IBM Israel Research Seminars

The vast number of services within organizations and on the Web raises the need for efficient retrieval solution, which will allow users to locate and analyze services. The general definition of a 'service', which describes an action taken according to a given input, consequences and qualities (e.g. price), is useful for describing a wide variety of resources in a range of fields. These fields include electronic commerce (e.g. flights), electronic government (e.g. tax payment) and more. Nowadays, considerable research is invested in leveraging logic-based models for service discovery and composition. These methods retrieve services with very high precision and certainty. However, they require the availability of full semantic models of the services, which is not likely to be feasible in many fields.

Our research proposes an approximate approach for service retrieval, making it usable for human users and with real-world services. The theory that lies at the heart of approximate service retrieval is aiming to describe relations between services, such as equivalence, resemblance and containment, in various levels of accuracy. The research uses fuzzy semantic networks as a conceptual model for service representation, analyzing relations between services by reasoning over the structural and behavioral compatibility of their respective semantic networks. The utilization of fuzzy models enables flexible query evaluation by approximating single and composed services. Furthermore, it allows services to be indexed from the Web and other semi-structured information sources.

In order to evaluate our approach, a search engine for business processes was constructed, supporting approximate retrieval and assembly of services. In the course of implementing the search engine, several challenges were addressed, resulting in methods for compact representing of fuzzy service descriptions and efficient algorithms for querying massive semantic networks.

About the Speaker
Eran Toch is perusing a Ph.D. (direct program) in the Information Systems Field, Faculty of Industrial Engineering and Management, the Technion --- Israel Institute of Technology. His research concerns novel methods for engineering knowledge-based information systems, including semantic models and information retrieval. For this work, he was awarded with the Israel Ministry of Science's Levi Eshkol scholarship for scientific achievement. Eran has been developing computer programs and games since the age of 12, and developing Web systems since 1995. He worked as a developer and product manager at the startup company Kinetica (acquired by NetVision at 1999), and holds a BS.c. in computer science and honors program from the University of Haifa. In addition with his Ph.D. studies, He serves as an adjunct lecturer in the Technion, teaching analysis and specification of information systems.