Combining partitions and modal logic for user modeling

Abstract

User modeling means acquiring, representing and managing assumptions about users of software systems. Inference processes as part of the representation component can be used to extend the base of assumptions, thus supporting the acquisition task. In the user modeling shell system BGP-MS, a partition approach to user model representation and reasoning has beentaken. It offers the possibility to maintain different types of assumptions about the user in a partition hierarchy; reasoning about the assumptions of one type is possible. In addition, useful inferences are Mready built into the partition mechanism of BGPMS. However, negative assumptions about what users do not know or want, or relationships between assumption types cannot be maintained in BGP-MS. Since such user modeling knowledge can be formulated with modal logic, a method was developed that combines techniques of reasoning with modal logic and the partition mechanism. Algorithms for translating modal expressions into first-order logic are employed. Thus, new representation and reasoning facilities were implemented into BGPMS. One of the main quMities of the combination method is that the advantages of the partition approach are preserved.