MaartenGr/KeyBERT
KeyBERT is a keyword extraction library that uses BERT embeddings to identify the most similar words and phrases to an entire document.
Not currently ranked — collecting fresh signals.
star history
It leverages transformer-based embeddings to create semantic representations of both documents and candidate phrases, then uses cosine similarity to rank phrases by their relevance to the source. The library supports configurable embedding models and includes techniques like Max Sum Distance and Maximal Marginal Relevance to improve diversity in extracted keywords.
Frequently asked
- What is MaartenGr/KeyBERT?
- KeyBERT is a keyword extraction library that uses BERT embeddings to identify the most similar words and phrases to an entire document.
- Is KeyBERT open source?
- Yes — MaartenGr/KeyBERT is open source, released under the MIT license.
- What language is KeyBERT written in?
- MaartenGr/KeyBERT is primarily written in Python.
- How popular is KeyBERT?
- MaartenGr/KeyBERT has 4.2k stars on GitHub.
- Where can I find KeyBERT?
- MaartenGr/KeyBERT is on GitHub at https://github.com/MaartenGr/KeyBERT.