SEMI-FND: Stacked ensemble based multimodal inferencing framework for faster fake news detection

Abstract

Fake News Detection (FND) is an essential field in natural language processing that aims to identify and check the truthfulness of significant claims in a news article to decide the news veracity. FND finds its uses in preventing social, political, and national damage caused due to the misrepresentation of facts that may harm a particular section of society. Further, with the explosive rise in fake news dissemination over social media, including images and text, it has become imperative to identify fake news faster and more accurately. This work investigates a novel multimodal stacked ensemble-based framework (SEMI-FND) for fake news detection. Focus is also kept on ensuring faster performance with fewer parameters. A deep unimodal analysis is done on the image modality to identify NasNet Mobile as the most appropriate model to improve multimodal performance further. For text, an ensemble of BERT and ELECTRA has been used. The approach was evaluated on Twitter MediaEval Dataset and Weibo Corpus. The suggested framework offered 85.80% and 86.83% accuracy on the Twitter and Weibo datasets. These reported metrics are superior when compared to similar recent works. This work also reports a reduction in training parameters compared to its potential counterparts. The developed framework offered an overall parameter reduction of at least 20% with a 2% and 60% reduction on the image and text modality, respectively. Therefore, based on the investigations presented, it is concluded that applying a stacked ensembling significantly improves FND over other approaches while also improving speed.