SOTA Search Over Academic Publications

TL;DR

Exa has launched a new search feature for academic papers and technical publications.
It uses a dedicated index of approximately 350 million publications and 30 million authors.
The search functionality supports natural language queries, even for vague or incomplete memories of papers.
Exa's search is designed to work semantically, understanding the meaning of queries rather than relying solely on exact keywords or titles.
Evaluations show Exa achieving 82.8% recall on its benchmarks, outperforming competitors like Perplexity and Google Scholar.
The system's architecture involves OCR and document parsing for PDFs, combined with metadata and web indexing for comprehensive searching.
The search API is available for use with the 'publication' category.

Continuar leyendo el artículo original