PerspectiveSpace

Abstract

Words mean different things to different people, and capturing these differences is often a subtle art. These differences are often "a matter of perspective," and perspective can be taken to be the set of beliefs held by a person as a result of their background, culture, tastes, and experience. Understanding perspective is often pivotal in resolving disputes, and such an understanding is useful for estimating the opinions a person would have on a matter, making recommendations to the person, and even understanding the language the person uses. Traditionally, perspectives are studied through the use of questionnaires and surveys that rarely leave people feeling like their opinions have been properly represented. In this paper, I propose a system for discovering distinct communities of people with coherent belief patterns, while providing a means to characterize those patterns. This system utilizes data on how people agree or disagree on assertions that they themselves have expressed. This system, called PerspectiveSpace, is an approach whereby elementary linear operations are used to perform calculations on user models and microtheories. PerspectiveSpace has applications ranging from discovering subcultures in a larger society to building community-driven web sites that adapt to individual perspectives, and three such applications are illustrated here. The first is the detection and amelioration of abusive user activity on a web site with community-generated content. The second is SlantExplorer, which is a tool that highlights and analyzes the perspectives underlying a document. The third is 2-wit, a novel movie recommender based on the perspectives people have about movies rather than simply the ratings they give them.