Applying Tools and Techniques of Natural Language Processing to the Creation of Resources for Less Commonly Taught Languages

Abstract

This paper proposes that research results from the area of naturallanguage processing could effectively be applied to creating softwareto facilitate the development oflanguage learning materials foranynaturallanguage. We will suggest that a knowledge-elicitationsystem called Boas, which was originally created to support amachine-translation application, could be modified to supportlanguage-learning ends. Boas leads a speaker of any natural Ianguage,who is not necessarily trained in linguistics, through a seriesof pedagogically-supported questionnaires, the responses to whichconstitute a" profile" of the language. This profile includes morphological,lexical and syntactic information. Once this structuredprofile is created, it can feed into virtually any type of system,including one to support language learning. Creating languagelearningsoftware using a system like this would be efficient in twoways: first, it would exploit extant cutting-edge research and technologiesin naturallanguage processin~ and second, it would permita single tool to be used for all languages, including less commonlytaught ones, for which limited funding for resource development isa bottleneck.