基于深度学习的虚假评论识别

Spam Review Detection Based on Deep Learning Framework

虚假评论识别在评论网站发展迅速的今天越来越重要。有效的区分虚假评论和真实评论是当今评论网站的急切需求。过去对于虚假评论识别主要是通过手工提取特征,用传统机器学习的算法来进行识别。但是通过依靠人对该领域的认识来提取特征会有一些信息上的流失。基于这样的因素,提出了深度学习的框架来对虚假评论进行识别。同时使用评论的内容和评论的其他信息来对评论进行真伪的鉴别。模型相对于传统的方法识别率有很大的提升。

Nowadays, spam review detection is more important in review websites. It is required to classify spam review and non - spam review validly. At present, the works of spam review detection extract features from human intuition and some domain knowledge and then utilize traditional machine learning method to detect spam review. However, use those method may cause the loss of information. Based on that, we inform a deep learning framework for spam re - view detection. We take advantages of both review context and be- havior information. Our method outperforms other strong baseline methods.