Translation method, machine translation model training method and device and storage medium

Abstract

The invention discloses a translation method, a machine translation model training method and device and a storage medium, and relates to the field of artificial intelligence. The method comprises: obtaining a sample statement marked with a reference translation result; inputting the sample statement into a to-be-trained machine translation model to obtain a sample translation result of the samplestatement, the machine translation model being a non-autoregressive model with to-be-adjusted model parameters; calculating a gradient estimation value of the word translation result according to thereference translation result, and obtaining the gradient estimation value through calculation of a reward value; and adjusting model parameters of the machine translation model according to a gradient estimation result. According to the method, the gradient estimation value is calculated through the reward value, the model parameters of the machine translation model are adjusted according to thegradient estimation value, the gradient estimation value is obtained with the reward value as the adjustment reference data to adjust the model parameters, and therefore sequence-level training of themachine translation model is achieved, and the translation accuracy of the trained machine translation model is improved.

Description

technical field [0001] The embodiments of the present application relate to the field of artificial intelligence, and in particular to a translation method, a training method of a machine translation model, a device, and a storage medium. Background technique [0002] Machine translation is a method of translating sentences in one natural language into sentences in another natural language through a computer. Usually, the machine translation is to translate sentences through a trained machine learning model. Schematically, a machine learning model After training with a large number of translation corpus samples, the user inputs the Chinese sentence "The housing prices continue to rise" into the machine learning model, and the output is the English translation "The housing prices continued to rise.". [0003] In related technologies, a non-autoregressive model is used to implement a machine translation process with high speed requirements. The model parameters of the non-auto...

AI Technical Summary

Problems solved by technology

[0005] The embodiment of the present application provides a translation method, a machine translation model training method, a device, and a storage medium, which can solve the problem that non-autoregressive models are difficult to capture the sequence dependencies between translations at different positions, and the generated translations are usually low in fluency. less effective problem

Images