A summarisation logic for structured documents

The logical approach to Information Retrieval tries to model the relevance of a document given a query as the logical implication between documents and queries. In early work, van Rijsbergen states that the retrieval status value of a document given a query is proportional to the degree of implication between a document and a query. Based on this, several probabilistic logics for information retrieval have been conceived, which add an additional layer of abstraction to the information retrieval task: probabilistic models for the retrieval of documents are expressed in those logics, rather than implemented directly.

The aim of the research presented here is to develop a logic for the IR task of document summarisation. Such a summarisation logic adds an abstraction layer to the summarisation task, similarly to the way a logic for document retrieval adds a layer to the retrieval task. Probabilistic models for document summarisation will thus be expressed as logical formulae, with the actual implementation hidden by the logical expressions. The "extract-worthyness" of textual components in this logic is measured as the degree of implication from those textual components to their surrounding contexts, providing a measure of how much said components are "about" the context within which they are situated.