GrammarSoft ApS

VISL and constraint grammar

This paragraph is still not written:

Something about the purpose of CG, introducing 'running text' and 'tags'

Constraint Grammar is not, as most other approaches to language named 'XX Grammar' are, a theory of grammar, but a formalism for rules enabling linguists to construct algorithms that correctly tag words in running text. The tags used in Constraint Grammar could be taken from some specific scientific theory of grammar, but usually comes from what may be called 'standard' or 'school grammar', the 'theory' of grammar used in primary and secondary schools and also usually seen as the basis that theoretical linguistics builds on. This includes part-of-speech (noun, verb, adjective etc.) and sentential function (subject, object, main verb, finite verb, adverbial etc.). These levels have already been treated within the VISL-project; the purpose of the project described below is, given a text already automatically tagged with POS and sentential function, to tag it with 'cases' as those introduced by Charles Fillmore's theory of Case Grammar (Fillmore, 1970,1987), which I will give an introduction to in section 2.1. This procedure also demonstrates how the output of one constraint grammar may be the input of another, so that for example one constraint grammar tags a raw text with parts of speech-tags, the next grammar uses these tags to tag the text with grammatical functions, which again is the input to my case-tagger.¹ The purpose of this analysis is, among others, to be used as part of the source language analysis component of a machine translation system.

• Strategies in Constraint Grammar