Mental Health Treatments Using an Explainable Adaptive Clustering Model

Abstract

AbstractIn this paper, we propose a model to help psychologists identify the most important aspects of emotions in mentally ill people. As a first step, we created emotional lexicon embeddings using natural language processing, followed by deep clustering based on the attention mechanism. The generated representation is used to evaluate patient-written text from an emotional perspective. To increase the patient authored emotional lexicon, we apply synonymous semantic expansion. The EANDC method is used to classify latent semantic representations according to context. This is an explainable attention network based adaptive deep clustering approach. To increase the explainability of the learning, we used similarity metrics to select the instances that label the text to optimize the following selection using the curriculum-based optimization technique. Experimental results show that increasing the number of synonyms based on emotion lexicons improves accuracy without negatively affecting performance.KeywordsNLPExplainableText clusteringHealthy treatments