An Attribute Enhanced Domain Adaptive Model for Cold-Start Spam Review Detection

Proceedings of the 27th International Conference on Computational Linguistics, pages 1884–1895 Santa Fe, New Mexico, USA, August 20-26, 2018. 1884 An Attribute Enhanced Domain Adaptive Model for Cold-Start Spam Review Detection Zhenni You 1, Tieyun Qian 1∗, Bing Liu 2 1 School of Computer Science, Wuhan University, Wuhan, China 2 Department of Computer Science, University of Illinois at Chicago, IL, USA ∗ Contact author {znyou,qty}@whu.edu.cn, liub@uic.edu Abstract Spam detection has long been a research topic in both academic and industry due to its wide applications. Previous studies are mainly focused on extracting linguistic or behavior features to distinguish the spam and legitimate reviews. Such features are either ineffective or take long time to collect and thus are hard to be applied to cold-start spam review detection tasks. Recent advance leveraged the neural network to encode the textual and behavior features for the cold- start problem. However, the abundant attribute information are largely neglected by the existing framework. In this paper, we propose a novel deep learning architecture for incorporating entities and their inherent attributes from various domains into a unified framework. Specifically, our model not only encodes the entities of reviewer, item, and review, but also their attributes such as location, date, price ranges. Furthermore, we present a domain classifier to adapt the knowledge from one domain to the other. With the abundant attributes in existing entities and knowledge in other do- mains, we successfully solve the problem of data scarcity in the cold-start settings. Experimental results on two Yelp datasets prove that our proposed framework significantly outperforms the state-of-the-art methods. 1 Introduction Online reviews and ratings are playing more and more critical roles in E-Commerce places. The Mintel flagship report showed that 69 percent of Americans seek out others’ advice online before making a purchase 1. This gives strong incentives for imposters to game the system. As a result, fraudulent reviews flood the e-market websites. Forbes news 2 reported that Amazon’s fake review problem is now worse than ever in 2017. Fake reviews are perceived as threats to the ecosystem of the e-business sites, companies, and users, and thus it becomes an urgent task to detect fake or spam reviews. Existing approaches in spam review detection mainly focused on exacting linguistic features and be- havioral features. However, linguistic features are ineffective when they are used to detect the real-life fake reviews (Mukherjee et al., 2013b; Wang et al., 2017b), and it usually requires a large number of samples to make the observations on behavior features. When dealing with the cold-start problem, i.e., a review is just posted by a new reviewer, it is hard to construct effective behavioral features for the new reviewer. More recently, Wang et al. (2017b) proposed a neural network model to jointly encode the textual and behavioral information into the review embeddings. This was a good try. However, due to the lack of sufficient information, the results reported in (Wang et al., 2017b) are not promising enough (with an accuracy less than 65%). This work …

Content truncated to 500 words