Translation rule extract method and translation method based on tree-to-tree translation model

Abstract

The invention provides a translation rule extract method based on a tree-to-tree translation model, comprising the following steps: 1), analyzing source language strings and target language strings in training corpus, and outputting the source language compact shared syntax forest of the training corpus and the target language compact shared syntax forest of the training corpus; and 2), extracting translation rules according to the source language compact shared syntax forest of the training corpus and the target language compact shared syntax forest of the training corpus. Since the testing corpus is translated according to the translation rules, the translation quality is effectively improved.

Description

technical field [0001] The present invention relates to the field of natural language processing, and more specifically, to a language translation rule extraction method and a translation method. Background technique [0002] The syntax-based translation model has gradually become the mainstream model of statistical machine translation. According to whether the syntax tree is built on the source language side and the target language side, the syntax-based translation model can be divided into the following three categories: 1) tree-to-string translation model, such as Liu Yang (Yang Liu) et al. disclosed at the 44th Annual Conference of Computational Linguistics (ACL 2006) in 2006; 2) the string-to-tree translation model, Ma Ku et al. 3) a tree-to-tree translation model, such as that disclosed by Eisler et al. at the 41st Annual Conference on Computational Linguistics in 2003. [0003] The existing tree-to-tree translation model establishes a unique syntactic tree on the so...

AI Technical Summary

Problems solved by technology

[0005] Therefore, the translation quality of existing tree-to-tree based translation models is not ideal

Images