FUZZY LOGIC AND MACHINE LEARNING FOR SOCIAL NETWORK TRUST CLASSIFICATION

Abstract

Differentiating between authentic and false news has become an increasingly pressing issue in this age of information overload, especially in the realm of public social media. In order to generate fact-checking categorisation results that are intelligible, a strategy is developed that is built on the synergy of a Neurosymbolic AI system that is driven by Fuzzy Logic approaches. Knowledge Graph Embedding (KGE) methods are the basis of the proposed fact-checking strategy. By projecting textual input into a vector space, it is able to extract involved entities as triples. These triples then generate graphs that effectively emphasise contextual information. Using fuzzy set modelling, which seeks to enhance the presentation of the final findings, the classification results are interpreted. Fuzzy variables with language phrases reflecting news dissemination are designed using the Hits@N measure. Afterwards, assessment of categorisation performance that mimics human performance is offered by using fuzzy rule design. By conducting experimental evaluations on the benchmark dataset, we demonstrate that our technique effectively distinguishes between true and fraudulent news. The clear explanations provided by the fuzzy rule design further boost its usefulness.