Deep Learning Based False Review Identification And Research

With the rapid development of mobile network payment and the e-commerce industry,people’s consumption channels have become more and more diversified,and online shopping has become the choice of most people.The rise of online shopping has been accompanied by a large number of online reviews,which influence merchants’ merchandise sales.The openness of the Internet is a double-edged sword that can both inform users of the pros and cons of a product and mislead them about it.There are some merchants or watermen who take advantage of this drawback to manipulate online reviews,so the number of fake reviews is gradually increasing.Nowadays,online reviews play an increasing role on e-commerce platforms,and determining the authenticity of online reviews is of great significance to both users and merchants.Therefore,it becomes especially critical to effectively identify the falsity of online reviews.The large volume of fake comment data and high concealment lead to some difficulties in correctly identifying fake comments,machine learning has been applied in fake comment identification,but the accuracy rate of model identification is not high.With the promotion of deep learning,it can be increasingly applied in the field of false comment identification,and this paper has carried out research work mainly from the following aspects.First,the obtained false comment dataset is subjected to relevant pre-processing operations to achieve the purpose of reducing the resource occupation rate and improving the efficiency of the experiment.At the beginning of the article,relevant algorithms of machine learning are used to identify the false comments,and the false comment recognition model is constructed by traditional machine learning to construct comparison experiments for evaluation and comparison.However,because traditional machine learning is used to extract relevant text features from comments for comment recognition,it does not consider the semantic information contained in the comment text,so the extracted text information is incomplete.To solve this problem a hybrid deep learning model with CNN+BiLSTM in parallel is introduced,and in order to enhance the weight of keywords and introduce attention after the CNN and BiLSTM layers respectively mechanism to differentiate the importance of keywords.However,the Glove and Word2vec models used in the hybrid deep learning model are based on static word vector feature extraction,and this model cannot accurately acquire the context contained in human language,and the acquisition of contextual semantics is also limited.Based on this problem,a false comment recognition model based on ERNIE and hybrid deep learning is proposed to optimize the pre-trained word vectors in the hybrid deep learning model.Finally,it is confirmed that the false comment recognition model based on ERNIE and hybrid deep learning has better experimental results through relevant comparison experiments.