CS 411 FORM 1
THESIS TITLE PROPOSAL
Proponent/s MORADOS, CHEENIE REYES 1510173141
Proposed Title Monitoring Depression of Students through Twitter Using Natural Language Processing Techniques
Field of PTitle Natural Language Processing
Background of the Study
Depression is a common mental disorder that negatively affects individual’s feelings, acts and how people think. Depression is now up to date; every individual who experienced depression thought that suicide would be the solution to escape. According to World Health Organization (WHO) In the Philippines, the suicide rate for men is 2.5 for every 10,000 populations and 1.7 for women because of depression. Although that there is a low suicide rate with other ASEAN nations, depression is still persistent. Another report showed 4.5 million Filipinos could be burdened with depression.
As social media turns out to be increasingly instilled in everyday lives, it’s easy to relate each platform to our physical environments. Social media are web based communication tools that empower individuals to collaborate with each other by both sharing and consuming information, one of this is Twitter, Twitter is progressively explored as a method for recognizing emotional well-being status, including melancholy and suicide, in the population, Twitter could be a resolved construct exclusively in light of the substance of the post, as judged by human coders and after that repeated by machine learning.
Machine learning forms are consequently; connected to evaluated whether an unequivocally concerning tweet could be distinguished naturally. The computer classifier accurately recognized 80% of ’emphatically concerning tweets and demonstrated expanding picks up in exactness; be that as it may, future enhancements are vital as a level was not came to as the measure of information expanded. The present examination exhibited that it is conceivable to recognize the level of worry among suicide-related tweets, utilizing both human coders, and a programmed machine classifier. Essentially, the machine classifier imitated the precision of the human coders. The discoveries affirmed that Twitter is utilized by people to express suicidality and that such posts evoked a level of worry that justified further examination.
The main objective of this study is to develop, create and evaluate a Monitoring Depression of Students through Twitter Using Natural Language Processing Techniques
? To develop a system that will be serving as a basis for monitoring depression.
? To create a system that can identify at risk-tweets and hence at-risk users to prevent suicides.
? To evaluate a system that will monitor the depression of students through twitter.
Significance of the Study
The results of this study can be utilized to assist the specialists to find a focused group of at-risk users and for raising a caution so that the relevant authorities could take necessary interventions to further analyse the predicted user to affirm his/her state of mental health.
Review of Related Literature
They developed this system to detect at-risk users or at-risk tweets to prevent suicides. This study used user-level classifier and tweets-classifier as well to detect depression. This study used Twitter user’s history as a tool to identify whether the users is depressed or not. In this study, the researchers used natural language processing and machine learning to easily detect at-risk users and tweets. (Jamil, 2017).
This study intends to contemplate the attributes of online depression communities. The analysis utilized numerous social media one of these is Twitter. They also used Psycholinguistic process and content topics extracted from the past. This study focuses on history. (Nguyen, Phung, Bo Dao, Venkatesh, 2015).
They developed this system to annotate scheme and corpus of depression-related tweets to identify depression to prevent suicides. They also used Twitter. In this study, they also used natural language processing to monitor major depressive disorder at the population level depending on the social media data. The researchers used machine learning such as – support vector machine, naïve bayes, and decision tree for predicting depression. (Mowery, Smith, Tyler, Craig and Conway, 2015).
They used hashtags to easily detect depression. Abdullah also used Twitter to determine who and how many Twitter users are depressed. (Abdullah, 2014).
In this research, the analysts focused on people who are in need a psychological guide to prevent them from suicide. In this study, they used Twitter to determine who and how many Twitter users are suicidal. (O’Dea, 2015).
The analysts used Natural Language Processing to influence induction about individual’s psychological states from what the users compose on Twitter or other social media. The researchers tokenized, highlight extraction and choice, parsing and machine learning orders. (Orooji, 2017). Most prominent types of web based life have been utilized as information hotspots for mental wellbeing applications. Number of clients, dialect (i.e. English) and accessibility of APIs increment the odds of a stage being utilized. Twitter is the most generally utilized wellspring of information chiefly in light of the fact that the accumulation of open information is simple. Facebook is moreover normal, frequently utilized by creators who additionally work for (or in association with) the organization. We just sought papers in English, and these for the most part discussed content written in English. A couple of special cases (e.g. Japanese) are said. It would appear that NLP in non-English dialects is an unexplored zone. This might be identified with the lower quality, or nonappearance of Natural Language Processing apparatuses in dialects other than English.
One of the most contemplated corpora of client produced writings is a suicide notes accumulation since it was utilized as a part of a common errand rivalry. Offer errands permit specialists from various controls to team up on a specific issue. (Calvo, David, Sazzad, Christensen, 2016) In particular, we directed a component removal concentrate to survey the education of each component gathering and an element disposal concentrate to decide the ideal capabilities for ordering Twitter tweets. We utilized a current, clarified Twitter dataset that was built in view of a various levelled model of despondency related manifestations. The dataset contains 9,473 comments for 9,300 tweets. Each tweet is commented on as no confirmation of sadness (e.g., “Natives fear a financial misery”) or proof of despondency (e.g., “discouraged over frustration”). On the off chance that a tweet is explained confirmation of gloom, at that point it is additionally commented on with at least one depressive side effects, for instance, discouraged disposition (e.g., “feeling down in the dumps”), bothered rest (e.g., “another anxious night”), or weariness or loss of vitality (e.g., “the exhaustion is excruciating”). For each class, each comment (9,473 tweets) is binaries as the positive class e.g., discouraged mood=1 or negative class e.g., not discouraged mood=0. (Mowery et al., 2016)
We implemented the Twitter gushing API to gather tweets containing key expressions produced from the APA’s rundown of hazard components and AAS’s rundown of caution signs identified with suicide. We arbitrarily researched the creators of these tweets to distinguish 60 bothered clients who every now and again composed about gloom, suicide, or self-mutilation. We too arbitrarily gathered 60 “regular” clients. An expert with mastery in emotional well-being research approved the determination of these upset and non-troubled clients. For each arrangement of clients, we gathered at most the last 50 tweets to make a database of 5,446 tweets, of which 2,381 are from bothered clients and 3,065 are from ordinary clients. (Vioules, Moulahi, Bringay, 2018)
We utilized the Twitter gushing API to gather tweets containing key expressions produced from the APA’s rundown of hazard components and AAS’s rundown of caution signs identified with suicide. We arbitrarily researched the creators of these tweets to distinguish 60 bothered clients who every now and again composed about gloom, suicide, or self-mutilation. We too arbitrarily gathered 60 “regular” clients. An expert with mastery in emotional well-being research approved the determination of these upset and non-troubled clients. For each arrangement of clients, we gathered at most the last 50 tweets to make a database of 5,446 tweets, of which 2,381 are from bothered clients and 3,065 are from ordinary clients. (Coppersmith, Dredze, Harman, 2014).
In this investigation, Nadeem, Horn, Coppersmith assembled data from the Shared Errand coordinators of the CLPsych 2015 gathering. This dataset was created from an amalgamation of clients with open Twitter accounts who posted an announcement as a articulation of finding, for example, “I was determined to have X today”, where X would speak to either despondency or PTSD.For every client, up to 3000 of their latest open tweets were incorporated into the dataset, and every client was secluded from the others. It ought to be noticed that this 3000 tweet constrain gets from Twitter’s authentic policies, and that most tweets thought long following a two-month timespan may perhaps lower the adequacy of a classifier, as appeared by Tsugawa et al. (Nadeem, Horn, Coppersmith, 2017).
Jamil, Z., Inkpen, D., Buddhitha, P., ; White, K. (2017). Monitoring Tweets for Depression to Detect At-risk Users. Proceedings of the Fourth Workshop on Computational Linguistics and Clinical Psychology –- From Linguistic Signal to Clinical Reality. doi:10.18653/v1/w17-3104
Nguyen, T., Phung, D., Dao, B., Venkatesh, S., ; Berk, M. (2014). Affective and Content Analysis of Online Depression Communities. IEEE Transactions on Affective Computing, (3).
Mowery, D., Smith, H. A., Cheney, T., Bryan, C., ; Conway, M. (2016). Identifying Depression-Related Tweets from Twitter for Public Health Monitoring. Online Journal of Public Health Informatics, (1).
Peer Review #2 of “Use of Twitter to monitor attitudes toward depression and schizophrenia: An exploratory study (v0.1)”. (2014).
O’dea, B., Wan, S., Batterham, P. J., Calear, A. L., Paris, C., & Christensen, H. (2015). Detecting suicidality on Twitter. Internet Interventions, (2).
Calvo, R. A., Milne, D. N., Hussain, M. S., & Christensen, H. (2017). Natural language processing in mental health applications using non-clinical texts. Natural Language Engineering, 23(05), 649-685. doi:10.1017/s1351324916000383
Mowery, D., Smith, H., Cheney, T., Stoddard, G., Coppersmith, G., Bryan, C., & Conway, M. (2017). Understanding Depressive Symptoms and Psychosocial Stressors on Twitter: A Corpus-Based Study. Journal of Medical Internet Research, (2).
Paul, M. J., Sarker, A., Brownstein, J. S., Nikfarjam, A., Scotch, M., Smith, K. L., & Gonzalez, G. (2015). Social Media Mining For Public Health Monitoring And Surveillance. Biocomputing 2016.
Vioules, M. J., Moulahi, B., Aze, J., & Bringay, S. (2018). Detection of suicide-related posts in Twitter data streams. IBM Journal of Research and Development, (1).
Coppersmith, G., Dredze, M., & Harman, C. (2014). Quantifying Mental Health Signals in Twitter. Proceedings of the Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal to Clinical Reality.
Zuorba, H. D., Olan, C. L., & Cantara, A. D. (2017). A Framework for Identifying Excessive Sadness in Students through Twitter and Facebook in the Philippines. Proceedings of the International Conference on Bioinformatics Research and Applications 2017 – ICBRA 2017.
Liu, S., Zhu, M., Yu, D. J., Rasin, A., & Young, S. D. (2017). Using Real-Time Social Media Technologies to Monitor Levels of Perceived Stress and Emotional State in College Students: A Web-Based Questionnaire Study. JMIR Mental Health, (1).
Dr. ALBERT A. VINLUAN
Dean, College of Computer Studies
Dr. MARC P. LAURETA
Program Head, Department of Computer Science
Engr. MARC ANGELO P. MARTINEZ
Research Coordinator, College of Computer Studies