BM5-SP-SC: A Dual Model Architecture for Contradiction Detection on Crowdfunding Projects

Despite the prevalence of scams in crowdfunding projects, currently, there is little research into identification fraudulent or infeasible projects. Since detecting projects challenging, most existing on fake information has focused news charity based social media, but very limited. Therefore, to solve this problem, we focus how detect knowledge extraction and contradiction detection. We proposed a novel method called BM5-SP-SC (BERT-MT5-Sentence Pattern-Sentiment Classification). BM5 (BERT-MT5), which built from combination key-BERT fine-tuned MT5 transformers, was used extract feature for training construct an adaptive model. The correct rate keywords extracted by our model up 72.7%, recall 100%, F-measure 84.19%. minimum train loss 0.1342, evaluation achieved 0.3064. BLEU score summary-to-keyword 37.336. Moreover, SP (Sentence Pattern) matching achieve extraction. Furthermore, SC (Sentiment Classification) build sentiment classifier thesaurus identifying Our overall accuracy 85.26%