Extracting REST resource models from procedure-oriented service interfaces

• We propose a technique for extracting REST resource models from procedural services.
• The proposed approach is highly automated, implementation-agnostic and efficient.
• We demonstrate the approach through a real-world case study.
• We evaluate the approach using 70 publicly available service interfaces.
• Results indicated significant productivity benefits, high efficiency and accuracy.

During the past decade a number of procedure-oriented protocols and standards have emerged for making service-offering systems available on the Web. The WS-* stack of protocols is the most prevalent example. However, this procedure and message-oriented approach has not aligned with the true potential of the Web's own architectural principles, such as the uniform identification and manipulation of resources, caching, hypermedia, and layering. In this respect, Resource Oriented Architectures based on the REST architectural style, have been proposed as a possible alternative to the operation-based view of service offerings. To date, compiling a REST API for back-end procedure-oriented services is considered as a manual process that requires as input specialized models, such as, service requirements and behavioral models. In this paper, we propose a resource extraction method in which service descriptions are analyzed, using natural language processing techniques and graph transformations, in order to yield a collection of hierarchically organized elements forming REST resources that semantically correspond to the functionality offered by the service. The proposed approach has been applied as a proof of concept with positive results, for the extraction of resource models from a sizable number of procedure-oriented Web Service interfaces that have been obtained from an open service directory.