根據統計, 在美國有18\%的人受精神疾病所苦, 精神疾病如同其他疾病一樣危及人們健康而且是需要被迫切解決的問題
隨著社群網站成為我們生活的一部分, 有許學者開始從資料科學的角度, 利用社群網站上豐富的資料來分析精神疾病, 並建立預測模型來偵測潛在的患者
在本研究中, 我們提出了一個可以有效蒐集患者資料的方法, 並利用情感分析技術來建立一個有效偵測患者的預測模型
我們提出的方法適用於各種精神疾病, 本研究則專注兩項有代表性的疾病: 躁鬱症以及邊緣性人格障礙

Mental disorders are currently affecting millions
of people – both offline and online – across the globe. The
challenge of mental disorders in general is that they are difficult
to detect on suffering patients. Conversely, someone can also
be misdiagnosed of having one type of mental illness when in
actuality they are suffering from another form of mental health
problem. In this paper, we aim at building a predictive model that
leverages language and behavioral patterns, used particularly
in social media, to determine whether a user is suffering from
two cases of mental disorder, namely Borderline Personality
Disorder (BPD) and Bipolar Disorder (Bipolar). This predictive model
is made possible by employing an effective data collection process,
coined as subconscious crowdsourcing, that helps to collect a
faster and more reliable dataset of patients, which is one of
the areas of improvement in this field of study. We also adopt
linguistic features used in previous several works and build on
top of those to improve the results obtained by our classification
model. Additionally, we apply sentiment analysis to foster further
understanding of a user’s emotional state with respect to a
mental disorder. Our experiments suggest that extracting specific
language patterns and social interaction features from reliable
patient datasets can greatly contribute to further analysis and
detection of mental disorders.

摘要: i
Abstract: ii
Acknowledgement: iii
List of Tables: viii
List of Figures: ix
1. Introduction: 1
2. Related Work: 4
3. Method: 8
3.1 Data Collecting: 9
3.2 Preprocessing: 10
3.3 Feature Extraction: 11
3.4 Classifier Training and Evaluation: 13
4. Experiment: 15
4.1 Data: 15
4.2 Prediction Performance: 16
4.3 Selection Bias Test: 18
4.4 Limited Tweets Test: 18
5. Demonstration: 20
6. Conclusion and Discussion: 22

[1] Carlos Argueta, Elvis Saravia, and Yi-Shin Chen. Unsupervised graph-based patternsextraction for emotion classification. InProceedings of the 2015 IEEE/ACM Interna-tional Conference on Advances in Social Networks Analysis and Mining 2015, pages336–341. ACM, 2015.[2] American Psychiatric Association et al.Diagnostic and statistical manual of mentaldisorders (DSM-5R). American Psychiatric Pub, 2013.[3] Murray R Barrick and Michael K Mount. The big five personality dimensions and jobperformance: a meta-analysis.Personnel psychology, 44(1):1–26, 1991.[4] David M Blei, Andrew Y Ng, and Michael I Jordan. Latent dirichlet allocation.theJournal of machine Learning research, 3:993–1022, 2003.[5] Glen Coppersmith, Mark Dredze, and Craig Harman. Quantifying mental health sig-nals in twitter. InProceedings of the Workshop on Computational Linguistics andClinical Psychology: From Linguistic Signal to Clinical Reality, pages 51–60, 2014.[6] Glen Coppersmith, Mark Dredze, Craig Harman, and Kristy Hollingshead. From24
adhd to sad: Analyzing the language of mental health on twitter through self-reporteddiagnoses.NAACL HLT 2015, page 1, 2015.[7] Munmun De Choudhury. Role of social media in tackling challenges in mental health.InProceedings of the 2nd international workshop on Socially-aware multimedia,pages 49–52. ACM, 2013.[8] Munmun De Choudhury, Michael Gamon, Scott Counts, and Eric Horvitz. Predictingdepression via social media. InICWSM, page 2, 2013.[9] Virgile Landeiro Dos Reis and Aron Culotta. Using matched samples to estimate theeffects of exercise on mental health from twitter. InProceedings of the Twenty-NinthAAAI Conference on Artificial Intelligence, 2015.[10] Glen A Coppersmith Craig T Harman and Mark H Dredze. Measuring post traumaticstress disorder in twitter. InIn ICWSM, 2014.[11] Matz S. Gosling S. Popov V. Kosinski, M. and D Stillwell. mypersonality project.http://mypersonality.org/wiki/doku.php?id=start/, 2012.[12] Andy Liaw and Matthew Wiener. Classification and regression by randomforest.Rnews, 2(3):18–22, 2002.[13] Margaret Mitchell, Kristy Hollingshead, and Glen Coppersmith. Quantifying the lan-guage of schizophrenia in social media.NAACL HLT 2015, page 11, 2015.[14] Megan A Moreno, Lauren A Jelenchick, Katie G Egan, Elizabeth Cox, Henry Young,Kerry E Gannon, and Tara Becker. Feeling bad on facebook: Depression disclosures
by college students on a social networking site.Depression and anxiety, 28(6):447–455, 2011.[15] Minsu Park, Chiyoung Cha, and Meeyoung Cha. Depressive moods of users portrayedin twitter. InProceedings of the ACM SIGKDD Workshop on healthcare informatics(HI-KDD), pages 1–8, 2012.[16] Sungkyu Park, Sang Won Lee, Jinah Kwak, Meeyoung Cha, and Bumseok Jeong.Activities on facebook reveal the depressive state of users.Journal of medical Internetresearch, 15(10), 2013.[17] James W Pennebaker, Roger J Booth, and Martha E Francis. Linguistic inquiry andword count: Liwc [computer software].Austin, TX: liwc. net, 2007.[18] Daniel Preotiuc-Pietro, Johannes Eichstaedt, Gregory Park, Maarten Sap, LauraSmith, Victoria Tobolsky, H Andrew Schwartz, and Lyle Ungar. The role of person-ality, age and gender in tweeting about mental illnesses.NAACL HLT 2015, page 21,2015.[19] Lenore Sawyer Radloff. The ces-d scale a self-report depression scale for research inthe general population.Applied psychological measurement, 1(3):385–401, 1977.[20] Camilo J Ruggero, Mark Zimmerman, Iwona Chelminski, and Diane Young. Border-line personality disorder and the misdiagnosis of bipolar disorder.Journal of psychi-atric research, 44(6):405–408, 2010.[21] Maarten Sap, Gregory Park, Johannes C Eichstaedt, Margaret L Kern, David Stillwell,