DASentimental: Detecting Depression, Anxiety, and Stress in Texts via Emotional Recall, Cognitive Networks, and Machine Learning

Ying Li,Massimo Stella,Thomas Trenholm Hills,Asra Fatima

doi:10.3390/bdcc5040077

Ying Li, Massimo Stella + Show 2 more

Open Access

https://doi.org/10.3390/bdcc5040077

Copy DOI

Abstract

Most current affect scales and sentiment analysis on written text focus on quantifying valence/sentiment, the primary dimension of emotion. Distinguishing broader, more complex negative emotions of similar valence is key to evaluating mental health. We propose a semi-supervised machine learning model, DASentimental, to extract depression, anxiety, and stress from written text. We trained DASentimental to identify how N = 200 sequences of recalled emotional words correlate with recallers’ depression, anxiety, and stress from the Depression Anxiety Stress Scale (DASS-21). Using cognitive network science, we modeled every recall list as a bag-of-words (BOW) vector and as a walk over a network representation of semantic memory—in this case, free associations. This weights BOW entries according to their centrality (degree) in semantic memory and informs recalls using semantic network distances, thus embedding recalls in a cognitive representation. This embedding translated into state-of-the-art, cross-validated predictions for depression (R = 0.7), anxiety (R = 0.44), and stress (R = 0.52), equivalent to previous results employing additional human data. Powered by a multilayer perceptron neural network, DASentimental opens the door to probing the semantic organizations of emotional distress. We found that semantic distances between recalls (i.e., walk coverage), was key for estimating depression levels but redundant for anxiety and stress levels. Semantic distances from “fear” boosted anxiety predictions but were redundant when the “sad–happy” dyad was considered. We applied DASentimental to a clinical dataset of 142 suicide notes and found that the predicted depression and anxiety levels (high/low) corresponded to differences in valence and arousal as expected from a circumplex model of affect. We discuss key directions for future research enabled by artificial intelligence detecting stress, anxiety, and depression in texts.

Highlights

Depression, anxiety, and stress are the three negative emotions most likely to be associated with psychopathological consequences [1,2,3]
We provide key results about several models using different combinations of network distances, further enriching the Emotional Recall Task (ERT) data with features coming from network navigation of semantic memory
Using valence and arousal of words expressed in suicide letters, we performed an additional validation of the results of DAS through the circumplex model of affect [11]

Summary

Introduction

Depression, anxiety, and stress are the three negative emotions most likely to be associated with psychopathological consequences [1,2,3]. Stress is associated with difficulty relaxing, agitation, and impatience (cf [1,5,6]) While these emotions are discrete and highly complex [3,6], they vary along a primary and culturally universal dimension of valence: perceived pleasantness [2,7,8]. Valence is often the only output of affect scales This is potentially problematic for measuring depression, anxiety, and stress, which are more complex [10]: These three distinct types of psychological distress are similar in valence but differ widely along other dimensions, such as arousal [11]. Depression, anxiety, and stress are difficult to distinguish using valence alone [3,12,13] This underlines the need for richer mappings between emotional dimensions and depression, anxiety, and stress (DAS) levels

Methods

Results

Discussion

Conclusion