Method and device for generating text abstract

Abstract

One or more embodiments of the invention disclose a method and a device for generating a text abstract, which are used for simply, conveniently and quickly extracting the text abstract related to a personalized theme. The method comprises the steps of obtaining a plurality of sentences in a target text; according to a preset theme classification mode, predicting first compactness between each sentence and a specified theme; determining the similarity between the sentences; according to the similarity between the sentences, performing at least one time of correction on the first compactness byutilizing a specified algorithm to obtain a second compactness between each sentence and the specified theme; and processing each sentence according to the second compactness to obtain a text abstract, related to the specified topic, of the target text.

Description

technical field [0001] This description relates to the technical field of text processing, and in particular to a method and device for generating text summaries. Background technique [0002] In the field of abstract generation for long texts, there are two mainstream ways to automatically generate text summaries: one is to find one or several sentences closest to the central idea from the original text based on Textrank, and the generation method is to read the original text by a computer; The other is a generative method based on a deep neural network. This method often requires a large amount of corpus annotations, and the cost is high. [0003] In the summary generation scenario, in addition to the conventional summary summary (this type of summary can usually be generated using Textran), some scenarios also require the generation of qualitative topic-related summaries, such as the generation of personalized summaries for users, content auditing Risk summary generation...

AI Technical Summary

Problems solved by technology

In this case, since Textrank mainly extracts the central sentence of the overall content, it is impossible to extract abstracts related to the target topic.

Although the abstract generation method based on the deep neural network can solve such problems through personalized annotation, the annotation cost of the abstract class is high, and it is often not easy to implement.

It can be seen that none of the current abstract generation methods can be well utilized in the generation of qualitative topic-related abstracts.

Images