摘要：The explosion of services like web services, APIs, Mashups, etc., makes how to find the right one you need a tough problem. Tags, as a kind of metadata have been widely used to annotate services. In this paper, we propose to use an ontology automatically built from tags to improve the performance of service searching. We use the famous Mashup directory, Programmable.com, to illustrate our approach: First, all metadata especially tags of mashups and APIs at Programmable.com are crawled and preprocessed by the suffix stripping algorithm to lower the noise. Second, a Mashup-Mashup Network (MMN) is constructed to represent the Mashups and their relationships. And the community detection technique in complex network theory is introduced to mine the community structures (the potential domains of these Mashups) in MMN. Based on the Mashup communities, we further group the tags into corresponding domains. Finally the tag ontology is built with reference to WordNet, and embedded in our own developed software service registry and repository (S2R2), to improve the performance of service searching. Comprehensive experiments are also conduct to testify the effectiveness of the proposed approach.
