KeyBERT

About

A lightweight Python library that leverages transformer models to create document-level keyword and keyphrase embeddings.

Key Features

• Uses BERT-based embeddings for semantic similarity
• Supports multiple vectorizers like Scikit-Learn and Flair
• Customizable diversity and MMR (Maximal Marginal Relevance) algorithms

Pros

• Extremely easy to implement
• Context-aware keyword extraction
• High flexibility with model backends

Cons

• Resource-intensive for very large documents
• Requires some knowledge of Python and Transformers