Context-aware similarity search of web service

Growing number of services need efficiently locating the desired web services. The similarity metric plays an important role in service search. However, most solutions only exploit information available in the descriptions of services and employ a single similarity metric. We observe that under different contexts the descriptions of services vary in their level of semantic ambiguity. It is hardly to apply one metric upon all services to filtering out those required by users. For services with more ambiguity, the metric with more restriction should be applied. As to those services sharing the same context and thus less ambiguity, the metric with less restriction is more appropriate. In this paper we show information exploitation of context of services to improve the searching accuracy. We firstly employ ideas from the semantic link network to infer the ambiguity and relatedness of the context of services, then utilize two kinds of similarity metrics on services with different contexts. The conservative metric is on those services with more ambiguity, while the relaxed metric is for those with less ambiguity. The experiments show that our solution outperforms some searching methods.