NSF Award
9302903
Kosaka 9303013 This is the second year of a three-year collaborative project between Monmouth College and New York University seeks to gather correspondences between expressions in source and target parallel corpora in Japanese and English with the objective of generalizing these correspondences to create rules for use in a translation system. The performance of translation systems has been often limited to a great extent by our ability to encode information about such correspondences manually. In this project the texts in the source and target languages are parsed and then syntactically regularized, producing regularized parse trees. A tree-matching procedure aligns the corresponding trees from the source and target texts, producing a set of detailed correspondences between source and target structures. A set of sublanguage (semantic) word classes is then defined and these correspondences are generalized to the extent possible using these word classes; the result is a set of rules for the transfer phase of a translation system. The approach uses corresponding programming language manuals written in Japanese and English.
