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

•A deep learning-based stacked ensembling approach to FND is developed and successfully tested.•Visual and lexical features are used as modalities.•Application of BERT and ELECTRA for extracting lexical features, is successfully demonstrated.•Developed approach is successfully tested on two datasets and found to be superior to recent counterparts.•The proposed framework is estimated to be faster than recent models.