A text mining based approach for web service classification

Web services have evolved as a versatile and cost effective solution for exchanging dissimilar data between distributed applications. They have become a fundamental part of service oriented architecture. However one of the major challenges in service oriented architecture is to figure out what a service does and how to use its capabilities without direct negotiation with the service provider. Discovering and exploring web services registered with Universal Description, Discovery and Integration registry or Web Services-Inspection documents requires exact search criteria such as service category, service name and service URL. Web Service Description Language (WSDL) document allows web services clients to learn operations, communication protocols and correct message format of service. Manually analyzing WSDL documents is the best approach but most expensive. This paper proposes a text mining approach to (1) automatically classify services to specific domains and (2) identify key concepts inside service textual documentation. This approach is validated on a dataset of 600 web services categorized into 8 fields yielding accuracy up to 90 %. Our classification approach can be used to focus user queries to a refined set of web service categories.