A Novel Knowledge Acquisition Framework for Claim Factuality Disambiguation
| dc.contributor.author | Aruna Shankar | |
| dc.date.accessioned | 2026-07-14T06:49:54Z | |
| dc.date.issued | 2025 | |
| dc.description | The speedy spread of fake news in the digital age has presented significant challenges to accurate and timely truth verification. Traditional fact-checking methods, while valued, are insufficient to address the complexity of online content. Although fact-checking approach offer scalability, they still face critical limitations, including the lack of evidence-supported semantic frameworks for mapping factual coherence between claims and evidence, as well as the absence of mechanisms to capture and represent claim-evidence discrepancies effectively. This research addresses the gap by proposing a novel knowledge acquisition framework that acquires evidence-based knowledge to model the factual coherence relationship between claim and evidence. By systematically mapping claims to retrieved evidence, the framework effectively disambiguates conflicting or ambiguous information, ensuring greater clarity and accuracy in the fact checking process. The significance of the knowledge-level disambiguation approach goes beyond the traditional fact checking methods, as it incorporates a contextual attention mechanism and semantic analysis to detect, represent, and resolve discrepancies. This capability is especially critical for handling complex claims that requires a multi-faceted evidence assessment framework. By ensuring a precise alignment between claim and evidence, the framework effectively resolves semantic uncertainties, thereby enabling a more accurate and context-aware fact-checking. The proposed knowledge level disambiguation mechanism was evaluated using various versions of BERT-based models such as BERT, RoBERTa, and BioBERT on diverse datasets such as FEVER, LIAR-PLUS, HoVer, and PubMed. The outcomes of this research highlight the effectiveness of the proposed hybrid knowledge-level disambiguation approach, which integrates BERT-based models with human-in-the-loop decision-making to ensure accurate claim interpretation, resolve ambiguities, and align claims with relevant evidence. As a result, the approach establishes a robust capability to address the iv evidence discrepancies. The findings further underscore the framework’s effectiveness in enhancing fact-checking approach by mitigating ambiguities and maintaining precise claimevidence alignment. This robust performance highlights the framework’s potential as a comprehensive solution for combating fake news, even in the presence of conflicting or incomplete information. Building on the current framework, future research will extend the hybrid approach to fact- check multimodal datasets, encompassing text, images, and videos. By integrating multiple data modalities, the approach supports more comprehensive and accurate verification in real-world scenarios, thereby further enhancing its adaptability and effectiveness in mitigating misinformation. | |
| dc.identifier.uri | https://scholarhub.unimas.my/handle/123456789/1175 | |
| dc.language.iso | English | |
| dc.publisher | UNIVERSITI MALAYSIA SARAWAK | |
| dc.relation.ispartofseries | Faculty of Computer Science and Information Technology | |
| dc.subject | Automated fact-checking, misinformation, semantic framework, evidence mapping, discrepancy detection, retrieval, context-based disambiguation | |
| dc.title | A Novel Knowledge Acquisition Framework for Claim Factuality Disambiguation | |
| dc.type | PhD |