Whereas temporal interval algebras appear to be the class of formalisms from Artificial Intelligence that is currently privileged by such researchers who are trying to handle temporal information as found in legal documents, there does not appear to be a consensus, within AI for Law, about a structured way of integrating the representation of the temporal data into the broader picture of treating the events or the overall plot. We describe here a knowledge representation formalism well-suited to take into account the temporal characteristics of narratives (of narrative documents). In these documents, the main part of the information content consists in the description of 'events' that relate the real or intended behaviour of some 'actors' (characters, personages, etc.). Narrative documents of an industrial and economic interest correspond, for example, to news stories, corporate documents (memos, policy statements, reports and minutes), normative and legal texts, intelligence messages, representation of the patient's medical records, etc. The formalism we present here is characterised by the following main properties: (i) it provides some general tools to deal with the 'fuzziness' which, in concrete situations, is inherently associated with the representation of any sort of 'timestamp'; (ii) it offers a way of implementing an efficient temporal reasoner, able to deal, for example, with the purely mechanical aspects of the well-known problem concerning the 'persistence of a situation'; (iii) it makes use of some second order representation tools (binding structures) to replace, to a certain extent, the interval algebra tools in the Allen style.
IntroductionTemporal data are ubiquitous in any computer application where information about real life is captured. Such is the case of legal records. In laws themselves, a temporal pattern can be extracted from the given article of law. Argumentation, hypotheticals, natural-language texts, are also likely to involve some (even trivial) temporal relation, even though the situation envisaged may be admittedly removed from reality. Database records in legal or judiciary applications are likely to include some attribute about time. The temporal raw data that can be extracted are clearly not always emplotted into a broad story; much less so in the computer representations adopted. Yet, if you think of legal narratives about which a court is called to decide, it is easy to see how the set of temporal relations cannot be reasonably divorced from the commonsense texture of actions, events, and situations. For example, let us recall the distinction between the supposedly object-level narrative, what semiologist of law Bernard Jackson (1988, 1996), has termed the 'semantics' of the legal narrative, and what the presentations in courts make out of it, at a meta-level so to speak, or, in Jackson's terms, the 'pragmatics' of the legal narrative. It is essential for the discipline of AI & Law, therefore, that a handy representational tool be 1 provided, which incorporates the ...