The Service-Oriented Computing paradigm aims to support automated discovery and selection of web services according to user?s requirements. At present, user?s requirements are often represented in certain existing standard interoperable service description languages such as WSDL/OWL-S. However, general service requestors may find such languages hard to use directly due to the reason that service requirements are often partially elicited and fragmented.

In this paper, we propose an automated Service Requirements Elicitation Mechanism (SREM) to help extract and accumulate relevant knowledge on service requirements. First, the SREM elicitation approach proposes to use a list of questions to narrow generic service requirements down to specific expressions of user preferences. Then, a service requirements and capability ontology is adopted to capture services requirements in breadth and precision. By integrating service requirements issued by different requestors, SREM provides non-trivial requirements guidelines and heuristic rules on service publication and discovery, also provided is a service requirements analysis mechanism that improves the accuracy of service discovery and efficiency of service composition continuously.