《Handling Cold-Start Problem in Review Spam Detection by Jointly Embedding Texts and Behaviors》阅读笔记

《Handling Cold-Start Problem in Review Spam Detection by Jointly Embedding Texts and Behaviors》阅读笔记

摘录

1. Compared with the recurrent neural network (RNN), the CNN can do a better job of modeling the different aspects of a review.

疑问

1. what is global behavioral information the traditional discrete features can not effectively record the global behavioral information
   阅读 (Wang et al., 2016).Learning to Represent Review
   with Tensor Decomposition for Spam Detection

2. Ren and Zhang (2016) have proved that the CNN can capture complex global semantic information and detect review spam more effectively, compared with traditional discrete manual features and the RNN model.
   阅读Deceptive Opinion Spam Detection Using Neural Network

3. CNNmodel filter层是否需要activation function

4. Rating embedding是如何使用的