Preprocessing module of multi-language intelligent preprocessing real-time statistical machine translation system

Abstract

The invention discloses a preprocessing module of a multi-language intelligent preprocessing real-time statistical machine translation system. The preprocessing module comprises a text preprocessing module and an automatic speech recognition result preprocessing module, wherein the text preprocessing module is used for carrying out word normalized operation, class recognition labeling and chuck and word order adjustment on languages input in a text manner; and the automatic speech recognition result preprocessing module is used for carrying out word normalized operation and punctuation prediction on the languages. The preprocessing module disclosed by the invention is capable of carrying out basic operations such as word normalized operation, class recognition labeling and chuck and word order adjustment on to-be-translated text languages, so that convenience is brought to the translation carried out on the to-be-translated language texts by a subsequent translation module; or the preprocessing module disclosed by the invention is capable of carrying out work normalized operation on speech languages or carrying out preprocessing such as prediction and the like on the punctuations in speech flows, so that convenience is brought to the translation carried out by a subsequent machine translation module. The preprocessing module has the effect of carrying out labelling and preferential translation on small-probability words, so as to improve the correctness of translating the small-probability words.

Description

Technical field [0001] The invention relates to the technical field of artificial intelligence machine translation, and in particular to a preprocessing module of a multi-language intelligent preprocessing real-time statistical machine translation system. Background technique [0002] Machine translation is a technology that uses computers to automatically translate human natural languages. It is a process of using computers to convert one natural language into another natural language, and the two natural languages ​​should be equivalent in meaning. [0003] At present, a relatively mature and mainstream machine translation method is based on statistics. The advantage of this method is that there is almost no need to manually write translation rules. All translation information is automatically learned from the corpus, so this method is the biggest To a great extent, the characteristics of computer high-speed operation are brought into play, and labor costs are greatly reduced. [...

AI Technical Summary

Problems solved by technology

[0005] At present, the function of the preprocessing module of machine translation is not perfect. Most of them are trained and translated by the translation module after simple typo judgment and punctuation prediction after receiving by the receiving module. This not only increases the difficulty of machine translation, but also for small probability Words, the translation module may have inaccurate translation problems

Images