Semi-supervised adversarial learning cross-language abstract generation method based on word alignment

Abstract

The invention relates to a semi-supervised adversarial learning cross-language abstract generation method based on word alignment, and belongs to the technical field of natural language processing. The method comprises the following steps: collecting news texts generated for training a Chinese-Vietnamese cross-language abstract , and obtaining existing Chinese-Vietnamese bilingual word vectors; pre-training a monolingual abstract model and semi-supervised adversarial learning by utilizing the Han-Vietnamese news text and the Han-Vietnamese bilingual word vectors respectively; utilizing a Bertencoder to respectively carry out vector representation on input Chinese-Vietnamese pseudo-parallel corpora; carrying out semi-supervised adversarial learning by combining the vectors obtained by theencoder with the Chinese-Vietnamese bilingual seed dictionary to obtain vectors mapped to the same semantic space; and taking the context text vector and the reference abstract which are mapped in thesame semantic space as input of a transformer decoder, and carrying out decoding to output a target language abstract. According to the method, the cross-language abstract generation task is realized, and the cross-language abstract effect is optimized.

Description

Technical field [0001] The present invention relates to the term-based semi-supervision confrontation learning cross-language abstract generation method, which belongs to the technical field of natural language processing. Background technique [0002] Cross-language summary generation is a hot problem of current natural language processing research. The problem of common concern in China and Vietnam, the relevant news reports also increased, using cross-language abstract approach to obtain the text summary information of Vietnamese news, to the timely understanding of the two countries, to promote the common development of the two countries It is of great significance. Currently, translation technology for small languages ​​is not yet mature, and different language text is difficult to represent the abstracts of cross-language news text under the same feature space. Therefore, it is of great significance to use artificial intelligent technology to automatically generate the abst...

AI Technical Summary

Problems solved by technology

[0003] The present invention provides a semi-supervised adversarial learning method for generating cross-language summaries based on word alignment, which is used to solve the problems that texts in the same language are difficult to represent in the same feature space, and how to use text representations in the same space to perform cross-language summarization tasks. And solved the problem that translation must be used to achieve cross-language summarization, but the translation effect is not good

Images