eduard-mir/Sentiment-Polarity-Entropy: Entropy-Based Polarity Analysis with Human Annotation and Cross-Model Comparison

This release consolidates the full project pipeline: construction of the adjective polarity/entropy corpus, the entropy-based analysis distinguishing encoded from inferred polarity, and a new human-annotated benchmark (959 sentences sampled from 200 stratified adjectives) used to validate automatic classification. It also adds a cross-model comparison against the resulting Gold Human Label, evaluating pysentimiento/RoBERTuito, Cardiff NLP's twitter-xlm-roberta-base-sentiment, ChatGPT, and Claude across low/mid/high entropy bands.