re-search/DocProduct

Medical Q&A with Deep Language Models

62
/ 100
Established

Combines fine-tuned BioBERT encoders with FAISS vector search for retrieving relevant medical information, then conditions a fine-tuned GPT-2 generator on retrieved context to produce answers. Built on TensorFlow 2.0 and trained on 700,000+ medical Q&A pairs from Reddit, HealthTap, and WebMD, with embedding-space metric learning via feedforward network heads to align question and answer representations.

571 stars and 10 monthly downloads. No commits in the last 6 months. Available on PyPI.

Stale 6m
Maintenance 0 / 25
Adoption 12 / 25
Maturity 25 / 25
Community 25 / 25

How are scores calculated?

Stars

571

Forks

157

Language

Jupyter Notebook

License

MIT

Last pushed

Mar 25, 2023

Monthly downloads

10

Commits (30d)

0

Dependencies

14

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/re-search/DocProduct"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.