|Title||Analyzing Clinical Depressive Symptoms in Twitter|
|Publication Type||Conference Paper|
|Year of Publication||2016|
|Authors||Amir Hossein Yazdavar, Hussein S. Al-Olimat, Tanvi Banerjee, Krishnaprasad Thirunarayan, Jyotishman Pathak, Amit Sheth|
|Conference Name||23rd NIMH Conference on Mental Health Services Research (MHSR): Harnessing Science to Strengthen the Public Health Impact|
|Conference Location||Bethesda, MD|
Twitter provides a rich source for studying people’s mood in order to detect depressive behaviors. We developed a novel technique to unobtrusively analyzes individual posts in social media to detect signs of depression that can be utilized to build a proactive and automatic screening tool for early recognition of clinical depression. Leveraging clinical definition of depression, we build a depression lexicon that contains common depression symptoms determined by experts such as from the established clinical assessment questionnaires PHQ-9. We expanded the terms expressing the nine PHQ-9 depression symptoms categories using Urban Dictionary and Big Huge Thesaurus. The lexicon contains depression-related symptoms that are likely to appear in the tweets of individuals either having depressive-like symptoms or suffering from depression. A subset of highly informative seed terms are selected from this depression lexicon for crawling depression-related tweets. For each lexical term, we calculate its association with all of the variations of the term “depress” using Pointwise Mutual Information (PMI) and Chi-squared test to quantify their correlation and thereby rank order them. We leverage Twitris, our social media analysis platform, to study language, sentiment, emotions, topics and people content-network of depressed individuals.
|Full Text|| |