Translation model training method, text translation method, device and equipment

Abstract

The embodiment of the invention provides a translation model training method, a text translation method, a text translation device and text translation equipment, and relates to the technical field of machine learning. According to the specific implementation scheme, a training data set used for training a translation model is obtained; identifying each target training sample in the training data set, wherein the specified label of each target training sample accords with a predetermined condition; performing specified preprocessing on each target training sample to obtain each first training sample; translating the plurality of to-be-utilized samples by utilizing a to-be-trained translation model to obtain an output result of each to-be-utilized sample; determining a model loss value based on the output result of each to-be-utilized sample and the specified label of the to-be-utilized sample; based on the model loss value, whether the translation model converges is judged, if yes, training is ended, the trained translation model is obtained, and if not, the step of obtaining the training data set used for training the translation model is executed again. Therefore, through the scheme, the translation accuracy of the translation model can be improved.

Description

technical field [0001] The present invention relates to the technical field of machine learning, and in particular, to a translation model training method, text translation method, device and equipment. Background technique [0002] In the field of machine translation, a source sentence containing a person's name needs to be correctly translated into the target language. For example, if the source sentence is "Hong Yu just went to school" and the target language is English, the source sentence needs to be correctly translated as "Hong Yu has just gone to school". Yugang goto school", which translates to a translation that includes the correct person's name. [0003] In the related art, a large number of source sentences containing personal names are used as training samples, and the translation result belonging to the target language corresponding to each training sample is used as the label of the training sample, and the translation model pair is trained by using the train...

AI Technical Summary

Problems solved by technology

[0004] However, in related technologies, since there may be unaligned data between some training samples and labels, learning these training samples will cause the model to learn wrong translation results of personal names

Images