Mixed attention mechanism text title matching method based on multi-task learning

Abstract

The invention relates to a mixed attention strategy text title matching method based on multi-task learning, and the multi-task learning of a model is embodied in that the model carries out a classification task 1 of an original text category and a classification task 2 of whether the text is a'title type 'article on an input text at the same time. The model is jointly trained through the multi-task learning model, and one task assists the other task in learning better parameters. According to the scheme, model parameters are adjusted by using back propagation of the classification task 1, so that the classification task 2 obtains better performance. According to the invention, key information is extracted from the text and is matched with the title, so that detection of the title type article is realized, and the detection precision and accuracy of the title type are obviously improved. According to the attention mechanism provided by the method, the correlation degree between each element and other elements can be calculated in one step, the calculation amount is small, and the efficiency is high.

Description

technical field [0001] The invention relates to the field of text processing, in particular to a multi-task learning-based mixed attention mechanism text title matching method. Background technique [0002] In the Internet age, based on the actual benefits brought by traffic hoarding and traffic incentives, the Internet text "headline party" has taken advantage of the trend, which has degraded the browsing experience of Internet users. And the platform where the "headline party" is active is experiencing the loss of users, which has an impact on the sustainable development of the platform. [0003] "Headline Party" refers to making eye-catching headlines on forums or media represented by the Internet to attract the audience's attention. The general term for purpose website editors, reporters, managers and netizens. The main behavior of "title party" is that the title of the post is seriously exaggerated, and the content of the post is usually completely irrelevant or has l...

AI Technical Summary

Problems solved by technology

Loose detection mechanism cannot effectively detect "headline party" text

Most of the "headline party" texts use legal words, and the number of words contained in the dictionary of sensitive words is limited. This method cannot be applied to this application scenario in practical applications.

Images