Text translation model training method and text translation method and device

Abstract

The embodiment of the invention provides a training method of a text translation model and a text translation method and device, and the training method comprises the steps: obtaining a parallel corpus set, and enabling each parallel corpus in the parallel corpus set to comprise a source language text, a machine translation text, a post-edited translation text and a revised translation text; sequentially training a first translation model, a second translation model and a third translation model according to the parallel corpus set, taking the trained third translation model as the text translation model, enabling the first translation model, the second translation model and the third translation model to share part of model parameters, and enabling the model parameters of the previous translation model after the training not to change along with the training of the subsequent translation model. According to the training method provided by the embodiment of the invention, a better model generalization effect can be obtained, the generalization ability of the model is improved, and finally, the translation result obtained by training is more accurate.

Description

technical field [0001] The present invention relates to the technical field of machine translation, and more specifically, to a training method for a text translation model. Background technique [0002] In the post-editing mode, the auxiliary translation tool calls the translation model results to give a first draft of translation, and then professional translators review and edit on this basis to produce high-quality translation results. In addition, in order to ensure a better translation effect, professional translation companies often invite revisers with higher translation qualifications to do more fine-tuning on the human-translated translation. Therefore, in this whole translation workflow, the generated corpus includes: the original text to be translated, the corresponding machine translation translation, the translation edited by the translator, and the translation "polished" by the reviewer (the final translation result) These four data. [0003] In many cases, ...

AI Technical Summary

Problems solved by technology

This way of outputting language is very different from the way humans write texts. When people write texts, they often write a draft first, and then polish the text based on this draft (similar to how a proofreading teacher works as a translator Polished based on the translations of the text), so there is a gap between machine-translated translations and human-written translations

Images