ScalaConsultants/Aspect-Based-Sentiment-Analysis

Implements a supervised "Professor" component that generates decision explanations by analyzing transformer self-attention patterns, enabling users to assess prediction reliability beyond raw sentiment scores. Built on BERT-based classification with a modular pipeline architecture (preprocessing → tokenization → encoding → model inference → review → postprocessing) that supports long-form documents via configurable text splitting and multi-word aspect definitions. Designed for production use with TensorFlow, SpaCy integration, and extensible pattern recognizers for custom explainability logic.

579 stars and 200 monthly downloads. Available on PyPI.