Extending Enterprise Service Design Knowledge Using Clustering Marcus Roy 1,2 , Ingo Weber 2,3,⋆ , and Boualem Benatallah 2 1 SAP Research, Sydney, Australia 2 School of Computer Science & Engineering, University of New South Wales 3 Software Systems Research Group, NICTA, Sydney, Australia {m.roy,ingo.weber,boualem}@cse.unsw.edu.au Abstract. Automatically constructing or completing knowledge bases of SOA design knowledge puts traditional clustering approaches beyond their limits. We propose an approach to amend incomplete knowledge bases of Enterprise Service (ES) design knowledge, based on a set of ES signatures. The approach employs clustering, complemented with vari- ous filtering and ranking techniques to identify potentially new entities. We implemented and evaluated the approach, and show that it signifi- cantly improves the detection of entities compared to a state-of-the-art clustering technique. Ultimately, extending an existing knowledge base with entities is expected to further improve ES search result quality. 1 Introduction In large-scale software development efforts, such as enterprise Service-Oriented Architectures (SOAs), ESs are commonly developed using service design guide- lines – guarded by SOA Governance [13,8,16]. These guidelines may include entities (e.g. Sales Order) and naming conventions used to construct unambigu- ous ES operation names, referred to as ES signatures. For instance, consider the ES signature “SalesOrderItemChangeRequestConfirmation In” from SAP’s ESR 1 . Although such service design knowledge is largely used to consistently design ES signatures, it can also be utilized for other applications, e.g. tools to automatically generate, duplicate-check and validate compliant ES signatures as well as to search for ESs. We tested the latter in an ongoing, separate stream of work, where we use service design knowledge in an entity-centric keyword search for ESs, with highly encouraging results. However, such service design knowl- edge can be incomplete, e.g. due to partial modeling, or become outdated as related systems and business requirements evolve over time. Particularly when customers tailor an off-the-shelf enterprise application to their specific need, they may disregard design guidelines when developing new ESs. Therefore, resulting ⋆ NICTA is funded by the Australian Government as represented by the Department of Broadband, Communications and the Digital Economy and the Australian Research Council through the ICT Centre of Excellence program. 1 Enterprise Service Registry: http://sr.esworkplace.sap.com C. Liu et al. (Eds.): ICSOC 2012, LNCS 7636, pp. 142–157, 2012. c  Springer-Verlag Berlin Heidelberg 2012