Detecting Knowledge Inconsistencies Across Text, Tables, and Knowledge Graphs
Kontrast compares table answers against Wikidata evidence to surface and classify cross-modal conflicts.
The paper frames the problem as modality-level inconsistency detection across Wikipedia text, tables, and knowledge graphs. Its taxonomy covers granularity gaps, direct conflicts, temporal changes, and incomplete KG structure. Experiments on Table-QA datasets found these inconsistencies were common and useful, though affected by Text-to-SPARQL errors and noise. ArXiv · AI/CL/LG's note
The paper frames the problem as modality-level inconsistency detection across Wikipedia text, tables, and knowledge graphs. Its taxonomy covers granularity gaps, direct conflicts, temporal changes, and incomplete KG structure. Experiments on Table-QA datasets found these inconsistencies were common and useful, though affected by Text-to-SPARQL errors and noise. ArXiv · AI/CL/LG's note
score 4