Web service discovery system using the vector space model and the normalized Google distance

Abstract

The Web is evolving from a collection of static pages to a collection of online services which interoperate in order to provide solutions to the requests of a particular user. In order to achieve this interoperation of services one must first discover web services which satisfy the functional requirements of the requester of a service. Web Services follow a number of standards such as WSDL which defines how the services should be described in a machine processable way, SOAP and REST which define the method of communication between the requester of the service and the provider of the service and also UDDI which is the standard used to facilitate discovery of such services. However public UDDI registries which facilitated Web Service Discovery have recently shut-down. In this project we propose a system which can be used as a registry for web services, where the provider of the service registers the developed service with the system, as well as an index of services using a focused web crawler for WSDL files. The system abstracts the description of a service from the WSDL document by storing the service description in a database following a particular specified schema. Thus the system can be used to register even services which are not WSDL described. Two algorithms were implemented which enable potential users to search for services registered within the system. The first algorithm uses the Vector Space Model, which represents every service as a vector and the cosine similarity between the query and each service vector is used for ranking. The second algorithm enhances the Vector Space Model with semantics by using the results of a search engine in the background. According to the page counts of the query and the words in the service description file, the Normalized Google Distance between the query and each word in the document is calculated and the documents are then ranked according to this semantic similarity measure.