CLASSIFYING EMPATHY IN TEXTUAL ANAMNESIS USING SINGLE LAYERED LSTM PERFORMANCE

Abstract

This study explores the classification of empathy in clinical conversations using Natural Language Processing (NLP) and a single-layer Long Short-Term Memory (LSTM) model. Empathy is a critical component of patient-centered care, yet detecting and analyzing it in text-based clinical anamnesis remains underexplored. The dataset, comprising transcribed clinical interviews annotated for affective and cognitive empathy, was processed using word embeddings combining Word2Vec and GloVe. An LSTM-based classifier was developed to identify empathetic expressions in text, achieving promising results with high accuracy and AUC scores. However, limitations include a simplified focus on two empathy dimensions and challenges due to dataset size and imbalance. These findings demonstrate the potential of LSTM models in understanding empathetic communication in healthcare, with implications for enhancing doctor-patient interactions through AI-driven insights.