Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Ontology Enhanced Claim Detection

Zehra Melce Hüsünbeyi, Tatjana Scheffler

TL;DR

The paper tackles automatic sentence-level claim detection under data scarcity by fusing ontology-based representations with text embeddings. It presents a multimodal architecture that combines BERT with OWL ontology embeddings (via OWL2Vec) derived from the ClaimsKG data, and systematically compares statistical and neural approaches on ClaimBuster and NewsClaims. Key findings show ontology-enhanced features boost performance on small datasets and help mitigate BERT-induced bias, while large datasets reduce the relative advantage of domain knowledge. The work highlights the practical value of integrating structured knowledge with neural models for robust, transparent fact-checking in low-resource settings and across multilingual contexts.

Abstract

We propose an ontology enhanced model for sentence based claim detection. We fused ontology embeddings from a knowledge base with BERT sentence embeddings to perform claim detection for the ClaimBuster and the NewsClaims datasets. Our ontology enhanced approach showed the best results with these small-sized unbalanced datasets, compared to other statistical and neural machine learning models. The experiments demonstrate that adding domain specific features (either trained word embeddings or knowledge graph metadata) can improve traditional ML methods. In addition, adding domain knowledge in the form of ontology embeddings helps avoid the bias encountered in neural network based models, for example the pure BERT model bias towards larger classes in our small corpus.

Ontology Enhanced Claim Detection

TL;DR

The paper tackles automatic sentence-level claim detection under data scarcity by fusing ontology-based representations with text embeddings. It presents a multimodal architecture that combines BERT with OWL ontology embeddings (via OWL2Vec) derived from the ClaimsKG data, and systematically compares statistical and neural approaches on ClaimBuster and NewsClaims. Key findings show ontology-enhanced features boost performance on small datasets and help mitigate BERT-induced bias, while large datasets reduce the relative advantage of domain knowledge. The work highlights the practical value of integrating structured knowledge with neural models for robust, transparent fact-checking in low-resource settings and across multilingual contexts.

Abstract

We propose an ontology enhanced model for sentence based claim detection. We fused ontology embeddings from a knowledge base with BERT sentence embeddings to perform claim detection for the ClaimBuster and the NewsClaims datasets. Our ontology enhanced approach showed the best results with these small-sized unbalanced datasets, compared to other statistical and neural machine learning models. The experiments demonstrate that adding domain specific features (either trained word embeddings or knowledge graph metadata) can improve traditional ML methods. In addition, adding domain knowledge in the form of ontology embeddings helps avoid the bias encountered in neural network based models, for example the pure BERT model bias towards larger classes in our small corpus.
Paper Structure (22 sections, 1 figure, 6 tables)

This paper contains 22 sections, 1 figure, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Data and task
  4. ClaimBuster
  5. NewsClaims
  6. Methodology
  7. Vectorization in Statistical Machine Learning Methods
  8. LIWC and other linguistic features
  9. Word2Vec for Feature Extraction
  10. Knowledge Graphs for Encoding of Metadata
  11. Neural Network Based Approaches
  12. Word2Vec Embedding
  13. OWL2Vec* Ontology Embedding
  14. BERT model
  15. Multimodal Fusion of BERT Model and Ontology Embeddings
  16. ...and 7 more sections

Figures (1)

  • Figure 1: Proposed model fusing textual and ontology features