Context-Aware Review Helpfulness Rating Prediction Jiliang Tang, Huiji Gao, Xia Hu and Huan Liu Computer Science and Engineering, Arizona State University, Tempe, AZ, USA {jiliang.tang, huiji.gao, xia.hu, huan.liu}@asu.edu ABSTRACT Online reviews play a vital role in the decision-making pro- cess for online users. Helpful reviews are usually buried in a large number of unhelpful reviews, and with the consis- tently increasing number of reviews, it becomes more and more difficult for online users to find helpful reviews. There- fore most online review websites allow online users to rate the helpfulness of a review and a global helpfulness score is computed for the review based on its available ratings. However, in reality, user-specified helpfulness ratings for re- views are very sparse - a few reviews attract large numbers of helpfulness ratings while most reviews obtain few or even no helpfulness ratings. The available helpfulness ratings are too sparse for online users to assess the helpfulness of reviews. Also the helpfulness of a review is not necessarily equally useful for all users and users with different background may treat the helpfulness of a review very differently. The user idiosyncracy of review helpfulness motivates us to study the problem of review helpfulness rating prediction in this paper. We first identify various types of context information, model them mathematically, and propose a context-aware review helpfulness rating prediction framework CAP. Experimental results demonstrate the effectiveness of the proposed frame- work and the importance of context awareness in solving the review helpfulness rating prediction problem. Categories and Subject Descriptors H.3.3 [Information Search and Retrieval]: Information filtering; J.4 [Computer Application]: Social and Behav- ioral Sciences Keywords Review Rating Prediction, Social Context, Review Recom- mendation, Content Context 1. INTRODUCTION Reviews, providing experiences with and opinions about products or services from other users, play a crucial role in Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full cita- tion on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or re- publish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from permissions@acm.org. RecSys’13, October 12–16, 2013, Hong Kong, China. Copyright 2013 ACM 978-1-4503-2409-0/13/10 ...$15.00. http://–enter the whole DOI string from rightsreview form confirmation. online communities such as e-commerce and product review sites, where users rely on reviews in their decision-making process. For example, users will select restaurants with good reviews in Yelp, and reviews about products in eBay are im- portant sources of information for users to make purchases. However, helpful reviews are usually buried in large numbers of useless reviews [15], and with the availability of massive reviews, it becomes increasingly difficult for online users to find helpful reviews. In an attempt to help online users identify helpful reviews, most online review websites implement a mechanism to allow users to rate the helpfulness of a review and then a global helpfulness score is computed for the review such as “20 out of 30 people found the following review helpful” in eBay and a score from 0 to 5 in Ciao. In reality, a large proportion of reviews obtain few or no helpfulness ratings, particularly the more recent ones and the available helpfulness ratings are too sparse for online users to assess the helpfulness of reviews [15]. For example, it is difficult for users to assess the helpfulness of a review in eBay with a score of “1 out of 1 people found review helpful”. There is recent work automat- ically predicting a global helpfulness score for a review [9, 13, 15]. However, a review is not necessarily equally useful for all users. For example, in eBay, a review’s helpfulness can have a score of “500 out of 1000 people found the fol- lowing review helpful”, which indicates that the other half do not think the review helpful or are indifferent. This user idiosyncracy of review helpfulness motivates us to study if we can predict review helpfulness rating for each user. We choose a product review site, a classical type of online review websites, to investigate if review helpfulness rating prediction can help mitigate the problem caused by user id- iosyncracy. Figure 1(a) gives an overview of product review sites where users have four different behaviors - connecting to other users, writing reviews, rating the helpfulness of re- views, and rating items. Figure 1(b) depicts the user help- fulness rating behavior and there are two types of ratings including item ratings and review helpfulness ratings. The review helpfulness rating is fundamentally different from the item rating. The former indicates “how does a user X rate a review from another user Y ?” while the latter denotes “how does a user X rate an item?”. These differences not only are useful to differ our studied problem from item rating predic- tion problem, but also present unique opportunities for us to investigate the review helpfulness rating prediction problem. First, the texts of reviews can affect how users rate the help- fulness of reviews [13], thus provide content context about reviews. Second, users play two roles in review helpfulness rating: authors - users who write reviews, and raters - users who rate reviews. Both raters and authors can rate items,