Statistical method and apparatus for learning translation relationships among phrases
AssigneeMicrosoft Corporation
InventorRobert C. Moore
Life of this patent
Estimating the term…
Abstract
The present invention learns phrase translation relationships by receiving a parallel aligned corpus with phrases to be learned identified in a source language. Candidate phrases in a target language are generated and an inside score is calculated based on word association scores for words inside the source language phrase and candidate phrase. An outside score is calculated based on word association scores for words outside the source language phrase and candidate phrase. The inside and outside scores are combined to obtain a joint score.
Claims28
A method of identifying a translation relationship between a phrase in a source language and a phrase in a target language, comprising:
- receiving access to an aligned pair of multiple word units, one being a source unit in the source language and another being a target unit in the target language, the source language phrase being identified in the source unit;
- generating at least one candidate phrase in the target unit, the candidate phrase being a hypothesized translation of the source language phrase;
- calculating a score for each candidate phrase including an inside component based on associations between words inside the source language phrase and candidate phrase and an outside component based on associations between words in the source unit but outside the source language phrase and words in the target unit but outside the candidate phrase;
- and
- identifying the translation relationship between the source language phrase and candidate phrase based on the score.
Dependencies are inferred from the claim text.