Hidden Factors and Hidden Topics strongUnderstandingstrong Rating.pdfVIP

Hidden Factors and Hidden Topics: Understanding Rating Dimensions with Review Text Julian McAuley Jure Leskovec Stanford University Stanford University jmcauley@ jure@ ABSTRACT to suggest new products they might enjoy [ 19], to summarize the In order to recommend products to users we must ultimately pre- important points in their reviews [ 13], or to identify other users dict how a user will respond to a new product. To do so we must who may share similar opinions [24]. Such tasks have been studied uncover the implicit tastes of each user as well as the properties of across a variety of domains, from hotel reviews [6] to beer [21]. each product. For example, in order to predict whether a user will To model how users evaluate products, we must understand the enjoy Harry Potter, it helps to identify that the book is about wiz- hidden dimensions, or facets, of their opinions. For example, to ards, as well as the user’s level of interest in wizardry. User feed- understand why two users agree when reviewing the movie Seven back is required to discover these latent product and user dimen- Samurai, yet disagree when reviewing Casablanca, it helps to know sions. Such feedback often comes in the form of a numeric rating that these films belong to different genres; these users may have accompanied by review text. However, traditional methods often similar preferences toward action movies