Abstract

In this paper, we show that there is a mismatch between information retrieval (IR) and database (DB) concepts, and we devise solutions for this problem. DB oriented approaches have to distinguish between the logical and the content structure of objects, and should also consider the layout structure. Data independence — not regarded in IR before — can be achieved by using the notion of vague predicates. Since IR is based on uncertain inference, data models with uncertainty are required for an integrated IR-DB system. For this purpose, we present a probabilistic relational algebra. As extensions, probabilistic Datalog yields a more expressive query language, whereas a probabilistic nested relational model is more appropriate for modelling document structures. 1

Citations