Multiple-Choice Question Answering Models for Automatic Depression Severity Estimation

Depression is one of the most prevalent mental health diseases. Although there are effective treatments, main problem relies on providing early and risk detection. Medical experts use self-reporting questionnaires to elaborate their diagnosis, but these have some limitations. Social stigmas lack awareness often negatively affect success self-report questionnaires. This article aims describe techniques automatically estimate depression severity from users social media. We explored pre-trained language models over subject’s writings. addressed task “Measuring Severity Signs Depression” eRisk 2020, an initiative in CLEF Conference. In this task, participants fill Beck Questionnaire (BDI-II). Our proposal explores application Multiple-Choice Question Answering (MCQA) predict user’s answers BDI-II questionnaire using posts These MCQA built BERT (Bidirectional Encoder Representations Transformers) architecture. results showed that multiple-choice question answering could be a suitable alternative for estimating degree, even when small amounts training data available (20 users).