How do falsehoods spread on the web? This and other questions related to the propagation of fake news and biased discourse in the public area have been drawing increasing interest in different communities from social sciences to artificial intelligence. Online discourse, i.e. claims and opinions shared online, together with their associated contexts (authors, salient entities, related events) constitute a valuable source of insights to these questions.
While knowledge graphs of today enable data reuse and federation, improving information retrieval on the web and facilitating knowledge discovery in various fields, they do not store information about claims and related online discourse data, making it difficult to access, query and reuse this wealth of information.
Hence, we have built ClaimsKG — a knowledge graph of fact checked claims and their related metadata, such as the claims’ truth labels, authors, topics, keywords and references. The knowledge graph is built by extracting information from a number of popular fact-checking web portals and structuring it following a specifically developed for this purpose conceptual data model mainly based on schema.org and the NIF vocabularies. Among other applications, ClaimsKG provides ground truth data for a number of tasks relevant to the analysis of societal debates on the web.
In this talk, I will first present three tools that operate on the data of ClaimsKG: (1) the Claims Explorer, which allows us to navigate through our data and perform faceted search; (2) the Claims Statistical Observatory, which helps uncover trends; and (3) the ClaimLinker, which allows to link arbitrary text to claims from ClaimsKG (or any other Knowledge graph containing claims), hence assisting the task of fact verification. I will further discuss perspectives on modeling claims in a generalized and contextualized manner, as well as related challenges such as claim disambiguation and the assessment of claim relatedness.