Improving Arabic Cognitive Distortion Classification in Twitter using BERTopic

Social media platforms allow users to share thoughts, experiences, and beliefs. These represent a rich resource for natural language processing techniques make inferences in the context of cognitive psychology. Some inaccurate biased thinking patterns are defined as distortions. Detecting these distortions helps restructure how perceive thoughts healthier way. This paper proposed machine learning-based approach improve distortions’ classi-fication Arabic content over Twitter. One challenges that face this task is text shortness, which results sparsity co-occurrence lack information (semantic features). The enriches rep-resentation by defining latent topics within tweets. Although classification supervised learning concept, enrichment step uses unsupervised learning. algorithm utilizes transformer-based topic modeling (BERTopic). It employs two types document representations performs averaging concatenation produce contextual embeddings. A comparative analysis F1-score, precision, recall, accuracy presented. experimental demonstrate our enriched representation outperformed baseline models different rates. encouraging suggest using distribution, obtained from BERTopic technique, can classifier’s ability distinguish between CD categories.