Enhancing Sentiment Analysis for Autistic Children: A Hybrid Approach Using SBERT and Ensemble Learning

Abstract

Autistic children utilise both verbal and nonverbal means of communication. Children diagnosed with autism predominantly communicate their feelings through verbal expression rather than physical gestures or behaviours. Hence, this work suggests a blending model that integrates Sentence-BERT as a language embedding model with Voting Classifier as a machine learning ensemble model. The proposed mixture model enhances the efficacy of sentiment analysis models for autistic children by incorporating both sentence embedding and semantic weights. This research presents a prospective paradigm for analysing the emotions of children with autism.