This paper proposes a lightweight mechanism to combine deductive and inductive inference engines and describes an application in which the resulting functionality is shown to add significant vale. Within the unified setting of deductive database technology and inductive logic programming, the combination of inference engines is much easier and the resulting systems much better integrated. This combination is important, e.g., if software agents are indeed to become widespread in the technical context of pervasive distributed computing that the web promises. Software agents will need to automate more decision-making tasks, for some of which they may need to learn new knowledge on the fly. The dynamic, volatile nature of the environment in which they are expected to operate make this need all the more acute. A recent, challenging example is the web services approach to distributed business computing. This paper describes an application of a combined inference engine built by the authors in which the problem of characterizing whether one web service is likely to be a substitute for another is solved in a fluent, disencumbered manner such as will be needed if it is to be deployed by software agents.
展开▼
