Personalized Recommendation of User Comments via Factor Models

Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, pages 571–582, Edinburgh, Scotland, UK, July 27–31, 2011. c⃝2011 Association for Computational Linguistics Personalized Recommendation of User Comments via Factor Models Deepak Agarwal Bee-Chung Chen Bo Pang Yahoo! Research 701 First Ave Sunnyvale, CA 94089 {dagarwal,beechun,bopang}@yahoo-inc.com Abstract In recent years, the amount of user-generated opinionated texts (e.g., reviews, user com- ments) continues to grow at a rapid speed: fea- tured news stories on a major event easily at- tract thousands of user comments on a popular online News service. How to consume subjec- tive information of this volume becomes an in- teresting and important research question. In contrast to previous work on review analysis that tried to filter or summarize information for a generic average user, we explore a dif- ferent direction of enabling personalized rec- ommendation of such information. For each user, our task is to rank the comments associated with a given article according to personalized user preference (i.e., whether the user is likely to like or dislike the comment). To this end, we propose a factor model that incorporates rater-comment and rater-author interactions simultaneously in a principled way. Our full model significantly outperforms strong baselines as well as related models that have been considered in previous work. 1 Introduction Recent years have seen rapid growth in user- generated opinions online. Many of them are user reviews: a best-seller or a popular restaurant can get over 1000 reviews on top review sites like Ama- zon or Yelp. A large quantity of them also come in the form of user comments on blogs or news arti- cles. Most notably, during the short period of time for which a major event is active, news stories on one single event can easily attract over ten thousand comments on a popular online news site like Yahoo! News. One question becomes immediate: how can we help people consume such gigantic amount of opinionated information? One possibility is to take the summarization route. Briefly speaking (see Section 2 for a more detailed discussion), previous work has largely formulated review summarization as automatically or manually identify ratable aspects, and present overall senti- ment polarity for each aspect (Hu and Liu, 2004; Popescu and Etzioni, 2005; Snyder and Barzilay, 2007; Titov and McDonald, 2008). A related line of research looked into predicting helpfulness of re- views in the hope of promoting those with better quality, where helpfulness is usually defined as some function over the percentage of users who found the review to be helpful (Kim et al., 2006; Liu et al., 2007; Danescu-Niculescu-Mizil et al., 2009). In short, the focus of previous work has been on dis- tilling subjective information for an average user. Whether opinion consumers are looking for qual- ity information or just wondering what other people think, each may have different purposes or prefer- ences that is not well represented by a generic av- erage user. If we think about how we deal with in- formation content overflow on the …

Content truncated to 500 words