Improving Heuristic-Based Temporal Analysis of Narratives with Aspect Determination

In previous work we presented an a lgo r i thm for tense interpretat ion wh ich employs a temporal focus to determine the intended temporal relations between the states and events mentioned in a narrative In this paper, we propose a new two-phased c lass i f ica t ion scheme for aspect Each situat ion described in an utterance is first classified as static (state) or dynamic (event) and i f dynamic as te l ic (event w i t h a culminat ion point) or atelic (evenl without a culmination point) Then, independent of the class the v iew of the situation is ident i f ied either as a point or as an interval We then demonstrate how the determination of aspect can be integrated into our tense interpretation a lgor i thm to produce a richer analysis of temporal relations Our classif icat ion for aspect is more detai led than most of the exist ing schemes a l lowing us to extract the interval relations between situations and cover a wide range of English narratives 1 I n t r o d u c t i o n Temporal analysis plays an important role in the understanding of natural language discourse and in particular, narratives and natural language input to plan recognition systems Its goal is to determine the temporal relations between the states and events mentioned in a discourse, such as whether one event occurs before or dur ing the t ime of another We generally refer to states and events as situations Our focus has been on the temporal analysis of "simple" narratives in which there is one speaker and al l utterances descnbe "actual" siluauons, which arc those asserted to have occurred to be occurring or to occur al a future t ime The reasons for these restr ict ions are to avoid conf l ic ts in bel ief w i t h 1 In this paper we restrict ourselves to qualitative temporal relations Quantitative information such as one event occurs before another by 5 minutes is not further distinguished different speakers and to exclude situations described in modal in tent ional , negated or f requentat ive contexts [Passonneau 1988] There are many factors that may be integrated into such temporal analysis inc lud ing tense aspect, temporal adverbials and connectives discourse structures and even real-world knowledge In Song and Cohen [1991] we presented an algorithm for temporal analysis that is based solely on tense interpretation The algor i thm expands the not ion of temporal focus in Webber [1988] lo determine the intended tempora l re lat ions in a narrative More specif ical ly we proposed a set of heuristics for determin ing the temporal ordering between a new situation and the situation in focus We also proposed a set of constraints for tracking the changes of a temporal focus The value of our algorithm is lo al low an implementation that can bt built into a working system Being heunstic-based the algor i thm may admit counter-examples, but u is designed to produce the preferred interpretation ol intended temporal ordenngs in the absence of high level information such as discourse structures and real world knowledge In this paper we further expand our algorithm for temporal analysis wi th aspect determinauon which is to analyze wh ich part of a situation is v iewed in an utterance We regard tense as a grammatical notion that is expressed solely by verb forms For example "John wrote a letter" is in the Simple Past tense since the verb "wri te" is in its past fo rm In contrast aspect is seen as a semantic not ion w h i c h has to be determined compositionally f rom certain lexical items in an utterance Tense and aspect together help provide a more accurate analysis for narratives al lowing us lo extract interval relations between entire situations and cover a wide range ol utterances The paper therefore demonstrates that it is possible lo develop heurist ics for temporal analysis wh ich incorporate aspectual information We also reveal the value of aspectual information in temporal analysis by 2 In Song (1991) we also considered those cases where lense sequences may lead lo more than one interpretation and how discourse structures and real-world knowledge can be used to resolve such ambiguities 1338 NATURAL LANGUAGE