Problem-Solving Modelling in Deductive Web Mining Vojtˇ ech Sv´ atek 1 , Martin Labsk´ y 1 , Annette ten Teije 2 , Miroslav Vacura 1 and Frank van Harmelen 2 1 Department of Information and Knowledge Engineering, University of Economics, Prague, W. Churchill Sq. 4, 130 67 Praha 3, Czech Republic {kavalec,svatek}@vse.cz 2 ..., Vrije Universiteit Amsterdam, ... Abstract. Deductive web mining recently gained on importance as sup- porting technology for building the semantic web. It is typically be- ing used in a stand-alone and ad hoc manner, however, its knowledge- intensive nature together with (presumed) ubiquitous usage calls for knowledge-level modelling. Yet, even deductive web mining is (also) data- intensive, and hence cannot be simply mapped on problem-solving meth- ods used in the context of traditional artificial intelligence. For this pur- pose, we developed a simple multi-dimensional framework, which enables to characterise deductive web mining tasks and methods at the abstract level. In the paper, we describe the framework, show some of its in- stantiations, and attempt to formalise them using the state-of-the-art technology of problem-solving modelling, namely the UPML language and tools. 1 Introduction Problem-solving methods (PSMs) became one of the hottest topics in knowl- edge engineering as early as in the mid 80s, and still attract considerable in- terest in the community. The KADS methodology achieved its industrial-level maturity in CommonKADS [16], which mimics many aspects of conventional software engineering while fulfilling the requirements of knowledge-intensive ap- plications. On the other hand, the IBROW project [1] represents the cutting- edge research in its endeavour for PSM-based component assembly ‘on the fly’. Until recently, PSMs have been understood as specific for knowledge-intensive but ‘data-temperate’ tasks. This contrasted with the proliferation of domain ontologies that are currently being adopted (in lightweight represenation for- malisms) in various domains far beyond traditional AI research. The question whether data-intensive tasks could benefit from the introduction of PSMs has obviously come to the mind of several researchers. The authors of the Internet Reasoning Service [6], developed within IBROW, suggest that ”more ‘mundane’ levels of service provision” should be supported in the future, and even mention data analysis. Information ‘brokers’ (and PSM libraries) for textual information access tasks have been designed by other groups within the same project [3, 2].