shibing624/similarity

Provides hierarchical similarity algorithms across word, phrase, sentence, and paragraph granularities using methods like synonym lexicon encoding, morpho-syntactic analysis, and cosine similarity with TF-IDF weighting. Integrates HowNet semantic primitives for word-level sentiment analysis and Word2Vec embeddings for synonym recommendation, with lazy-loaded models and customizable training on user corpora. Designed as a modular, low-coupling NLP toolkit targeting Chinese text processing with plain-text dictionary distribution for transparency.

1,571 stars.