US 5,687,383 A · Translation rule learning scheme for machine translation

Abstract

A translation rule learning scheme for a machine translation using a computer, in which an original sentence in a first language and a translated sentence in a second language of a bilingual text are analyzed to obtain dependency structures indicating a predicate and a case information associated with the predicate in each language; case frame correspondences between the first and second languages are estimated by estimating correspondences of words in the dependency structures in the first and second languages according to a bilingual dictionary; the estimated case frame correspondences are modified by comparing the case information in the dependency structures in the first and second languages and estimating a missing case information in one language from an existing case information in another language; and translation rules are extracted from the modified case frame correspondences.

Claims20

Dependent claims

A method of translation rule learning for a machine translation, comprising the steps of:

analyzing an original sentence in a first language and a translated sentence in a second language of a bilingual text to obtain dependency structures indicating a predicate and a case information associated with the predicate in each language;
estimating case frame correspondences between the first and second languages by estimating correspondences of words in the dependency structures in the first and second languages according to a bilingual dictionary;
modifying the estimated case frame correspondences by comparing the case information in the dependency structures in the first and second languages and estimating a missing case information in one language from an existing case information in another language;
and
extracting translation rules from the modified case frame correspondences.

2 more elements

A translation rule learning system for a machine translation, comprising:

a memory for storing a bilingual text containing original sentences in a first language and translated sentences in second language;
analysis means for analyzing an original sentence in the first language and a translated sentence in the second language of the bilingual text to obtain dependency structures indicating a predicate and a case information associated with the predicate in each language;
estimation means for estimating case frame correspondences between the first and second languages by estimating correspondences of words in the dependency structures in the first and second languages according to a bilingual dictionary;
modification means for modifying the estimated case frame correspondences by comparing the case information in the dependency structures in the first and second languages and estimating a missing case information in one language from an existing case information in another language;
and
extraction means for extracting translation rules from the modified case frame correspondences.

3 more elements

Dependencies are inferred from the claim text.

Translation rule learning scheme for machine translation

Life of this patent

Translation rule learning scheme for machine translation

Life of this patent

Abstract

Claims20