With a grant from the SIDN Fund, the Utrecht Data School built the BOLD Profiler. This application detects bias in commonly used knowledge bases such as Wikidata.

Disinformation is not always spread with a deliberate plan, but can also (unintentionally) originate from incorrect information, missing data or misrepresentation. For example, Wikidata is considered very reliable, but is not always objective or complete. The skewed representation of facts in Wikidata inspired researchers Mirko Schäfer, Mel Chekol, Wessel Radstok and Egor Dmtriev to think about ways to profile data in order to identify bias. In 2021 they started working on the issue of bias in knowledge graphs.

This resulted in the BOLD Profiler. This tool allows the user to inspect any knowledge graph and map the distribution of facts represented in the graph. This enables persons without knowledge of semantic web technologies (e.g. SPARQL) to engage with knowledge graphs and retrieve information about their data. The tool makes it easier for (data)journalists, developers, data scientists or knowledge professionals to verify whether the data source is suitable for its intended purpose an whether it contributes to improving open data. Think for example of designers of automatic answering services, fact checkers, Wikimedia product managers.

BOLD is the result of the ‘Grip op desinformatie’ Call of the SIDN Fund.

More information on BOLD:
Paper: paper-5.pdf (ceur-ws.org)
Sofware:
BOLD Profiler, http://176.34.138.230/
BOLD user manual, https://egordm.github.io/BOLD/user_manual/
BOLD documentation, https://github.com/EgorDm/BOLD
If you would like to access the BOLD Profiler, please get in contact with Mirko Schäfer, m.t.schaefer@uu.nl