Machine translation system based on human-computer interaction and automatic optimization method thereof

Abstract

The invention provides a machine translation system based on human-computer interaction and an automatic optimization method thereof, and the method comprises the steps: completing the directional enhancement of a parallel corpus in a user application field through manual translation corpus by means of a loopback translation technology; generating monolingual corpora in the user application field based on user terms and target language data of manual translation corpora through a text generation technology, and translating the monolingual corpora in the user application field back to generate parallel corpora biased to the user application field; fusing parallel corpora in all user application fields on the basis of the native parallel corpora for automatic online training of a machine translation model. Automatic optimization of a machine translation model is completed online by utilizing high-quality parallel corpora accumulated by a man-machine translation system, so that links of corpus management and construction, machine translation, machine-aided translation and online training form a closed loop, and the translation capability of a man-machine translation platform on data in a specific user application field is continuously improved.

Description

technical field [0001] The invention relates to the technical field of natural language processing, in particular to a machine translation system based on human-computer interaction and an automatic optimization method thereof. Background technique [0002] With the rapid development of deep learning technology, machine translation based on deep neural network has achieved a qualitative leap in quality. At the same time, more and more companies have begun to launch human-computer interaction machine translation platforms (hereinafter referred to as human-machine translation platforms) to realize the integration between machine translation and machine translation translation (CAT). Human translators use the human-computer translation platform to post-edit the machine-translated text results, and then add the post-edited sentence pairs as translation resources to the memory database, and serve as recommendations for the next translation of similar texts. At the same time, huma...

AI Technical Summary

Problems solved by technology

If a certain scale of high-quality parallel corpus of manual translation can be incorporated into the training corpus, the translation ability of the model in the user application field can be effectively improved. However, the scale of manual translation corpus is limited, so simply add this part of the corpus to the training set for model training play a limited role

[0003] In addition, the term translation function currently provided by most platforms is basically implemented by first translating and then replacing. Although this method can ensure that the professional term is translated correctly, it loses the context information of the term during the translation process, and cannot guarantee the final translation. The sentence of is the optimal result

Images