Tag Ontology Automatic Building for Semantic Searching of Services: a Case Study on Mashup Services

摘要

—The explosion of services like web services, APIs,Mashups, etc., makes how to find the right one you needa tough problem. Tags, as a kind of metadata have beenwidely used to annotate services. In this paper, we proposeto use an ontology automatically built from tags to improvethe performance of service searching. We use the famousMashup directory, Programmable.com, to illustrate ourapproach: First, all metadata especially tags of mashupsand APIs at Programmable.com are crawled and preprocessedby the suffix stripping algorithm to lower the noise.Second, a Mashup-Mashup Network (MMN) is constructedto represent the Mashups and their relationships. And thecommunity detection technique in complex network theory isintroduced to mine the community structures (the potentialdomains of these Mashups) in MMN. Based on the Mashupcommunities, we further group the tags into correspondingdomains. Finally the tag ontology is built with referenceto WordNet, and embedded in our own developed softwareservice registry and repository (S2R2), to improve the performanceof service searching. Comprehensive experimentsare also conduct to testify the effectiveness of the proposedapproach.