Mapping the Media Universe Using User-Generated Data in Online Social Networks

Abstract

The way people watch movies and TV has been going through great changes in recent years, one being that people are increasingly willing to share their TV watching habits with friends (and strangers) through online social network platforms. In this work, we collect usage data from the online social network tvtag and propose a data structure to represent and efficiently retrieve similarity information about movies and TV shows. We refer to this structure as "map of media", because it consists of a multi-dimensional Euclidean space, where each item is represented by a set of coordinates and the distance between them represents the similarity. We propose several metrics to evaluate the resulting structure and show that, besides computational efficiency, the proposed approach provides a high quality measure for item similarity. Moreover, the quality increases with increasing dimensionality of the map.