The accuracy of content-based recommender systems tends to depend on the way similarity is defined. In this paper, we will explore different ways to measure similarity for a news recommender system based on news headlines. We will compare human judgements of similarity with Lin’s taxonomy-based measure and the WASP measure that uses annotated corpus data. The main aim of this work is to better understand similarity, so that it can be used to explain recommendations to users.
|Title of host publication||Workshop on Recommender Systems and Intelligent User Interfaces|
|Subtitle of host publication||2nd international workshop on web personalisation,recommender iystems and intelligent iser interfaces|
|Publication status||Published - 2006|
- recommender systems