Detecting Depression in Social Media using Fine-Grained Emotions

Proceedings of NAACL-HLT 2019, pages 1481–1486 Minneapolis, Minnesota, June 2 - June 7, 2019. c⃝2019 Association for Computational Linguistics 1481 Detecting Depression in Social Media using Fine-Grained Emotions Mario Ezra Arag´on⋆, A. Pastor L´opez-Monroy†, Luis C. Gonz´alez-Gurrola‡ and Manuel Montes-y-G´omez⋆ ⋆ Instituto Nacional de Astrof´ısica, ´Optica y Electr´onica (INAOE), Mexico. † Centro de Investigaci´on en Matem´aticas (CIMAT), Mexico. ‡ Facultad de Ingenier´ıa, UACh, Mexico. mearagon@inaoep.mx, pastor.lopez@cimat.mx, lcgonzalez@uach.mx, mmontesg@inaoep.mx Abstract Nowadays social media platforms are the most popular way for people to share information, from work issues to personal matters. For ex- ample, people with health disorders tend to share their concerns for advice, support or sim- ply to relieve suffering. This provides a great opportunity to proactively detect these users and refer them as soon as possible to profes- sional help. We propose a new representation called Bag of Sub-Emotions (BoSE), which represents social media documents by a set of fine-grained emotions automatically generated using a lexical resource of emotions and sub- word embeddings. The proposed representa- tion is evaluated in the task of depression de- tection. The results are encouraging; the usage of fine-grained emotions improved the results from a representation based on the core emo- tions and obtained competitive results in com- parison to state of the art approaches. 1 Introduction Mental Disorders affect millions of people around the world. Out of these disorders, depression has been ranked among the most common, even with a high incidence in mortality rates (Kessler et al., 2017; Mathers and Loncar, 2006). It is imperative then, to come with effective approaches to detect depression before it causes irreparable damage to mere individuals that suffer it and their loved ones. In a connected world where we live, it is very normal to share personal information, matters and concerns in social media platforms. This fact poses an opportunity, since the understanding of depression through the analysis of social media documents increases the chances to detect people that present signs of depression and could lead to provide them professional help (Guntuku et al., 2017; Pestian et al., 2010). Several works in literature have explored how to use linguistic and sentiment analysis to detect depression (Xue et al., 2013). For example, in (Huang et al., 2014) the authors applied sentiment analysis (SA) to assign polarity to tweets. They count the number of positive, negative, neutral words, and the ratio of the negative and positive words, and found that depressed users post longer emotional tweets. The work of Wang et al. (2013) enriched SA with features derived from psycho- logical research like the use of first person pro- nouns, user social interaction and user behaviors in micro blogs. An interesting finding is that the time of the posts is useful to detect people with high risk of committing suicide. In a recent work, Chen et al. (2018) proposed to use emotions with the aim to identify depression on Twitter users. That study openly exposed the potential of using discrete emotions as features, instead of only using linguistic …

Content truncated to 500 words