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

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