Modelling and exploiting the temporal information associated with complex 'narrative' documents

Abstract

This paper supplies some details on the solutions adopted in a narrative knowledge representation language (NKRL) context for representing/managing the temporal information included within 'narrative' multi-media documents. They are based on the association of two notions, the 'category' and the 'perspective' of dating. The first allows us to describe the logical implications of the distribution on the time axis of the timestamps associated with an event. The second takes into account the different types of fuzziness with which these timestamps can be captured. From an operational point of view, the two notions have also an important inferential function since they represent the basic building blocks of an advanced indexing system for NKRL knowledge bases. This allows us to reduce to sets of simple comparison operations the computational procedures needed for selecting all the data characterised by temporal information that are congruent with the temporal criteria specified in a given query.