ClaimsKG: A Knowledge Graph of Fact-Checked Claims

Abstract

Various research areas at the intersection of computer and social sciences require a ground truth of contextualized claims labelled with their truth values in order to facilitate supervision, validation or reproducibility of approaches dealing, for example, with fact-checking or analysis of societal debates. So far, no reasonably large, up-to-date and queryable corpus of structured information about claims and related metadata is publicly available. In an attempt to fill this gap, we introduce ClaimsKG, a knowledge graph of fact-checked claims, which facilitates structured queries about their truth values, authors, dates, journalistic reviews and other kinds of metadata. ClaimsKG is generated through a semi-automated pipeline, which harvests data from popular fact-checking websites on a regular basis, annotates claims with related entities from DBpedia, and lifts the data to RDF using an RDF/S model that makes use of established vocabularies. In order to harmonise data originating from diverse fact-checking sites, we introduce normalised ratings as well as a simple claims coreference resolution strategy. The current knowledge graph, extensible to new information, consists of 28,383 claims published since 1996, amounting to 6,606,032 triples.