scite

Discovers relevant scientific articles by querying a proprietary indexed database of millions of papers using semantic search and citation context analysis. The system parses citation statements from papers to understand whether citations are supportive, contradictory, or methodological, enabling context-aware retrieval beyond keyword matching. Results are ranked by citation sentiment and relevance to the query.

Automatically analyzes citation statements within papers to classify whether each citation is supportive, contradictory, or methodological using trained NLP models. The system extracts citation context windows, applies multi-class classification, and assigns confidence scores. Results are surfaced in the UI with highlighted citation text and sentiment labels.

Extracts and enriches bibliographic metadata from scientific papers including authors, affiliations, publication date, journal, abstract, and keywords using OCR, PDF parsing, and entity extraction. The system normalizes author names, disambiguates affiliations, and links papers to external identifiers (DOI, PubMed ID, arXiv ID). Enriched metadata is stored and indexed for search and filtering.

Enables researchers to input a specific research claim or hypothesis and automatically retrieves papers that support, contradict, or provide methodological context for that claim. The system uses semantic matching to find relevant papers, then surfaces citation sentiment to show agreement/disagreement. Results are organized by evidence strength and citation count, creating an evidence map for the claim.

Provides a shared workspace where research teams can create, organize, and annotate collections of papers with collaborative features. Users can tag papers, add notes, highlight key findings, and share collections with team members. The system tracks changes, enables commenting on papers, and integrates with reference management tools. Collections are versioned and can be exported in standard formats.

Exposes REST and/or GraphQL APIs that allow developers to programmatically query the scite index, retrieve citation sentiment data, and integrate scite capabilities into external applications. APIs support filtering by citation sentiment, paper metadata, and date ranges. Rate limiting and authentication via API keys enable scalable access. Response formats include JSON with structured citation context and metadata.

Analyzes citation patterns and sentiment distributions across papers to identify research trends, consensus, and emerging disagreements in a field. The system aggregates citation sentiment data, tracks how citation patterns change over time, and identifies papers that are frequently cited with contradictory sentiment. Results are visualized as trend charts and consensus heatmaps showing agreement/disagreement over time.

Evaluates paper quality and reliability using multiple signals including citation sentiment distribution, citation count, author reputation, journal impact factor, and peer review status. The system aggregates these signals into a reliability score that indicates how much supporting evidence exists for a paper's claims. Scores are displayed alongside search results and in paper detail views.