Towards the Automatic Selection of Prepositions in the Interactive Computer-Assisted Translation of English to French

Prepositions account for a large percentage of interactions in the word-sense selection phase of computer-assisted translation. Reducing the amount of human input during interactive translation greatly improves the cost-effectiveness of the process, and the elimination of interaction on selected prepositions results in a 30% reduction in over-all interactions. In-depth studies of twelve common prepositions, based on analysis of a 500,000-word corpus of Latter-day Saint English, provides insights into ways of reducing the interaction on these prepositions to an acceptable level. Junction Grammar allows for structural disambiguation but as of yet has not isolated possible internal structure of prepositions. Analysis of each preposition examines the feasibility of accommodating variant meanings of preposition by including information on the different nodes of referment.

Thesis Author: DeSantis, Christopher C.

Year Completed: 1980

Thesis Chair: Eldon Lytle