A multi-label text classification method and device based on multi-task learning

Abstract

The present application provides a multi-task learning-based multi-label text classification method and device, the method comprising: obtaining training samples, and establishing a multi-task learning model for multi-label text classification; wherein, the multi-task learning model includes sharing layer and many specific task layers; the shared layer is used to obtain the context-related information in the text information; the multi-specific task layer is used to carry out multi-task classification for the feature vector output by the shared layer; when obtaining the text information to be classified, The classification label of the text information to be classified is obtained based on the multi-task learning model. This method enables more accurate multi-label text classification.

Description

technical field [0001] The invention relates to the technical field of information processing, in particular to a multi-label text classification method and device based on multi-task learning. Background technique [0002] With the development of the Internet and social media, there is already a large amount of text information on the Internet, including Wikipedia entries, academic articles, news reports, and various after-sales service reviews, and these text information contains a lot of valuable information , the existing text classification technology can roughly extract specific information. For example, consumers’ satisfaction with the product or service can be known by sentiment analysis of after-sales reviews. By classifying news data, news reports can be roughly distinguished. Domains, relationships in knowledge graphs, etc. can be obtained by classifying sentences in Wikipedia data. [0003] In a word, text classification is an extremely important technology. At ...

AI Technical Summary

Problems solved by technology

[0004] The above-mentioned technologies are all applicable to single-label text classification, and it is impossible to predict the various labels included in the text. For multi-label text classification, the current mainstream method is to train m classifiers, and then the final result of these m classifiers The advantage of this method is that multi-label can be obtained without changing the algorithm, but its disadvantage is that the m classifiers are isolated from each other. Applying this network structure to related Multi-label text cannot give accurate labels

Images