Combining Collaborative Filtering and Semantic-Based Techniques to Recommend Components for Mashup Design

Mashup editors enable end-users to mix the functionalities of several applications to derive a new one. However, when the end-user faces the development of a new mashup application s/he has to cope with the abundance of services and information sources available on the Web, and with complex operations like filtering and joining. Thus, even a simple to use mashup editor is not capable of providing adequate support, unless it embeds intelligent methods to process the semantics of available mashups and rank them based on how much they meet user needs. Most existing mashup editors process either semantic or statistical information to derive recommendations for the mashups considered suitable to user needs. However, none of them uses both strategies in a synergistic way. In this paper we present a new mashup advisory approach and a system that combines the statistical and semantic based approaches, by using collaborative filtering techniques and semantic tagging, in order to rank mashups based on user goals. We have proven the validity of the proposed approach through experimental sessions based on data from the ProgrammableWeb repository.