Process-annotated service discovery facilitated by an n-gram-based index

Service discovery in web service infrastructure must consider abstract processes to guarantee successful business interactions with the matched services. Unfortunately, matchmaking process annotations requires a significant processing effort, thus, finding relevant service providers in a large collection is time consuming and requires more efficient access methods than sequential scanning. The paper presents an approach supporting more efficient querying of abstract processes by introducing an abstraction to reduce the computational complexity of the comparison operation. Experimental studies on a collection of over 800 abstract business processes derived from the RosettaNet PIP specification show that our approach outperforms naive sequential scanning by up to an order of magnitude.