← Go back

Profile page

Umair Qudus

Early stage researcher

Work Description

Title: Establishing the veracity of Knowledge Graphs

Current approaches for fact validation can be categorized into two families: (i) approaches that use unstructured textual sources and (ii) approaches that use structured knowledge graphs as reference knowledge. Few of these latter approaches are: (a) meta-path-based approaches, (b) predicate-path-based, and (c) corroborative-path-based approaches. All these approaches can be used to check a single fact. However, when it comes to calculating the veracity of a complete knowledge graph, none of these approaches are applicable due to their high runtime.

The following major accomplishments are expected in this research project:

  1. Systematic survey of the field of fact checking over knowledge graphs. This will be designed as a living document and updated during the project.
  2. A scalable hybrid approach utilizing both document corpora and knowledge graphs for the sake of fact checking. We will aim to outperform the state of the art w.r.t. both runtime and F-measure.
  3. A novel approach to evaluate the veracity of large-scale knowledge graphs efficiently. Such an approach is currently not available.
  4. Provide evidence and semi-automatic means to fix erroneous content.