A Novel Hybrid Fuzzy-WASPAS-ART Framework for Evaluating Emotional Intelligence and Communication Effectiveness Using LLM Empathy and BERT Deep Features

Evaluating emotional intelligence (EI) and communication ability from digital text is becoming increasingly crucial in education, mental health, and human-computer interaction. The present project offers a novel hybrid framework that combines fuzzy decision-making, sentiment analysis, empathy scoring, and deep semantic feature extraction into a comprehensive assessment of EI in written communication. Features from textual data are generated with BERT embeddings and reduced with principal component analysis (PCA), while VADER scores normalize sentiment and a LLaMA-based large language model score provides empathy scores. Fuzzy logic with triangular membership functions addresses the inherent uncertainty associated with emotional expression and translates the numerical scores into linguistic categories that can be more easily interpreted. The framework facilitates multi-criteria decision-making using the Weighted Aggregated Sum Product Assessment (WASPAS)-method and allows for integrated scores of sentiments, empathy, and semantic features. Results from our evaluation of a large emotion-labeled corpus suggest that empathy, as assessed by advanced language models, is the variable with the most significant negative impact on communication quality over and above sentiment. The framework allows for interpretable, easily scalable and actionable approaches to next-generation emotion-aware systems and informs the development and deployment of emotionally intelligent technology in the real world by opening avenues for benchmarking.