Poster:Research on multi-feature fusion false review detection based on DistilBERT-BiLSTM-CNN

In the context of rapid expansion of e-commerce and social media, fake reviews have become a prominent issue, misleading consumers and damaging the reputation of businesses. To more effectively identify and filter these false reviews, this paper introduces a multi-feature fusion strategy and proposes a composite architecture based on DistilBERT-BiLSTM-CNN. Experimental results on the Amazon dataset show that compared to the fake review detection methods of LSTM, HAN, BERT, RoBERTa and ELECTRA models, our model’s method improved by approximately 4.2%, 3%, 2.6%, 1.3% and 0.8% respectively, achieving a maximum accuracy of 91.5%.