AAGCN-ACSA
AAGCN-ACSA is a Python implementation of the Beta Distribution Guided Aspect‑aware Graph model for Aspect Category Sentiment Analysis with affective knowledge. The framework builds aspect‑aware graphs using Beta‑distributed weights derived from external affective commonsense sources such as SenticNet and ConceptNet. It provides both a non‑BERT version and a BERT‑based version, and requires Python 3.6, PyTorch 1.0, SpaCy and NumPy. Users train the model with the scripts train.py or trainbert.py, selecting the architecture via the --model flag. The code supports research in sentiment analysis and can be adapted for related natural language processing tasks involving aspect extraction.