A Method for Improving the Quality of English-Chinese Machine Translation Based on Data Selection

Abstract

The invention discloses a method for improving English-Chinese machine translation quality on the basis of data selection. The method includes presenting data by the aid of bag-of-word presentation forms again; presenting the distances among sentences by the aid of cosine computation processes and acquiring final scores of each sentence pair by means of cosine correlation computation; sorting universal data by the aid of the scores, ultimately selecting the related data and systematically training machine translation systems. The method has the advantages that the time costs and the storage space costs in statistic machine translation system training procedures can be reduced, and the data volumes of the training data can be reduced as compared with systems trained by the aid of multi-field universal data; the selected data and to-be-tested data come from the same field and are related to one another relatively in the aspect of contents, and accordingly the performance of the systems trained by the data selected by the aid of the method can be theoretically superior to the performance of the machine translation systems trained by all the data.

Description

technical field [0001] The invention belongs to the technical field of data selection, in particular to a method for improving the quality of English-Chinese machine translation based on data selection. Background technique [0002] With the introduction of the IBM statistical model, statistical-based machine translation methods have gradually replaced rule-based translation methods and become the mainstream machine translation methods at this stage. The basic idea is to use statistical methods to automatically learn translation knowledge from large-scale bilingual corpora and build translation models. [0003] In traditional statistical machine translation, the quality of the corpus directly determines the quality of the final translation system. In this era of information explosion, the information on the Internet is growing exponentially, and it also provides a large amount of monolingual or bilingual corpus for machine translation. [0004] Theoretically, as the amount...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to provide a method for improving the quality of English-Chinese machine translation based on data selection, aiming to solve the problem that most machine translation systems at the present stage require a large amount of training time during the entire training process, and also require huge disk space to store data and models, while at the same time, the translation system trained with a large amount of multi-domain training data cannot achieve the best translation results for a specific domain

Images