Downlink generation method and device based on double-layer attention joint learning

Abstract

The invention discloses a method and a device for generating a lower copy based on double-layer attention joint learning, which adopt a novel'corpus-phrase-word 'three-layer text structure and use a double-layer attention joint learning model to generate the lower copy on the basis of the novel'corpus-phrase-word' three-layer text structure. A double-layer attention mechanism is introduced into the model to respectively capture attention information of a phrase layer and a word layer, and a joint learning method is used externally to carry out iterative training on all clauses of couplets, so that model parameters are optimized. According to the method, rich experiments are carried out, the BLEU value of the double-layer attention joint learning model is 0.374, and the score is obviously superior to that of an existing model using word embedding. Meanwhile, the effectiveness of the method is also verified in terms of indexes such as word number consistency rate and manual evaluation.

Description

technical field [0001] The invention belongs to the technical field of computer data processing, and in particular relates to a method and a device for generating downlinks based on two-layer attention joint learning. Background technique [0002] Couplet is a unique and long-standing traditional Chinese art. The art form of couplet is rigorous, requiring the same number of characters in the upper and lower couplets, relative semantics, and harmonious intonation. With the development of deep learning technologies in the field of natural language, more attention has been paid to the text generation task for Chinese. Among them, the automatic generation of couplets is a very innovative research. Different from tasks such as dialogue systems, machine translation, and poetry generation, the task of generating downlinks not only requires effective output, but also needs to meet the requirements of word count, semantics, and intonation between the output downlinks and the input up...

AI Technical Summary

Problems solved by technology

The model that uses "word" embedding will output word by word in the generation of couplets, and it is difficult to preserve the charm of the original word-to-word generation of couplets in the way of generating couplets from word to word, which makes the current method of generating couplets unable to meet the needs of couplets well.

Images