Detecting fence-sittings among categories of web services

Finding clusters of Web services benefits their sorting and discovery, moreover, certain services sitting on the fence between different services clusters could improve the quality of sorting and help users with cross-domain background to retrieve or combine the services correctly. In this paper, we extract words from the description field of Web services on the level of semantic, and connect these words together to form a Weight Words Network (WWN), in which the edge weight denotes correlation of words in description field. On the basic of complex network community detection, we discover the fence-sitting words in WWN firstly. And then we map these words to the corresponding services, getting the fence-sitting services among different services categories. We compare the result of experiment on a well-known Web services test collection, OWLS-TC, with a handmade classification, and find out the corresponding real-life meaning of fence-sitting services to evaluate the reasonability of our method.