Hironsan/bertsearch

Converts document text into 768-dimensional BERT embeddings and stores them as dense vectors in Elasticsearch, enabling semantic similarity search beyond keyword matching. The Docker Compose setup orchestrates BERT inference and Elasticsearch services, with Python scripts handling vectorization of documents and index creation. Supports multiple pretrained BERT variants (Base/Large, cased/uncased, multilingual) with a Flask web interface for querying indexed documents.

898 stars. No commits in the last 6 months.