A deceptive reviews detection model: Separated training of multi-feature learning and classification

作者：

Highlights：

• Separated training of multi-feature learning and classification is proposed.

• Extract different semantic features (local, temporal, weighted) independently.

• Make full use of different features to learn better feature representation.

• Analyze the results from the perspective of part-of-speech.

摘要

•Separated training of multi-feature learning and classification is proposed.•Extract different semantic features (local, temporal, weighted) independently.•Make full use of different features to learn better feature representation.•Analyze the results from the perspective of part-of-speech.

论文关键词：Deceptive reviews detection,Separated training,Convolutional neural network,Recurrent neural network,Self attention

论文评审过程：Received 4 February 2021, Revised 9 July 2021, Accepted 23 September 2021, Available online 28 September 2021, Version of Record 1 October 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115977