel and Commuting and Art and Culture, but the Low Ranked reviewers also include Sports and Social in their reviews – see attachment X.
From the result hereby analysed – Text Classification and Topics Extraction, it is possible to say that each coding seems to culminate in the same conclusions. The most important themes mentioned by the reviewers in their reviews on TripAdvisor about Bermuda, are related with Lifestyle and Leisure, Tourism, Travel and Commuting, and a little bit of Art and Culture. These three themes are the base of the e-WOM that the reviewers are creating online regarding Bermuda and their experiences in the island.
The next discussion to be having is related with the results obtained from the Sentiment Analysis performed – Global Sentiment Analysis and Topics Sentiment Analysis.
Firstly, the general results of the Global Sentiment Analysis indicate that globally the reviewers were feeling positive when writing about their stays and experiences in the island, as an average polarity of 4.10 was obtained. A score like this indicates a positive sentiment towards the destination.
Moreover, on a general level it can be said when the reviewers are in agreement with the content they are mentioning, they are feeling more positive sentiments, than when they are in disagreement – Agreement (4.35) and Disagreement (3.79). Another interesting fact is that, when the reviewers are subjective – letting their personal feelings and ideas influence what they are saying, they also tend to be more positive than when being objective – Subjective (4.11) and Objective (4.05). Finally, the irony analysis indicates that when being ironic, the reviewers are also happier and feel more positive sentiments than when not being ironic – Ironic (4.21) and Non-ironic (4.10).
Moving on to the Category coding, the overall results indicate that the Hotel category was the one with the biggest satisfaction associated, with an average polarity of 4.18, while the Restaurant one, was the most negative according to the results (4.01). The Hotel was actually the category with the most positive sentiments attributed across the several tests – Agreement, Objective, Subjective, Ironic and Non-Ironic. Only the Restaurant category got the highest average polarity under the reviews in Disagreement, but it was in fact the category with the most negative sentiments in the reviews in Agreement, Subjective, Ironic and Non-Ironic. Regarding the Tourism Experience Model, the Re-Discover experience was the one that globally contained the reviews with the most positive sentiments (4.12), while the Holist experience the most negative ones (3.83). This time it is possible find results more spread across the different aspects studied, but it is important to refer that again, even the sentiments considered more negative are still pretty positive in all experiences. Finally, the results of the Membership Level of Reviewers coding indicate that the Low Ranked reviewers obtained more positive sentiments in all of the aspects studied in comparison with the High Ranked ones.
At last, the Sentiment Analysis indicates that the reviewers in Bermuda were felt the most positive sentiments regarding matters of Science, Technology and Services, and the most negative ones related to Religion, Travel and Measures. One important fact, is that even being the most negative ones, these last three topics are still above the neutral score, and therefore are still positive.
Regarding the Category coding, the topics related with the Hotel category were the ones that the reviewers were more positive about (3.55) and the Leisure Activities the more negative ones (3.47). In attachment X it is possible to see the most important and least important topics according to each category. Moreover, the Tourism Experience Model tells us that the Holist experience is the one containing the most positive topics, while the Knowledge Seeker one the lowest (Attachment X). Finally, towards Bermuda the High Ranked ones obtained the most positive experiences in the island when in comparison with the Low Ranked ones. The most important topics for the HRR are Lifestyle, Services and Technology, and the least important are Tourism, Religion and Travel. The LRR obtained the best experiences in topics like Sciences, Social Events and Technology, and the most negative ones with Travel, Weather and Meteorology and Measures.
6. CONCLUSIONS AND IMPLICATIONS
The next chapter is structured in three different subchapters: Discussion, Theoretical and Managerial Implications and Limitations and Future Research. It is important to refer that, besides the literature already studied, the Discussion was based on an analysis grounded bu the results achieved – such analysis can be found in Annex X.

6.1 DISCUSSION
The literature has pointed out multiple times the importance of the online decision making, during a customer buying process, being influenced by User Content Generated (UCG) – namely Online User Reviews (OCR) (Zhou & Duan, 2016), as well as that content being an important source of information in the tourism industry for travellers (Pan, et al., 2007). The amount of reviews obtained in August of 2017 in the TripAdvisor platform regarding Madeira and Bermuda indicates a strong commitment by the online community, and it has potential to impact on the online travel booking (Peng & Chen, 2013).
The first study (Study 1) lays down on a netnographic process and aims studying the reviews extracted from each island in eight different codings: rating, category, membership level of reviewers, language type, tourism experience model, content, symbology and positive/negative information.
Regarding the Madeira’s results there are multiple data that shows how positive the reviewers have evaluated their stays in the island – average rating of 4.59, plus every category being classified over 4 and especially the fact that in 98.78% of the reviews it is possible to find positive information. Such facts are believed to persuade readers to purchase the destination by increasing their booking intention (Chen ; Tseng, 2011; Cheng, et al., 2017). Moreover, the fact that negative information is present in 33.45% of the reviews can be a positive factor in enhancing the reviewers credibility (Schuckert, et al., 2015; Zhong ; Daniel Leung, 2013), as well as firms being able to identify possible services or products imperfections (Chen ; Tseng, 2011).
The type of language mostly used by the reviewers is literal (vs figurative) – 79.27%, which is perceived by authors as being more useful for the readers (Fang, et al., 2016), thus increasing the purchase buying intention (Ashraf, et al., 2016). Also, the reviewers are mainly Low Ranked – 61.15%, which according to the literature indicts they are more willing to accept the information (Chen, 2015), but having LRR using figurative language – 22.51% of the reviews, may be a threat to the tourism of Madeira, once potential consumers tend to lower their booking intention against reviews with such characteristics (Wu, et al., 2017). The results of the Tourism Experience Model indicate that it is a place where people tend to perform recreational activities in order to get closer to their existential being (Gnoth ; Matteucci, 2014).
Concerning the Bermuda’s results there are also numerous indicators that led us to believe the reviewers’ experiences in the island was positive: an average rating of 4.43, all of the categories were evaluated with a rating above 4 and the fact of existing positive information in 96.70% of the reviews that were extracted. According to the literature studied these results point out that the e-WOM presented online in TripAdvisor is capable of persuading the readers in increasing the customer’s booking intention (Chen & Tseng, 2011; Cheng, et al., 2017). Moreover, it was found negative information about Bermuda in 32.28%, which can be turned into a positive factor by increasing the reviewers’ credibility (Schuckert, et al., 2015; Zhong ; Daniel Leung, 2013), and allowing brands/firms to re-adjust their strategies based on the flaws presented (Chen ; Tseng, 2011).
The percentage of reviewers using literal (vs figurative language) was massive – 80.31%, growing the usefulness of the review in the eyes of the readers (Fang, et al., 2016), consequently intensifying the customer’s buying intention (Ashraf, et al., 2016). The vast majority of reviewers were classified in the TripAdvisor platform as Low Ranked – 56.69%, which according to the literature states that the reviewers are more agreeable in accepting other’s information (Chen, 2015). A problem might arise due to having 21.11% of the reviews written by LRR with figurative language, as the literature points out that it can reduce the customer’s booking intention (Wu, et al., 2017). The results of the Tourism Experience Model were identical to the ones already mentioned in the Madeira’s case. According to the literature studied and the results obtained, Bermuda is perceived as a destiny where people, especially through recreational activities tend to get closer to their existential being (Gnoth ; Matteucci, 2014).
The second study (Study 2) contributes to the existing literature in a different scenario, as it results of a Text Mining process, and its’ objective was to determine the main themes being mentioned by the reviewers – Text Classification and Topics Extraction, as well as studying their sentiments through the reviews extracted – Sentiment Analysis.
Relating to the Madeira’s results, the Text Classification and the Topics Extraction showed very close results, as fifteen out of the twenty most important concepts analysed by the Topics Extraction, are appropriate in the two most important mentioned themes: Lifestyle and Leisure and Travel, Tourism and Commuting. The results of such finding have therefore made an important contribution to the literature, by highlighting the main themes of conversation among reviewers, and their importance in the creation of e-WOM about the destination. The results of the Category and Tourism Experience Model among these analysis are also congruent with the information given above. It was concluded that there is no significant difference between the themes mentioned between LRR and HRR.
Moving on to the Global Sentiment Analysis, it is clear that the sentiments demonstrated by the reviewers are extremely positive: an average polarity of 4.68; even the reviewers who manifested disagreement showed positive sentiments; every score above 3 in every coding and component (Agreement, Irony and Subjectivity) studied. Such extreme sentimentalism demonstrated, outcomes in reviews considered as more useful (Fang, et al., 2016), influencing positively the customer’s decision making (Ashraf, et al., 2016). A higher polarity influences, in a higher level, potential tourists of a destination (Alaei, et al., 2017), and having in mind that positive customer satisfaction is capable of transforming those potential customers in real ones (Chang, et al., 2009; Ribbink, et al., 2004), it can be infered that the reviews extracted are capable of converting readers in customers of a destination. In fact, the literature also shows that in order to increase the booking intention, destinations are dependent of positive emotions being demonstrated in UCG (Wang, 1999), as happens in the Madeira’s case. The findings also prove that despite reviewers express disagreement, they continue to recommend the destination and informing potential customers, has pointed out in the literature (Bilro, 2017).
It is also possible to say that these reviewers are engaged with the destination, by their role as active and co-creational members of this online community (Brodie, et al., 2011; Kumar, et al., 2010), as well as the customer satisfaction demonstrated (Chang, et al., 2009; Ribbink, et al., 2004). There are also indications of having loyal visitors of the island not only through the customer satisfaction or the co-creational role, but also by the fact that it is possible to find recommendations to others – e.g: Study 1 – Positive Information Results, spreading positive e-WOM (Islam, et al., 2012; Zeithaml, et al., 1996).
Lastly, the Madeira’s results showed through the Topics Sentiment Analysis that the topics with the higher positive connotation – according to the reviewers, were: Hotel, Services and Gastronomy; while the most negative ones were: Measures, Tourism and Weather and Meteorology. As it was already mentioned, this analysis shows which aspects of the island are negatively impacting the tourists in a higher preponderance, and allows the responsible entities of the Tourism of Madeira to identify these imperfections and re-develop their strategies (Chen & Tseng, 2011). The results also showed that the Restaurant category contains the topics being most positively evaluated – sentimentaly, while the Cultural Activities the worst in terms of positive sentiments. This is also linked with the experiences lived by the reviewers, as Cultural Activities are more involved with Knowledge Seeker it makes sense that it arises as the experience with the least positive sentiments involved, while Pure Pleasure was the one containing the best topics experiences. Regarding the last coding, the results show that the Low Ranked reviewers appear to be more satisfied with the topics that they mention online (3.56), but the difference is not huge in comparison with the High Ranked ones (3.49).
Moving on to the Bermuda’s results obtained in Study 2, the Text Classification and the Topics Extraction assume an even higher preponderance, as twenty out of twenty major topics analysed by the Topics Extraction, are suitable for the two most mentioned themes in the reviews extracted: Lifestyle and Leisure and Travel, Tourism and Commuting. These findings allow us to make a contribution to the literature, as it has identified the two major themes of discussion between reviewers online regarding Bermuda. Moreover, the results obtained in the Category and Tourism Experience Model codings appear to be on the same page as the general information, given that the two themes above mentioned are linked toall categories and almost every experience. The difference of themes being mentioned by the two types of reviewers is not significant, as the majority is the same.
The results of the Global Sentiment Analysis indicates that the reviewers felt positive sentiments: an average polarity of 4.10; the reviewers even manifesting disagreement in any part of their reviews showed sentiments above neutral; every score, generally, is above 4 except for disagreement (3.79). It is therefore possible to conclude that readers will find the reviews as more useful into making a decision (Fang, et al., 2016), which should be a positive one (Ashraf, et al., 2016). Also, has seen in the literature, the higher the average polarity score, more probable it is that the reviews will influence potential tourists of a destination, as well as the higher the probability of turning potential customers into real ones (Alaei, et al., 2017; Chang, et al., 2009; Ribbink, et al., 2004). It is possible to conclude that the reviews extracted, through the results obtained, are capable of acting as a positive mediator into bringing new tourists into the island. A positive score in the reviews considered as in disagreement (3.79) also confirms that despite not being 100% in favor of everything, the positive customer satisfaction makes reviewers continue to recommend the destination and spread information to potential customers (Bilro, 2017).
Furthermore, the results obtained – a clear positive customer satisfaction, together with the fact that reviewers playa part in an online community (as TripAdvisor) with an active and co-creational role, allows to conclude that the reviewers are engaged with this destination (Brodie, et al., 2011; Kumar, et al., 2010; Chang, et al., 2009; Ribbink, et al., 2004). User engagement plays a crutial role in influencing the loyalty of a brand, especially through online communities – like TripAdvisor (Zheng, et al., 2015). A sign of e-loyalty is the recommendation of the destination to others, by spreading positive e-WOM, as it is verified in this analysis and Study 1 (e.g: Study 1 – Positive Information Results) (Islam, et al., 2012; Zeithaml, et al., 1996).
At last, the Topics Sentiment Analysis demonstrate that the topics being mentioned online showing the most positive sentiments were: Sciences, Technology and Services; while the most negative sentiments ones were: Religion, Travel and Measures. Such analysis allows the respective entities of the Tourism of Bermuda to take action into re-adapting their strategies and “repair” the topics causing less positive sentiments in the tourists (Chen ; Tseng, 2011). The results of the Category coding states that Hotel is the category with the topics causing the most positive sentiments in the reviewers, while Leisure Activities the most negative ones. On the Tourism Experience Model level, the Holist experience contains the best topics while Knowledge Seeker the worst. Regarding the Membership Level of Reviewers the results were opposite to the Madeira, as High Ranked ones (3.52) felt highly satisfied than the Low Ranked ones (3.49).
This study also allows to perform a comparison between the results and conclusions obtained for both the islands of Madeira and Bermuda. The two following paragraphs are the last point of order of this Discussion:
Firstly, from the reviews extracted in August of 2017 in the TripAdvisor platform regarding both islands, it was concluded that the netnographic process rised some common conclusions, as the set of both reviews extracted are able to persuade readers into increasing their purchase decision, as well as increase their booking intention. The negative content is similar and allows the respective entities of both islands to re-define their strategies if necessary. The type of language used is also similar in both cases, reviewers clearly have a tendency to use literal (vs figurative) language, as well as being Low Ranked (vs High Ranked), which may be a threat to both destinations as potential customers tend to lower their reservation intention after reading a figurative language review written by a Low Ranked reviewer. Also, both destinations position themselves identical in the Tourism Experience Model – top left quadrant, which indicates that these islands are perceived by the reviewers as places where they can get closer to their existential being through recreational activities. Given the amount of positive information obtained vs negative one, it was also possible to conclude that in both cases, reviewers are tending to promote positively the destinations, as well as the brands and firms that operate in that industry. Lastly, regarding emojis/emoticons, it was clear that there is a clear preference to use them owing to positive content (vs negative), but its significance, in relation to the number of reviews, is pratically non-existent.
To conclude, the text mining analysis under the Meaning Cloud platform through the Text Classification and Topics Extraction processes indicate that there are two themes that are clearly under the scope of the reviewers when mentioning each island in the online community: Lifestyle and Leisure and Tourism, Travel and Commuting. In the Bermuda’s perspective all of the 20 major topics extracted fit in one of those themes (or both), while in the Madeira’s results 15 of the 20 major topics are suitable in one, or both, of those themes. Other themes assume an important secondary role, in the Madeira’s case it is important to refer Art and Culture and Environment, Weather and Energy, while in Bermuda’s reviews Art and Culture. The results of the Global Sentiment Analysis in both islands were quite positive, but in the Madeira’s reviews the sentiment attributed by the reviewers was higher. According to the literature studied it was possible to conclude that both set of reviews can be considered useful for potential customers. In the Madeira’s case, and having in mind that the results achieved a higher average polarity, it can be stated that the set of reviews extracted in this case is capable of influencing in a higher level potential tourists of a destination into becoming real ones, in comparison with Bermuda. Moreover due to the extreme sentimentalist obtained in the Madeira’s reviews, and following the same logic, the customer booking intention is higher, when compared with Bermuda. A common ground was obtained when Disagreement reviews are concerned, as reviewers despite showing relutance in a determined subject still continue to spread positive information, creating good content e-WOM. The reviewers, in both cases, were perceived as being engaged with the destinations, which together with the customer satisfaction obtained, and their co-creational role in the TripAdvisor community indicates having “loyal customers” – inducing that the strategies of “customer re-purchases”, performed by the entities responsible for the tourism of both destinations, are being correctly done.

6.2 THEORETICAL AND MANAGERIAL IMPLICATIONS
This chapter intends to transmit clearly which are the implications given by this study on a theoretical and managerial level. Relating the first, it will follow a structure of new methologies used; extension of the literature that it reached; and application of existing literature on a new perspective. Regarding the second, it will give directions to the responsible entities of the tourism of both destinations based on the results obtained.
Theoretically, and on a first basis, this study has given a unique perspective on studying together, both the islands of Madeira and Bermuda on a reviewer’s perspective. Combining the Kozinets netnographic process – with a set of unique codings as: rating, category, membership level of reviewers, language type, tourism experience model, content, symbology and positive/negative information; with a text mining process (namely, Text Classification, Topics Extraction and Sentiment Analysis) was also a new approach to the existing literature.
Secondly, the Tourism Experience Model created by (Deans ; Gnoth, 2012) was used in a unique way, as this study is able to provide an innovative approach in characterizing the two island destinations according to the different experiences proposed by the model.
Thirdly, the structure of this study can be used for further research using any set of destinations, as well as it is a flexible methodology, which allows, for instance, to change the codings according to the objective of the project.
Managerially speaking, this thesis – based on the results obtained, is capable of providing useful and suitable insights for the tourism entities of both destinations.
The first advice given is to promote Low Ranked Reviewers into High Ranked ones, due to the fact that the language effect is attenuated when reviews are written by the second (Wu, et al., 2017) – this strategy might overcome lowering the customer’s booking intention caused by LRR using figurative language. It could be done through a partnership between the refered entities of each island and the firms responsible for the Social Networking Sites.
Secondly, based on the results of the Text Classification and Topics Extraction, it is important that the communication/promotion strategies of the destinations (and the firms and brands that operate in the tourism industry) are based on Lifestyle and Leisure and Tourism, Travel and Commuting. Other themes might also play an important secondary role in these campaigns, such as Art and Culture and Environment, Weather and Energy.
Thirdly, one consideration when promoting the islands should include the type of experiences that the island is known and sought by. The fact that both islands are mainly experienced through their Recreational Activities, makes it a key element when promoting it to potential tourists.
In fourth, and last place, this study permits firms to recognize the negative information circulating through the TripAdvisor online community regarding a destination, which allows to re-define strategies into turning that negative information into positive. On the other hand, the already positive information being mentioned, can also be analysed to reinforce the strengths of the destinations.
6.3 LIMITATIONS AND FUTURE RESEARCH
The last subchapter of the Conclusions and Implications concerns Limitations and Future Research. In a strategic point of view it is possible to say that the limitations are avenues for future investigation. The following enlisted topics present the current limitations of this study, as well as possible future researches associated:
1. This study was limited to a certain amount of reviews presented in August of 2017 in the TripAdvisor platform, which provide a limited amount of data (1783 reviews). In future research the sample obtained, for Madeira and Bermuda, may be enlarged to provide more accurate results;
2. The conducted study aimed merely two island destinations, in order to obtain more complete and rigorous results, the amount of island destinations should be increased;
3. This study based its’ results using only one online community – TripAdvisor, which may or may not have skewed the results. In future research, more than one Social Networking Site should be used, to provide a comparison between the results obtained in each one of them;
4. In future research, the methodology used can be applied to any destinations that are intended to be studied – which means that it is possible to study any destinations presented in the TripAdvisor platform using this methodology;
5. Lastly, the methodology of this study was limited to two different techniques: Netnography and Text Mining. In future research further techniques may be used to confront the results – e.g: questionnaries, interviews, etc., as well as form a new set of techniques to study these islands, or other destinations.

9. REFERENCES
Alaei, A. R., Becken, S. ; Stantic, B., 2017. Sentiment Analysis in Tourism: Capitalizing on Big Data. Journal of Travel Research, Volume n.d, pp. 1-17.
Anderson, R. E. ; Srinivasan, S. S., 2003. E-Satisfaction and E-Loyalty: A Contingency. Psychology ; Marketing, 20(2), p. 123–138.
Ashraf, A. R., Thongpapanl, N. ; Spyropoulou, S., 2016. The connection and disconnection between e-commer businesses and their customers: Exploring the role of engagement, perceived usefulness and perceive ease-of-use. Eletronic Commerce Research and Applications, Volume 20, pp. 69-86.
Barger, V., Peltier, J. W. ; Schultz, D. E., 2016. Social media and consumer engagement: a review and research agenda. Journal of Research in Interactive Marketing, 10(4), pp. 268-287.
Bermuda Tourism Authority, 2018. 2017 Visitor Arrivals Report, Bermuda: GotoBermuda.
Berry, L. L., Carbone, L. P. ; Haeckel, S. H., 2002. Managing the Total Customer Experience. MIT Sloan Management Review, 43(3), pp. 1-6.
Bigné, L. ; Gnoth, J., 2005. The Theme Park Experience: An Analysis of Pleasure, Arousal and Satisfaction. Tourism Management, Volume 26, pp. 833-844.
Bijmolt, T. H. A. et al., 2010. Analytics for Customer Engagement. Journal of Service Research, 13(3), pp. 341-356.
Bilro, R. G., 2017. The role of consumer-brand engagement in a digital marketing era. ISCTE-IUL – Instituto Universitário de Lisboa, p. 83.
Biran, A. ; Hyde, K., 2013. New perspectives on dark tourism. International Journal of Culture, Tourism and Hospitality Research, 7(3), pp. 191-198.
Biran, A., Poria, Y. ; Oren, G., 2011. Sought experiences at (dark) heritage sites. Annals of Tourism Research, 38(3), pp. 820-841.
Bolton, R. N., 1998. A Dynamic Model of the Duration of the Customer’s Relationship with a Continuos Service Provider: The Role of Satisfaction”. Marketing Science, 17(1), pp. 45-65.
Bolton, R. N. & Drew, J. H., 1991. A Multistage Model of Customers’ Assessment of Service Quality and Value. Journal of Consumer Research, 17(4), pp. 375-384.
Bosque, I. ; Martín, H., 2008. Tourist Satisfaction a Cognitive-Affective Model. Annals of Tourism Research, 35(2), pp. 551-573.
Bowden, J., 2009. The process of customer engagement: A conceptual framework. Journal of Marketing Theory and Practice, 17(1), p. 63–74.
Brakus, J. J., Schmit, B. H. ; Zarantonello, L., 2009. Brand Experience: What Is It? How Is It Measured? Does It Affect Loyalty?. Journal of Marketing, Volume 73, pp. 52-68.
Brodie, R. J., Hollebeek, L. D., Juric, B. ; Ilic, A., 2011. Customer Engagement: Conceptual Domain, Fundamental Propositions, and Implications for Research. Journal of Service Research, 14(3), pp. 252-271.
Castro, J., 2013. Search Engine Journal. Online
Available at: https://www.searchenginejournal.com/applying-aida-digital-marketing/67755/
Chang, H. H., Wang, Y.-H. ; Yang, W.-Y., 2009. The impact of e-service quality, customer satisfaction and loyalty on e-marketing: Moderating effect of perceived value. Total Quality Management, 20(4), p. 423 –443.
Chen, C. C. ; Tseng, Y.-D., 2011. Quality evaluation of product reviews using an information quality framework. Decision Support Systems, Volume 50, pp. 755-768.
Cheng, I. C. C., Lam, L. W., Chow, C. W. ; Fong, L. H. N., 2017. The effect of online reviews on hotel booking intention: The role of reader-reviewer similarity. Journal of Hospitality Management, Volume 66, pp. 54-65.
Chen, J., 2015. Examining the influence of membership levels on consumers’ decision making. Procedia Computer Science, Volume 55, pp. 1258-1264.
Cheung, C. M., Lee, M. K. & Rabjohn, N., 2008. The impact of eletronic word-of-mouth: the adoption of online opinions in online customer communities. Internet Research, 18(3), pp. 229-247.
Cohen, E., 1979. A phenomenology of tourist experiences. Journal of the British Sociological Association, 13(2), pp. 179-201.
Confente, I., 2015. Twenty-Five Years of Word-of-Mouth Studies: A Critical Review of Tourism Research. International Journal of Tourism Research, Volume 17, pp. 613-624.
Deans, K. R. & Gnoth, J., 2012. Developing the tourism experience model. Otago, Australian & New Zealand Marketing Academy.
Direção Regional de Estatística da Madeira, 2017. DREM. Online
Available at: https://estatistica.madeira.gov.pt/en/download-now-3/economic/turismo-gb/turismo-noticias-gb/tourism-press-release-current-gb/1553-14-02-2018-in-2017-madeira-s-hotel-industry-registered-new-records-in-overnight-stays-and-revenue.html
Accessed 22 April 2018.
Dodds, P. S. et al., 2015. Human language reveals a universal positivity bias. Proceedings of the National Academy of Sciences, 12(8), pp. 2389-2394.
Doorn, J. v. et al., 2010. Customer Engagement Behavior: Theoretical Foundations and Research Directions. Journal of Service Research, 13(3), pp. 253-266.
Duffy, L. & Overholt, J., 2013. Seeking authenticity: re-conceptualizing adventure tourism. Illuminare: A Student Journal in recreation, Parks and Leisure Studies, 11(1), pp. 45-59.
Eastin, M. S., Daugherty, T. & Burns, N. M., 2011. Handbook of Research on Digital Media and Advertising: User Generated Content Consumption. New York: Information Science Reference.
Fang, B., Ye, Q., Kucukusta, D. & Law, R., 2016. Analysis of the perceived value of online tourism reviews: Influence of readability and reviewer characteristics. Tourism Management, Volume 52, pp. 498-506.
Fan, W., Wallace, L., Rich, S. & Zhang, Z., 2006. TAPPING THE POWER OF TEXT MINING. COMMUNICATIONS OF THE ACM, Volume 49, pp. 77-82.
Feldman, R., 2013. Techniques and Applications for Sentiment Analysis. Communications of the ACM, 56(4), pp. 82-89.
Fernandes, T. & Esteves, F., 2016. Customer Engagement and Loyalty: A Comparative Study Between Service Contexts. SERVICES MARKETING QUARTERLY, 37(2), p. 125–139.
Forbes, K. A., 2018. bermuda-online.org. Online
Available at: http://www.bermuda-online.org/climateweather.htm
Accessed 23 April 2018.
Gentile, C., Spiller, N. & Noci, G., 2007. How to Sustain the Customer Experience: An Overview of Experience Components that Co-create Value with the Customer. European Management Journal, 25(5), pp. 395-410.
Geyskens, I., Jan-Bennedict, S. E. & Nirmalya, K., 1998. Generalizations About Trust in Marketing Channel Relationships Using Meta-Analysis. InternationalJournal of Research in Marketing, 15(3), pp. 223-248.
Gilmore, J. H. & Pine, B. J., 2002. Differentiating hospitality operations via experiences: Why selling services is not enough, s.l.: Cornell Hotel and Restaurant Administration Quarterly.
Gnoth, J. & Deans, K., 2012. Developing the tourism experience model. Otago, Otago University.
Gnoth, J. & Matteucci, X., 2014. A phenomenological view of the behavioural tourism research literature. INTERNATIONAL JOURNAL OF CULTURE, TOURISM AND HOSPITALITY RESEARCH, 8(1), pp. 3-21.
Gobinath J, D. G. & Vidyapeetham, A. V., 2016. Online Reviews: Determining the Perceived Quality of Information. Jaipur, India, Intl. Conference on Advances in Computing, Communications and Informatics (ICACCI).
Goldenberg, J., Libai, B. & Muller, E., 2001. Talk of the network: a complex systems look at the underlying process of word-of-mouth. Marketing Letters, 12(3), pp. 211-223.
Grewall, D., Levy, M. & V.Kumar, 2009. Customer Experience Management in Retailing: An Organizing Framework. Journal of Retailing, Volume 85, pp. 1-14.
Hearst, M., 2003. What is Text Mining?, Berkeley: Japan Advanced Institute of Science and Technology.
Herskovitz, S. & Crystal, M., 2010. The Essential Brand Persona: Storytelling and Branding. Journal of Business Strategy, 31(3), pp. 21-28.
Hippner, H. & Rentzmann, R., 2006. Text Mining. Informatik-Spektrum, 29(4), p. 287–290.
Hirschman, E. C. & Holbrook, M. B., 1982. Hedonic Consumption: Emerging Concepts, Methods and Propositions. Journal of Marketing, Volume 46, pp. 92-101.
Hollebeek, L. D., Glynn, M. S. & Brodie, R. J., 2014. Consumer Brand Engagement in Social Media: Conceptualization, Scale Development and Validation. Journal of Interactive Marketing, Volume 28, p. 149–165.
Howard, J. A. & Sheth, J. N., 1969. The Theory of Buyer Behavior. Journal of the American Statistical Association, pp. 467-487.
Humberstone, B., 2011. Embodiment and social and environmental action in nature-based sport: spiritual spaces. Leisure Studies, 30(4), pp. 495-512.
Islam, M., Khadem, M. & Sayem, A., 2012. Service quality, customer satisfaction and customer loyalty analysis in Bangladesh apparel fashion retail: an empirical study, International Journal of Fashion Design. Technology and Education, 5(3), pp. 213-224.
Kaur, A. & Chopra, D., 2016. Comparison of text mining tools. Noida, India, 2016 5th International Conference on Reliability, Infocom Technologies and Optimization (ICRITO) (Trends and Future Directions).
Kem, Z. Z., Sesia, Z. J., Chung, C. M. & Matthew, L. K., 2014. Examining the influence of online reviews on consumers’ decision-making: A heuristic–systematic model. Decision Support Systems,, Volume 67, pp. 78-89.
Kirilenko, A. P., Stepchenkova, S. O., Kim, H. ; Li, X. (., 2017. Automated Sentiment Analysis in Tourism: Comparison of Approaches. Journal of Travel Research, Volume n.d, pp. 1-14.
Kirilenko, A. P., Stepchenkov, S. O., Kim, H. ; Li, X. (., 2017. Automated Sentiment Analysis in Tourism: Comparison of Approaches. Journal of Travel Research, 00(0), pp. 1-14.
Kozinets, R. V., 2015. Netnography: Redefined. 2 ed. London: Sage.
Kozinets, R. V., Dolbec, P.-Y. ; Earley, A., 2014. Netnographic Analysis: Understanding Culture through Social Media Data. In: U. Flick, ed. Sage Handbook of Qualitative Data Analysis. London: Sage, pp. 262-275.
Kumar, V. et al., 2010. Undervalued or Overvalued Customers: Capturing Total Customer Engagement Value. Journal of Service Research, 13(3), pp. 297-310.
Kusumasondjaja, S., Shanka, T. ; Marchegiani, C., 2012. Credibility of online reviews and initial trust: the roles of reviewer’s identity and review valence. Journal of Vacation Marketing, 18(3), pp. 185-195.
Lean, G., 2012. Transformative travel: a mobilities perspective. Tourist Studies, 12(2), pp. 151-172.
Lee, M. & Youn, S., 2009. Eletronic word of mouth (eWOM): How eWOM platforms influence consumer product judgement. International Journal of Advertising , 28(3), pp. 473-499.
Lemon, K. N. & C.Verhoef, P., 2016. Understanding Customer Experience Throughout the Customer Journey. Journal of Marketing: AMA/MSI Special Issue, Volume 80, pp. 69-96.
Liu, B. & Zhang, L., 2012. A SURVEY OF OPINION MINING AND SENTIMENT ANALYSIS. In: C. C. Aggarwal & C. Zhai, eds. MINING TEXT DATA. Charu C. Aggarwal • ChengXiang Zhai ed. New York: Springer: Springer, pp. 415-463.
LLC, M., 2018. meaningcloud. Online
Available at: https://www.meaningcloud.com/solutions/media-analysis
Accessed 19 August 2018.
Loureiro, S. M. C., 2015. The Role of Website Quality on PAD, Attitude and Intentions Visit and Recommend Island Destination. International Journal of Tourism Research, 17(6), pp. 545-554.
Lu, W. & Stepchenkova, S., 2012. Ecotourism experiences reported online: Classification of satisfaction attributes. Tourism Management, Volume 33, pp. 702-712.
Madeira, V., 2018. Visit Madeira. Online
Available at: http://www.visitmadeira.pt/pt-pt/a-madeira
Accessed 22 April 2018.
Maitland, R. & Ritchie, B., 2009. City Tourism: national capitals perspective. Wallingford: CABI.
McIntosh, A., 1999. Into the tourist’s mind: understanding the value of the heritage experience. Journal of Travel ; Tourism Marketing, 8(1), pp. 41-64.
Mendes, J. d. C., Valle, P. O. d., Guerreiro, M. M. ; Silva, J. A., 2010. The tourist experience: Exploring the relationship between tourist satisfaction and destination loyalty. ORIGINAL SCIENTIFIC PAPER, 58(2), pp. 111-126.
Meyer, C. ; Schwager, A., 2007. Understanding Customer Experience. Harvard Business Review, 85(2), pp. 117-126.
Mont, O. ; Plepys, A., 2003. Customer satisfaction: review of literature and application to the product-service systems, Lund, Sweden: International Institute for Industrial Environmental Economics at Lund University.
Moore, S. G., 2015. Attitude Predictability and Helpfulness in Online Reviews: The Role of Explained Actions and Reactions. Journal of Consumer Research, Volume 42, pp. 30-44.
Mou, J., Shin, D.-H. ; Cohen, J., 2017. Understanding trust and perceived usefulness in the consumer acceptance of an e-service: a longitudinal investigation. Behaviour ; Information Technology, 36(2), pp. 125-139.
Norman, A., 2012. The varieties of the spiritual tourist experience. Literature ; Aesthetics, 22(1), pp. 20-37.
Oliver, R. L., 2010. Satisfaction: A Behavioral Perspective on the Consumer. Second Edition ed. Armonk, New York: M.E. Sharpe, Inc.
Otto, J. E. ; Ritchie, J. R. B., 1996. The service experience in tourism. Tourism Management, 17(3), pp. 165-174.
Pan, B., Maclaurin, T. ; Crotts, J. C., 2007. Travel Blogs and the Implications for Destination Marketing. Journal of Travel Research, Volume 46, pp. 35-45.
Peng, X. H. ; Chen, W., 2013. Tourist Behaviour in Online Booking: A new research agenda. Communications in Information Science and Management Engineering, 3(6), pp. 280-285.
Peterson, R. A., 1995. Relationship Marketing and the Consumer. Journal of the Academy of Marketing Science, 23(4), pp. 278-281.
Pucinelli, N. M. et al., 2009. Customer Experience Management in Retailing: Understanding the Buying Process. Journal of Retailing, Volume 85, pp. 15-30.
Purnawirawan, N., Pelsmacker, P. D. ; Dens, N., 2012. Balance and Sequence in Online Reviews: How Perceived Usefulness Affects Attitudes and Intentions. Journal of Interactive Marketing, Volume 26, p. 244–255.
Ramya, N. ; Ali, S. M., 2016. Factors affecting consumer buying behavior. International Journal of Applied Research, 2(10), pp. 76-80.
ReviewTrackers, 2018. reviewtrackers.com. Online
Available at: https://www.reviewtrackers.com/online-reviews-survey/
Accessed 16 September 2018.
Ribbink, D., Riel, A. C. v., Liljander, V. ; Streukens, S., 2004. Comfort your online customer: quality, trust and loyalty on the internet. Managing Service Quality, 14(6), pp. 446-456.
Ritchie, J. ; Crouch, G., 1997. Roles and Contributions to Destination Competitiveness. Cha-Am, Thailand, Proceedings of the 47rd Congress of the Association International d’Experts Scientifi ques du Tourisme.
Rose, S., Clark, M., Samouel, P. & Hair, N., 2012. Online Customer Experience in e-Retailing: An empirical model of Antecedents and Outcomes. Journal of Retailing, 88(2), pp. 308-322.
RTP, 2018. Rtp.pt. Online
Available at: https://www.rtp.pt/madeira/economia/madeira-bate-recorde-de-turistas-em-2017-_15696
Accessed 22 April 20128.
Russell, A. & Levy, S., 2012. The temporal and focal dynamics of volitional reconsumption: a phenomenological investigation of repeated hedonic experiences. Journal of Consumer Research, 39(2), pp. 341-359.
Sáenz-Navajas, M.-P., Ballester, J., Peyron, D. & Valentin, D., 2014. Extrinsic attributes responsible for red wine quality perception: A cross cultural study between France and Spain. Food Quality and Preference.
Schmitt, B., 1999. Experiental Marketing. Journal of Marketing Management, 15(1-3), pp. 53-67.
Schmitt, B., Brakus, J. J. & Zarantonello, L., 2015. From experiential psychology to consumer experience. Journal of Consumer Psychology, Volume 1, pp. 116-171.
Schmunk, S., Hopken, W., Fuchs, M. & Lexhagen, M., 2014. Sentiment Analysis: Extracting Decision-Relevant Knowledge from UGC. In: X. P. & T. I., eds. Information and Communication Technologies in Tourism 2014. New York: Springer, pp. 253-265.
Schuckert, M., Liu, X. & Law, R., 2015. HOSPITALITY AND TOURISM ONLINE REVIEWS: RECENT TRENDS AND FUTURE DIRECTIONS. Journal of Travel & Tourism Marketing, Volume 32, p. 608–621.
Shaikh, T. & Deshpande, D., 2016. A Review on Opinion Mining and Sentiment Analysis. IJCA Proceedings on National Conference on Recent Trends in Computer Science and Information Technology, NCRTCSIT 2016(2), pp. 6-9.
Sheth, J. N. & Parvatlyar, A., 1995. Relationship Marketing in Consumer Markets: Antecedents and Consequences. Journal of the Academy of Marketing Science, 23(4), pp. 255-271.
Sheth, J. N., Sisodia, R. S. & Sharma, A., 2000. The Antecedents and Consequences of Customer-Centric Marketing. Journal of the Academy of Marketing Science, 28(1), pp. 55-66.
Smith, C., 2018. DMR: Business Statistics. Online
Available at: https://expandedramblings.com/index.php/tripadvisor-statistics/
Accessed 16 September 2018.
Soliman, H. S., 2011. Customer Relationship Management and Its Relationship to the Marketing Performance. International Journal of Business and Social Science , 2(10), pp. 166-182.
Spinakis, A. & Chatzimakri, A., 2005. Comparative Study of a Text Mining Tools. Some Industrial Applications of Text Mining, 232(2005), pp. 223-232.
Sreejesh, S. & Ponnam, A., 2017. Investigating the process through which e-servicescape creates e-loyalty in travel and tourism websites. Journal of Travel & Tourism Marketing, 34(1), pp. 20-39.
Srinivasan, S. S., Andersona, R. & Ponnavolu, K., 2002. Customer loyalty in e-commerce: an exploration of its antecedents and consequences. Journal of Retailing, Volume 78, pp. 41-50.
Srivastava, A., Sahami, M. & Kumar, V., 2009. Text Mining: Classification, Clustering and Applications. New York: Chapman and Hall/CRC.
Stebbins, A., 2007. Serious Leisure: A Perspective for our time. New Brunswick, NJ: Transaction.
Thompson, C. J., Locander, W. B. & Pollio, H. R., 1989. Putting Consumer Experience Back into Consumer Research: The Philosophy and Method of Existential-Phenomenology. Journal of Consumer Research, 16(2), pp. 133-146.
TripAdvisor, 2017. TripAdvisor. Online
Available at: https://tripadvisor.mediaroom.com/us-about-us
Accessed 22 April 2018.
TripAdvisor, 2018. TripAdvisor. Online
Available at: https://www.tripadvisor.pt/Tourism-g189165-Madeira_Islands-Vacations.html
Accessed 22 April 2018.
TripAdvisor, 2018. TripAdvisor. Online
Available at: https://www.tripadvisor.co.za/Tourism-g147255-Bermuda-Vacations.html
Accessed 23 April 2018.
Turner, V., 1973. The centre out there: Pilgrim’s goal. History of Religions, 12(1), pp. 191-230.
Uriely, N., 2005. The Tourist Experience: Conceptual Developments. Annals of Tourism Research, 32(1), p. 199–216.
Valdivia, A., Luzón, M. V. ; Herrera, F., 2017. Sentiment Analysis in TripAdvisor. IEEE Computer Society, Volume 17, pp. 1541-1672.
Vega, A., Casielles, R. ; Martín, A., 1995. La Calidad Percibida Del Servicio en Estabelecimientos Hoteleros de Turismo Rural. Papers de Turisme, Volume 19, pp. 17-33.
Verhoef, P. C. et al., 2009. Customer Experience Creation: Determinants, Dynamics and Management Strategies. Journal of Retailing, 85(1), pp. 31-41.
Verhoef, P. C., Reinartz, W. J. ; Krafft, M., 2010. Customer Engagement as a New Perspective in Customer Management. Journal of Service Research, 13(3), pp. 247-252.
Wang, N., 1999. RETHINKING AUTHENTICITY IN TOURISM EXPERIENCE. Annals of Tourism Research, 26(2), pp. 349-370.
Wilson, E. ; Harris, C., 2006. Meaningful travel: women, independent travel and the search for self and meaning. TOURISM, 54(2), pp. 161-172.
WordAtlas, 2018. worldatlas. Online
Available at: https://www.worldatlas.com/webimage/countrys/namerica/bermuda/bmland.htm
Accessed 22 April 2018.
World Tourism Organization, 2017. UNWTO e-library. Online
Available at: https://www.e-unwto.org/doi/pdf/10.5555/unwtotfb0060010020122016201712
Accessed 25 April 2018.
Wu, L., Shen, H., Fan, A. ; Mattila, A. S., 2017. The impact of language style on consumers0 reactions to online reviews. Tourism Management, Volume 59, pp. 590-596.
Yan, Q. et al., 2016. E-WOM from e-commerce websites and social media: Which will consumers adopt?. Electronic Commerce Research and Applications, Volume 17, pp. 62-73.
Ye, Q., Law, R. ; Gu, B., 2009. The impact of online user reviews on hotel room sales. International Journal of Hospitality Management, Volume 28, pp. 180-182.
Ye, Q., Zang, J. ; Law, R., 2009. Sentiment Classification of Online Reviews to Travel Destinations by Supervised Machine Learning Approaches. Expert Systems with Applications, 36(3), pp. 6527-6535.
Yoo, C., Sanders, G. ; Moon, J., 2013. Exploring the effect of e-WOM participation on e-Loyalty in e-commerce. Decision Support Systems, Volume 55, pp. 669-678.
Zaglia, M. E., 2012. Brand communities embedded in social networks. Journal of Business Research, Volume 66, pp. 216-223.
Zahra, A. ; McIntosh, A., 2007. Volunteer tourism: evidence of cathartic tourist experiences. Tourism Recreation Research, 32(1), pp. 115-119.
Zeithaml, A., Berry, L. ; Parasuraman, A., 1996. The Behavioral Consequences of Service Quality. Journal of Marketing, 60(2), pp. 31-46.
Zhang, K. Z., Zhao, S. J., Cheung, C. M. ; Lee, M. K., 2014. Examining the influence of online reviews on consumers’ decision-making: A heuristic–systematic model. Decision Support Systems, Volume 67, p. 78–89.
Zheng, X., Cheung, C. M., Lee, M. K. & Liang, L., 2015. Building brand loyalty through user engagement in online brand communities in social networking sites. Information Technology & People, 28(1), pp. 90-106.
Zhong, L. & Daniel Leung, R. L. &. B. W., 2013. eTourism in China: A Review of the Chinese Language Literature. Asia Pacific Journal of Tourism Research, 18(5), pp. 464-482.
Zhou, W. & Duan, W., 2016. Do Professional Reviews Affect Online User Choices Through User Reviews? An Empirical Study. Journal of Management Information Systems, 33(1), pp. 202-228.

Post Author: admin