Article Preview
Top1. Introduction
Web-service discovery is a major challenge for automatic web service selection and integration. Discovery depends not only on web service availability but also on the ability to understand the objective and the function of the available web services, using the descriptive details. Several approaches and techniques have been proposed to address this challenge, including syntactic-based and semantic-based approaches (Batra & Bawa, 2010; Klusch & Kaufer, 2006). Such approaches assume that correct and valid service descriptions already exist. In practice, the reality is different.
The major concern with semantic Web service discovery is the vast majority of already existing web services that are not associated with semantics. In general, the semantic approach suffers from “a cold-start problem” because it assumes that a corpus of previously annotated services is available (Batra & Bawa, 2010; Klusch & Kaufer, 2006; Kiefer, Bernstein, Lee, Klein, & Stocker, 2007). These findings provide momentum to reconsider the syntactic approach, which depends on the web service description to empower the other semantic initiatives. SAWSDL (Semantic Annotation for WSDL) (Farrell & Lausen, 2007) is an example of such new approach. In conclusion, Web-service semantics should be considered an added layer to the standard Web-service description part.
We analyzed more than 35,000 web services from eight well-known datasets (Al-Masri & Mahmoud, 2012; Service-Repository, 2011; Service Platform, 2012; OPOSSum, 2012; Zheng & Lyu, 2012). In these datasets, we found that 49% of registered services are active and 51% are not active. Further analysis of WSDL-based web services shows that only 19% have a web service description. Therefore, 81% of web services are not being considered in the syntactic discovery phase. Our findings are also supported by previous studies (Sabou, Wroe, Goble, & Mishne, 2005; Fan & Kambhampati, 2005).
Recently, quality of service (QoS) in web service discovery has gained considerable attention as a vital research topic (Garofalakis, Panagis, & Sakkopoulos, 2004; Sabou, Wroe, Goble, & Mishne, 2005; Al-Masri & Mahmoud, 2007). In this context, it is important to distinguish between two research domains within web service discovery. The first approach focuses largely on Web-service functionality and performance, such as response time, latency, availability, accessibility, and security. The second approach focuses on the quality of a Web-service interface, largely assessing the Web-service interface in terms of complexity. The importance of the second approach lies in the fact that without easily understandable Web-service interface functionality, it is not reasonable to expect successful discovery or usability (Garofalakis, Panagis, & Sakkopoulos, 2004).
Our research aims at introducing a new approach for fixing the current problem of poorly written Web-service descriptions and providing practical control over Web-service interface quality to minimize and avoid such bad practices. This approach will help service providers in improving their service quality by satisfying discovery needs and measures. Service registry moderators and brokers will gain the advantage of providing valid and well-defined web services based on better classification and clustering that supports service discovery and composition. Consumers will effectively use the shared interface quality metrics during discovery to select services that match their quality needs and development constraints with minimal effort.
We focus on schema clarity and WSDL-based Web services due to their wide implementation and acceptance in the WS provider community. We primarily depend on extracting meaningful words from the used elements and parameter names to empower the description part. We use the developed clarity and quality metrics in measuring the quality of Web-service interfaces during Web-services registration and updates. Finally, we share the measured metrics so that developers, registry moderators, and consumers can gain the benefits of our approach during the discovery and selection phases.