A Language-Neutral Representation of Temporal Information

We propose a framework for representing semantic tense that is language-neutral, in the sense that it represents what is expressed by different tenses in different languages in a shared formal vocabulary. The proposed framework allows the representation to retain surface distinctions for particular languages, while allowing fully semantic representations, such as a representation of event sequence, to be derived from it. The proposed framework also supports the incorporation of semantic tense information that does not derive from grammatical tense, but derives instead from other expressions such as time adverbials. The framework is currently implemented in NLPWin, a multi-lingual, multi-application natural language understanding system currently under development at Microsoft Research, but the representational framework is in principle independent of any particular system.


> Publications > A Language-Neutral Representation of Temporal Information