Content-based recommender systems

Abstract

The goal of the thesis is to evaluate content-based recommender systems in the domain of video games. The thesis compares approaches based on Linked Open Data and natural language parsing(NLP) with traditional approaches which are only based on NLP methods. The purpose of a recommender system for an user is to present the most relevant products from a larger set of products. In the domain of this project the set of products are Greenlight submissions and the relevant submissions are those which a specific Steam user would want to discover, rate and eventually buy. Recommender systems can very roughly be categorized into two distinct approaches (i ) content-based and (ii ) collaborative. A third approach (iii ) hybrid combines the other two approaches.