Statistical machine translation method using dependency forest

Abstract

The present invention relates to a dependency forest for effectively processing a plurality of dependency trees, and which uses a plurality of dependency trees for tree-based statistical machine translation. The method of the present invention uses the dependency forest to generate translation rules and dependency language models, and applies the generated translation rules and dependency language models when converting native language text into target language text, thereby improving translation performance.

Description

technical field [0001] The present invention relates to a statistical machine translation method using a dependency forest, in particular to a statistical machine translation method using a dependency forest, which can improve the translation ability by performing the following operations: performing dependency analysis on a bilingual corpus to generate multiple A dependency tree, combine the generated multiple dependency trees to generate a dependency forest, use the dependency forest to generate translation rules and dependent language models, and then apply the generated translation rules and dependent language models when converting the source language text to the target language text. Dependent language model. Background technique [0002] figure 1 The dependency tree of the sentence "He saw a boy with a telescope" in English is shown. Such as figure 1 As shown, the arrow points from the child node to the parent node. A parent node often represents the head of a chi...

AI Technical Summary

Problems solved by technology

[0018] A recent dependency parser (parser) has high performance (English 91%, Chinese 87%), but this dependency parser is not enough for statistical machine translation

Because string-to-tree systems rely on the best tree for parameter evaluation, the quality of rule tables and dependency language models can suffer from errors in parsing and thus translation errors

Images