ARSYS -- Article Recommender System

In a world flooded by information, a filtering mechanism is compulsory. Recommender systems have taken a huge leap towards this goal, significantly improving the user experience in the online environment. There are two main approaches, content-based and collaborative filtering, both with advantages and drawbacks. We propose an article recommender system that integrates content based, collaborative and metadata recommendations, allowing users to select the method that best suits their needs. The first approach uses keywords in order to find similar articles, given a query or an entire document. Collaborative filtering is implemented using a P2P network in which data is distributed evenly across all peers. The last technique uses data from a semantic repository containing information about articles (e.g. title, author, domain), which can be interrogated using natural language-like queries. In addition, we present in detail the results obtained from employing the P2P network in terms of providing timely responses to the collaborative filtering technique and of ensuring reliability through data replication.