Text abstract generation method, intelligent terminal and storage medium

Abstract

The invention discloses a text abstract generation method, an intelligent terminal and a storage medium, and the method comprises the steps: obtaining the input information of an abstract to be generated, and converting the input information into digital vector data; Inputting the digital vector data into a bidirectional recurrent neural network for operation through an encoder, and outputting sentences to a selection gate network to generate a new gate state sequence; Adding a Gaussian window to the gate state sequence output in the selection gate network through the attention layer, performing calculating to obtain a vector containing context information at the current moment, and outputting the vector to a decoder; And generating an output result with the highest prediction probabilitycorresponding to a real training result by decoding through a decoder, and taking the output result as the abstract of the input information. The text abstract is rapidly and accurately generated according to the requirements of the user, the purpose of accurately expressing the article can be achieved, the reading habit of the user is better met, and the user can rapidly obtain the main information of the article according to the text abstract.

Description

technical field [0001] The present invention relates to the field of computer application technology, in particular to a method for generating a text summary, an intelligent terminal and a storage medium. Background technique [0002] With the rapid development of science and technology, all kinds of articles and materials on the Internet are dazzling. These messy articles are often of little use value, and many attractive titles are not true, which makes it easy to be "headline party". Deceit, especially after reading some materials and then discovering that they are not what you want, not only wastes a lot of time, but also greatly reduces work efficiency. In order to determine whether the materials you currently have are what you really want , to find the content you are satisfied with more quickly, you need to quickly browse and identify unknown information, so as to get the valuable information you really need. [0003] The existing relatively mature abstract generatio...

AI Technical Summary

Problems solved by technology

[0002] With the rapid development of science and technology, all kinds of articles and materials on the Internet are dazzling. These messy articles are often of little use value, and many attractive titles are not true, which makes it easy to be "headline party". Deceit, especially after reading some materials and then discovering that they are not what you want, not only wastes a lot of time, but also greatly reduces work efficiency. In order to determine whether the materials you currently have are what you really want , to find the content you are satisfied with more quickly, you need to quickly browse and screen the unknown information, so as to get the valuable information you really need

[0003] The existing relatively mature abstract generation methods generally adopt the extraction method. The core idea is to find some keywords and sentences in the original text to form an abstract. This method has great limitations. Keywords, The generation of sentences is generally determined by searching for tags with high frequency after word segmentation of the article. It is difficult for the generated abstract to accurately capture all the information of the article.

As a result, the summary effect is unsatisfactory and cannot meet the needs of users

[0004] That is to say, the core idea of ​​the traditional method is to extract key sentences or words with high frequency in the original text, and splicing these words or sentences together to generate a summary. Since there is no correlation between the keywords before and after, the generated summary is often The words do not express the meaning, are obscure and difficult to understand, and cannot clearly express the core idea of ​​the original text, requiring users to "half-read and half-guess", which cannot achieve the real purpose of the abstract

Images