Open Issue: Tense and Aspect

Mentor: EmilyBender (ebender at u dot washington dot edu, http://faculty.washington.edu/ebender)

Co-mentors welcome!

Problem statement

We are in need of a standardized representation of tense and aspect within MRS. This representation should be able to encode all contrasts which are syntactically marked, while underspecifying further distinctions in such a way that they can be computed if necessary from the MRS without reference to the syntax and without `overgenerating’ (finding readings which are not licensed, given the syntax). This representation should target, in the first instance, grammatical tense and aspect (as opposed to lexical aspect or aktionsart).

Deep/shallow integration

For the purposes of RMRS-based integration, the ideal system would allow for a mapping between morpohological (or morphosyntactic) tense/aspect and semantic tense/aspect. That is, a mapping between what can be found through morphological analysis and the representations built by deep grammars.

Matrix customization

The ideal system will define a superset of predicates (or operators or types to serve as the value of features) from which individual grammars can draw. This is a prerequisite to building a tense/aspect section for the Matrix customization system (http://www.delph-in.net/matrix/modules.html). A complication is that contrasts which are routinely marked in some languages are routinely left underspecified in others. Indeed, some languages grammaticize mostly aspect while others grammaticize mostly tense.

Autogeneration of transfer rules

To the extent that grammars can draw from a common vocabularly of tense/aspect information, the relations among which we understand, it should be possible to automatically generate transfer rules between any two grammars based on how they are generated from the Matrix customization system.

Existing literature

Goss-Grubbs, David. 2005. An Approach to Tense and Aspect in MRS. Master’s Thesis. University of Washington.

Copestake, Ann, Dan Flickinger, Carl Pollard, and Ivan A. Sag. 2005. Minimal Recursion Semantics: An Introduction. Research on Language and Computation.

Bender, Emily M. and Dan Flickinger. 2005. Rapid Prototyping of Scalable Grammars: Towards Modularity in Extensions to a Language-Independent Core. Proceedings of IJCNLP-05 (Posters/Demos), Jeju Island, Korea.

Last update: 2011-10-09 by anonymous [edit]