The Semantic Scholar Academic Graph (S2AG)
Alex D. Wade
Abstract
The Semantic Scholar Academic Graph, or S2AG (pronounced ”stag”), is a large, open, heterogeneous knowledge graph of scholarly works, authors, and citations that powers the Semantic Scholar discovery service. S2AG currently contains over 205M publications, 121M authors, and nearly 2.5B citation edges. Semantic Scholar integrates metadata from Crossref, PubMed, Unpaywall, and other sources. In addition, through partnerships with academic publishers and through web-crawling, we source and process the full-text of nearly 60M full-text publications in order extract and classify the document structure, including references, citation contexts, figures, tables, and more. S2AG is available via an open API as well as via downloadable monthly snapshots. In this talk, we will describe the S2AG resource as well as the Semantic Scholar Open Research Corpus, or S2ORC (pronounced ”stork”), a general purpose, multi-domain corpus for NLP and text mining research.