Learning parliamentary profiles for recommendation tasks

Abstract

We consider the problem of building a content-based recommender system in a parliamentary context, which may be used for two different but related tasks. First, we consider a filtering task where, given a new document to be recommended, the system can decide those Members of the Parliament who should receive it. Second, we also consider a recommendation task where, given a request from a citizen, the system should present information on those deputies that are more involved in the topics of the request. To build the system we collected, for each Member of the Parliament, the text of corresponding speeches within the parliament debates and generated, with different techniques, a profile that was used to match against the input (document or request). We tested our methods using the documents of the regional Andalusian Parliament at Spain, obtaining promising results.