A model for time granularity in natural language

Abstract

We propose a model for dealing with time granularity in natural language. In contrast with many other fields where granularity levels are essentially quantitative, in natural language we are confronted with a more subtle and qualitative kind of granularity where the different levels are not always precisely defined. Our model representing such phenomena is based on time units and intervals. Time units are considered as indivisible chunks of time and corresponding in many aspects to the usual notion of temporal points with the difference that the former are durative whereas the latter represent traditionally instantaneous events. Relations between time units and/or intervals are exhaustively studied and a relation algebra is defined which extends the traditional algebra on points and intervals. In particular a containment relation is introduced in order to express changes in granularity. A logic with restricted quantifiers is proposed for formalizing temporal knowledge and some examples are discussed which show the relevance of the model for natural language.