FuDoCS: A Web Service Composition System Based on Fuzzy Dominance for Preference Query Answering

Modern enterprises are increasingly moving towards a service oriented architecture for data sharing by putting their data sources behind services, thereby providing an interoperable way to interact with their data. This class of services is known as DaaS (Data-as-a-Service) services. DaaS Composition is a powerful solution to answer the user’s complex queries by combining primitive DaaS services. User preferences are a key aspect that must be considered in the service composition process. A more general and suitable approach to model preferences is based on fuzzy sets theory [3]. Fuzzy sets are very well suited to the interpretation of linguistic terms and constitute a convenient way for a user to express her/his preferences. For example, when expressing preferences about the price of a car, users often employ fuzzy terms like rather cheap, affordable, etc. However as DaaS services proliferate, a large number of candidate compositions that would use different (most likely competing) services may be used to answer the same query. Hence, it is important to set up an effective service composition framework that would identify and retrieve the most relevant services and return the top-k compositions according to the user preferences. Challenges. Consider the services from the car e-commerce in Table 1 (i.e., typical services that can be provided by systems like the e-Bay). The symbols “$” and “?” denote inputs and outputs of services, respectively. Services providing the same functionality belong to the same service class. For instance, S21, S22, S23, S24 belong to the class S2. Each service has its (fuzzy) constraints on the data it manipulates. For instance, the cars returned by S21 are of cheap price and short warranty. Assume that the user Bob wants to buy a car. He sets his preferences and submits the following query Q1: “return the French cars, preferably at an affordable price with a warranty around 18 months and having a normal power with a medium consumption”. Answering Q1 raises the following challenges: (i) how to understand the semantics of the published services to select the relevant ones that can contribute