Title of session: zelph: Enhancing Wikidata Quality Through Semantic Network Analysis and Contradiction Detection
Session description:
This presentation introduces zelph, an innovative open-source semantic network system specifically designed to improve Wikidata’s data quality. zelph uniquely represents knowledge in a way that allows inference rules to be defined within the network itself, enabling powerful automated reasoning about Wikidata’s vast knowledge graph.
I will demonstrate how zelph:
- Successfully processes the complete Wikidata dataset (1.7 TB) while maintaining memory efficiency
- Detects logical contradictions within the knowledge graph that would otherwise remain hidden
- Makes automatic inferences based on existing relationships, uncovering implicit knowledge
- Represents relations and rules as first-class nodes, enabling meta-reasoning about relationships themselves
- Treats both facts and rules uniformly, whether they come from custom zelph scripts or Wikidata’s knowledge structure
The presentation will include live demonstrations of contradiction detection and inference, showing real examples from Wikidata. I’ll present a roadmap for how zelph could be integrated as a practical tool for Wikidata maintainers, and discuss collaboration opportunities to develop inference rules that reflect community consensus about ontological relationships.
As an open-source project, zelph aims to become a valuable addition to the Wikidata ecosystem, helping to ensure knowledge integrity while reducing maintenance burden.
Username for contact: Acrion-dev
Session duration: 35min
Session type: presentation
Language of session: English
Prerequisites:
- Basic understanding of knowledge graphs and semantic relationships
- Familiarity with Wikidata’s structure (properties and items) is helpful but not required
- No programming knowledge necessary to understand the concepts presented
Any other details to share?
The technical proof-of-concept has already been completed, demonstrating zelph’s ability to handle the scale and complexity of Wikidata. The project is available at https://github.com/acrion/zelph and https://zelph.org, where interested participants can explore the code and documentation.
Interested? Add your username below: