Analysis of Consistency between Wikidata and Wikipedia Categories

Abstract

Wikipedia categories play a significant role in organizing articles by topic. They form a hierarchy, which groups related articles into larger collections. Wikidata provides a corresponding item for each category and allows to define membership of other items to the specific category by a SPARQL query or by specifying classes and properties. This provides us with multiple, redundant sources of category membership which may deviate quite substantially. In this paper, we investigate inconsistencies between Wikipedia and Wikidata category members and analyze possible reasons. We propose a candidate category generation and evaluation workflow that traverses the category hierarchy of Wikipedia in all available languages and compares the results with information obtained from Wikidata. This workflow can be executed either online using the publicly available endpoints or offline based on the provided dumps. Furthermore, we formulate concrete suggestions to harmonize category membership definitions between Wikipedia and Wikidata.