Probabilistic, object-oriented logics for annotation-based retrieval in digital libraries

In this paper we introduce POLAR, a probabilistic object-oriented logical framework for annotation-based information retrieval. In POLAR, the knowledge about digital objects, annotations and their relationships in a digital library repository can be modelled considering certain characteristics of annotations and annotated objects. Insights about these characteristics are gained by an analysis of the annotation models behind existing systems and a discussion of an object-oriented, logical view on relevant objects in a digital library. Retrieval methods applied in a digital library should take annotations into account to satisfy users' information needs. POLAR thus supports a wide range of flexible and powerful annotation-based fact and content queries by making use of knowledge and relevance augmentation. An evaluation of our approach on email discussions shows performance improvements when annotation characteristics are considered