An unsupervised topic-sentiment joint probabilistic model for detecting deceptive reviews

作者：

Highlights：

• First considering topic, then sentiment, which is consistent to human written habits.

• Improve typical LDA model with topic and sentiment.

• Superior to baseline models in dealing with balanced dataset.

• Have superiority in detecting big imbalanced real-life deceptive reviews.

摘要

•First considering topic, then sentiment, which is consistent to human written habits.•Improve typical LDA model with topic and sentiment.•Superior to baseline models in dealing with balanced dataset.•Have superiority in detecting big imbalanced real-life deceptive reviews.

论文关键词：Deceptive review detection,Topic-sentiment joint probabilistic model,Latent dirichlet allocation,Gibbs sampling

论文评审过程：Received 21 October 2017, Revised 12 June 2018, Accepted 2 July 2018, Available online 21 July 2018, Version of Record 1 August 2018.

论文官网地址：https://doi.org/10.1016/j.eswa.2018.07.005