Abstract text title generation method based on eye movement attention

Abstract

A text title generation method based on eye movement attention comprises the steps that S100, word embedding operation is conducted on a character sequence of a source text, and then the character sequence is input into a bidirectional encoder; S200, the bidirectional encoder extracts context semantic vectors from the embedded representation of the source text; S300, calculating eye movement attention weights of the words and the sentences; S400, the context semantic vector is taken as the initial state of the first layer of the decoder, and firstly the hidden layer state matching degree of the decoder and the bidirectional encoder are calculated; Further an attention context weight and an eye movement attention context weight vector of the word is calculated by utilizing the obtained matching degree; and S500, the eye movement attention context weight vector, the hidden state of the first layer of the decoder at the current moment and the hidden state of the second layer of the decoder at the previous moment are input into the second layer of the decoder to generate a corresponding title. According to the method, the accuracy and the effectiveness of the generated title are improved, so that the generated title is closer to an artificial title.

Description

technical field [0001] The disclosure belongs to the field of natural language processing and automatic text summarization, in particular to a method for generating abstract text titles based on eye movement attention. Background technique [0002] In recent years, with the popularity of online platforms, users browse a large amount of text information every day. However, if the text is too long, users will find it tedious to read. In view of the fact that the text title can provide users with the most basic information of the text, and at the same time save the time for users to browse the text. Therefore, people hope to solve the problem of information overload by generating high-quality titles to mine the key information of the text. Unfortunately, manual generation of text titles requires considerable manpower, time and other resources. Not only that, the rampant "title party" even prevents users from knowing the real information of the text through the title. Theref...

AI Technical Summary

Problems solved by technology

Therefore, the generative method is more in line with the process of artificially generating titles, but it cannot capture the key information of the text well.

At the same time, most of the above methods use linguistic knowledge such as grammar and context to grasp the important information of the text, lack of consideration for the user's attention on the text, and ignore basic human behavior data such as eye movement data, resulting in generated titles limited expressiveness

Images