ThanhHung2112/Semantic_chunking

The library computes sentence embeddings using Sentence Transformers models, then groups consecutive sentences by cosine similarity to detect semantic boundaries—splitting only when similarity drops below a configurable threshold. It provides tunable parameters for maximum chunk size and similarity thresholds, enabling fine-grained control over segmentation granularity for downstream NLP tasks like retrieval-augmented generation or document analysis.

Used by 1 other package. No commits in the last 6 months. Available on PyPI.