Search results

Fake news refers to intentionally fabricated or misleading information designed to deceive the public and manipulate opinions for personal, political or financial gain. Most existing detection methods primarily focus on capturing language features from news content. However, these methods neglect the varying importance of different news entities. Additionally, these methods tend to overlook the auxiliary role of external knowledge, resulting in an incomplete understanding of the entity. To address these issues, this paper aims to propose a Dual-layer Semantic Information Extraction Network with External Knowledge (DSEN-EK) for fake news detection.

This approach is proposed to comprise three parts: Dual-layer Semantic Information Extraction Network, Entity Integration Network with External Knowledge and Classifier. Specifically, Dual-layer Semantic Information Extraction Network is designed to enhance relationships between entities and the influence of important entity representations. The Entity Integration Network with External Knowledge is proposed to extract entity descriptions from external knowledge bases.

The DSEN-EK model performs well on the Liar, Constraint, Twitter15 and Twitter16 data sets, achieving accuracy of 98.02%, 94.61%, 90.09% and 93.65%, respectively. These results highlight its effectiveness in detecting fake news across different types of content.

The Dual-layer Semantic Information Extraction Network is proposed to capture the relationships between entities and enhance the continuous semantic information of the news. The Entity Integration Network with External Knowledge is designed to enrich entity descriptions, leading to a more comprehensive capture of semantic details.