EPEN – A Personalized Recommendation System for News Articles

Starting at the beginning of 2013, we intiated a new project in the area of recommender systems. The project is called EPENDevelopment of a personalized recommendation system for news articles, (German: Entwicklung von Personalisierten Empfehlungssystemen für Nachrichten).

The project is funded through the ZIM program the BMWi. CC IRML is undertaking this project together with Plista  and will research the following topics:

  • Create novel cross-topic (cross-domain) recommendation algorithms: We want to find the best strategies to deduce the interest of a user for one news domain, e.g. sports, based on the interest of the interests in another domain, e.g. culture.
  • Find the optimal ways to combine weights from different recommendation algorithms: To offer good recommendations, different algorithms have to be combined to take into account the interests of a user, the context, the location, etc. We want to find intelligent strategies to combine different algorithms to present recommendations that really meet the users’ interests and needs.
  • New evaluation metrics to better compare real-life (online) performance of recommendation approaches against old-fashioned offline evaluations (RMSE, Precision and Recall etc.).

We will publish results on this blog on a regular basis, so stay tuned. In the meantime check out the Plista Contest where you already can test your algorithm in a real-time setting. A Java-based implementation with a Most Popular recommender is available here, to help you get started.



Leave a Reply

You can use these HTML tags

<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>