Behavioral Sentiment Analysis of Depressive States

Abstract

The need to release accurate and incontrovertible diagnoses of depression has fueled the search for new methodologies to obtain more reliable measurements than the commonly adopted questionnaires. In such a context, research has sought to identify non-biased measures derived from analyses of behavioral data such as voice and language. For this purpose, sentiment analysis techniques were developed, initially based on linguistic characteristics extracted from texts and gradually becoming more and more sophisticated by adding tools for the analyses of voice and visual data (such as facial expressions and movements). This work summarizes the behavioral features accounted for detecting depressive states and sentiment analysis tools developed to extract them from text, audio, and video recordings.