gsarti/covid-papers-browser

Enables semantic search across 44,000+ COVID-19 papers using domain-specific BERT models (SciBERT, BioBERT, CovidBERT) fine-tuned on natural language inference tasks via Sentence-Transformers to produce universal sentence embeddings. The interactive browser supports querying paper abstracts with cached embeddings for fast retrieval after initial GPU-accelerated corpus processing. Includes reproducible training workflows on SNLI/MultiNLI datasets and evaluation against the STS benchmark.

183 stars. No commits in the last 6 months.