Structured document retrieval using Dempster-Shafer's Theory of Evidence: Implementation and evaluation

Documents are often structured, for example into chapters, each with sections, and so on. The representation of structured documents should provide for a focussed retrieval of those components, individual or aggregated, of the documents that are most relevant to an information need. In previous work, we developed a model for the representation and the retrieval of structured documents. This model is based on Dempster-Shafer's theory of evidence. In this paper, we propose an implementation of our model to evaluate its e ectiveness. We are particularly interested in studying whether the representation of aggregated components within the Dempster-Shafer theory allows for focussed retrieval of structured documents. We accomplish a basic implementation so that we can concentrate solely on evaluating the appropriateness of the representation. We carried out a number of experiments which indicate that the use of the DempsterShafer theory in modelling the representation of aggregated components is e ective.