Maureen and Mike Mansfield Library

Type of Document	Master's Thesis
Author	McCloy, Daniel Richard
URN	etd-03012007-112716
Title	Automated Adaptation Between Kiranti Languages
Degree	Master of Arts
Department	Linguistics
Advisory Committee	Advisor Name Title Dr. Anthony Mattina Committee Chair Dr. David E. Watters Committee Member Dr. Mizuki Miyashita Committee Member Dr. Nancy Mattina Committee Member
Keywords	disambiguation Kiranti typology automated adaptation
Date of Defense	2006-11-30
Availability	unrestricted
Abstract McCloy, Daniel, M.A., December 2006 Linguistics Automated Adaptation Between Kiranti Languages Chairperson: Dr. Anthony Mattina Minority language communities that are seeking to develop their language may be hampered by a lack of vernacular materials. Large volumes of such materials may be available in a related language. Automated adaptation holds potential to enable these large volumes of materials to be efficiently translated into the resource-scarce language. I describe a project to assess the feasibility of automatically adapting text between Limbu and Yamphu, two languages in Nepal’s Kiranti grouping. The approaches taken—essentially a transfer-based system partially hybridized with a Kiranti-specific interlingua—are placed in the context of machine translation efforts world-wide. A key principle embodied in this strategy is that adaptation can transcend the structural obstacles by taking advantage of functional commonalities. That is, what matters most for successful adaptation is that the languages “care about the same kinds of things.” I examine various typological phenomena of these languages to assess this degree of functional commonality. I look at the types of features marked on the finite verb, case-marking systems, the encoding of vertical deixis, object-incorporated verbs, and nominalization issues. As this Kiranti adaptation goal involves adaptation into multiple target languages, I also present a disambiguation strategy that ensures that the manual disambiguation performed for one target language is fed back into the system, such that the same disambiguation will not need to be performed again for other target languages.
Files	Filename Size mccloydaniel.pdf 1.03 Mb