Abstract—The ubiquity of social media provides a rich opportunity to enhance the data available to mental health clinicians and researchers, enabling a better-informed and better-equipped mental health field. Datasets originating from social networks are very valuable to many fields such as sociology and psychology. However, the supports from technical perspective are far from enough, and specific approaches are urgently in need. This paper applies data mining to psychology area for detecting depressed users in Twitter. Firstly, a sentiment analysis method is proposed utilising vocabulary and man-made rules to calculate the depression inclination of each micro-blog. We present analysis of mental health phenomena in publicly available Twitter data, demonstrating how rigorous application of simple natural language processing methods can yield insight into specific disorders as well as mental health writ large, along with evidence that as-of-yet undiscovered linguistic signals relevant to mental health exist in social media.
Introduction
The rise of online social network provides unprecedented opportunities for solving problems in a wide variety of fields with information techniques . While mental health issues pose a significant health burden on the general public, mental health research lacks the quantifiable data available to many physical health disciplines. This is partly due to the complexity of the underlying causes of mental illness and partly due to longstanding societal stigma making the subject all but taboo.This paper applies data mining techniques to psychology, specifically
the field of depression, to detect depressed users in social network services (SNS). The expansion of data mining to psychology is of great technical and social significance. It is proved that the proposed model in this paper could effectively help for detecting depressed ones and preventing suicide in online social networks. Moreover, population-level analysis via traditional methods is time consuming, expensive, and often comes with a significant delay. Of the numerous health topics for which social media has been considered, mental health may actually be the most appropriate. A major component of mental health research requires the study of behaviour, which may be manifest in how an individual acts, how they communicate, what activities they engage in and how they interact with the world around them including friends and family. Additionally, capturing population level behavioural trends from Web data has previously provided revolutionary capabilities to health researchers . Thus, social media seems like a perfect fit for studying mental health in both individual and overall trends in the population. How does a service like Twitter inform our knowledge in this area? Numerous studies indicate that language use, social expression and interaction are telling indicators of mental health Twitter and other social media provide a unique quantifiable perspective on human behavior that may otherwise go unobserved, suggesting it as a powerful tool for mental health researchers.
2 Related Work
2.1 Research of Depression in Psychology
Depression is the world’s fourth largest disease and will be in the second place in 2020 according to World Health Organisation statistics. The main clinical symptom of depressed patients is lasting depressed state of mood and lack of positive emotions. They prefer to be alone rather than together with others. What’s more, most of depressed patients suffer from chronic insomnia.
The research of depression in social network in psychology comes in two types: one is to discover disciplines of a crowd of depressed users the other
is to look into a specific case elaborately. Literature observes linguistic markers of depression through collecting posts by depressed and non-depressed
individuals from twitter. It analyses the text with LIWC, a computerised word counting tool, and shows that the online depressed writers use more
first person singular pronouns but less first person plural pronouns, more negative emotion words but less positive emotion words. Literature discusses the
relations between SNS behaviours and depression levels based on Zoufan Event. It is established by questionnaire and statistic tools, and reveals that frequencies of the original posts could indicate micro-bloggers’ depressed levels. Also, the period of time users post twitter is a consideration as most depression patients suffer from chronic insomnia.
These researches of depression features in the perspective of psychology provide reliable background knowledge for our study. However, when comes to data analysis problems, only some simple statistic tools are designed for them, which
undoubtedly limit their researches. Therefore a specific data mining technique to detect depressed users is designed in this study based on their results.
Without converting to a PS, EPS, TIFF, PDF, or PNG file: Microsoft Word, Microsoft PowerPoint, or Microsoft Excel. Though it is not required, it is recommended that these files be saved in PDF format rather than DOC, XLS, or PPT. Doing so will protect your figures from common font and arrow stroke issues that occur when working on the files across multiple platforms. When submitting your final paper, your graphics should all be submitted individually in one of these formats along with the manuscript.
2.2 Sentiment Analysis Techniques
As the cardinal symptom of depression is severe negative emotions and lack of positive emotions, sentiment analysis is the most important step in depression detection. Sentiment analysis aims at mining users’ opinions and sentiment polarity from the texts they posted. Recently many progresses have been made in sentiment analysis on Twitter data. These researches include two aspects:
• Subject-independent analysis, namely judging the polarity of the tweets without considering if it is relevant to a subject. The main approaches are based on hashtags, smileys and some abstract features.
• Subject-dependent analysis, namely judging the polarity of the tweets based on the given subject. The sentiments of the tweets as positive, negative or neutral in, according to not only the abstract features but also the target-dependent features, which refers to the comments on the target itself and the related things. Little study has been made for solving
problems in a specific field, although analysis strategy differs a lot for different fields. For example, depression sufferers tend to think the topic about “death”, so this kind of words should be paid special attention to when constructing the vocabulary. Inspired by the work in literature, abstract features and target-dependent features are taken into account. This study stresses the particularity of depression and twitter content, and the whole model is specifically designed based on them.
3. Algorithm
Sentiment Analysis is a part of machine learning algorithm, works in R. We shall try to analyse the streamed data from Twitter,
Basically, We have two types of words: Positive words, and Negative words.
These positive and negative words are our Database.
Here, We pass a sample statement, and it is matched against the positive words.
Next, we try computing an aggregate score for each statement:
Example: I feel lonely and very depressed, but my friends will help me through it!
Score = (-1) + (+1) + (+1) = +ve
Therefore, we conclude that the above sample statement is a positive one!
4. Results
5. Conclusions
A model for detecting depressed users in social network based on sentiment analysis is proposed in this paper. The sentiment analysis method pays special attention to the characteristics of depression.We demonstrate quantifiable signals in Twitter data relevant to depression and negative words.
6. References
Aggarwal, C.C.: Social Network Data Analytics. Springer, New York (2011)
2. Hsieh, Y., Bolan, J.E.: Predicting Processing Difficulty in Chineses Syntactic Ambiguity Resolution: a Parallel Approach. Poster, The 84th Annual Meeting of the Linguistic Society of America, Baltimore, MD (2010)
3. World Health Organization, http://www.who.int/en/
4. Ramirez-Esparza, N., Chung, C.K., Kacewicz, E., Pennebaker, J.W.: The Psychology of Word Use in Depression Forums in English and in Spanish: Testing Two Text Analytic Approaches. In: Proceedings of the International Conference on Weblogs and Social Media, pp. 102–108. Menlo Park, CA:AAAI Press (2008)
5. Moreno, M., Jelenchick, L., Egan, K., Cox, E., Young, H.,Gannon, K.,etal.: Feeling Bad on Facebook: Depression Disclosures by College Students on Social Networking Site. Depression and Anxiety. 28, 447–455
6.http://ai2-s2-pdfs.s3.amazonaws.com/7398/1bfbe4fa535af965c9521e80f434afda9a5c.pdf
https://www.cs.jhu.edu/~mdredze/publications/2014_acl_mental_health.pdf