Combining collaborative filtering and search engine into hybrid news recommendations
Toon De Pessemier, Sam Leroux, Kris Vanhecke, Luc Martens · Ghent University Academic Bibliography (Ghent University) · 2015
Recommender systems have proven their usefulness in many classical domains, such as movies, books, and music, in helping users to overcome the information overload problem.When properly configured, recommender systems can also act as a supporting tool for content selection and retrieval in more challenging fields, such as news content.The short life span of news items and the demand for up-todate recommendations require a specially tailored approach.This paper proposes a hybrid recommender system using a search engine as a content-based approach and combining this with collaborative filtering for diversifying the user profiles.Based on similar users, user profile vectors are extended with related terms interesting to read about.The recommender system is fed real-time streams of news content originating from different sources.The resulting recommendations are clustered into topics and presented through a web application.This paper demonstrates that the advantages of both search engine and collaborative filtering can be successfully combined into a recommender system for domains with transient items, such as news.