4 April 2014: Recommender systems can be seen as a practical alternative to traditional search. They can satisfy the users’ needs for relevant information without the overhead of having the users explicitly state a query.
The query is therefore derived from both the user preferences and the application context. Recommender systems have proved their business value in many contexts already, ranging from e-shopping sites (e.g. Amazon) to very different settings such as television.
One of the most favored approaches to recommending is Collaborative Filtering (CF). CF is a technique to filter or evaluate items through the opinions of other people. It makes use of peer user ratings in order to provide recommendations on the items that are unknown but may interest the target user. Collaborative Filtering-based recommender systems are typically at the core of many of today’s mainstream recommendation engines (e.g. Amazon, Netflix, etc.). Despite its commercial success, it suffers from a number of limitations such as the cold-start problem and privacy concerns. At Telefonica Research, we are working on pushing the state-of-the-art in recommender systems by developing novel algorithms that are able to overcome some of the shortcomings of today’s systems.
