Extreme multi-label classification data enhancement method based on label and text block attention mechanism

Abstract

The invention discloses an extreme multi-label classification data enhancement method based on a label and text block attention mechanism. The method comprises the following steps: selecting an original data set; the high-level semantic representation of each word in the text is learned through BERT; the method comprises the following steps: segmenting a text into a plurality of text blocks with equal length, and averaging the high-level semantic representation of each word in the text blocks to obtain the representation of the whole text blocks; and calculating the relevancy between the representation of each text block and the vector representation of the label through an attention mechanism, fusing the representations of all the text blocks, obtaining a complete label-text block relation model after training, performing data enhancement according to the relevancy, and finally outputting a new data set after enhancement. According to the method, the relationship between the label and the text block is considered, the relevance between the label and the text is learned by using the model, the unimportant text block in the original data is replaced based on the text block associated with the long-tail label, and the multi-label classification effect of various existing models on a new data set is remarkably improved.

Description

technical field [0001] The invention belongs to the technical field of computer applications, and in particular relates to data mining and extreme multi-label classification, in particular to an extreme multi-label classification data enhancement method based on a label and text block attention mechanism. Background technique [0002] In recent years, with the rapid development of the Internet, platforms such as social media and e-commerce websites have accumulated a large amount of labeled text data. Due to the large number of label sets, the extreme multi-label classification task is to find the most relevant labels of the text from a large number of label sets. Data mining through extreme multi-label classification tasks is of great significance to the development of various industries. For example, the analysis of commodity evaluation data in e-commerce websites can help merchants understand consumers' purchasing tendencies, and then provide them with effective decision ...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to solve the problem that the existing technology cannot obtain better classification results on labels with a small number of occurrences, and to provide an extreme multi-label classification data enhancement method based on the label and text block attention mechanism, which can enhance labels with a small number of occurrences related data, thereby improving the classification performance of various models on such labels

[0008] The present invention believes that increasing the number of "long tail" labels (labels with low frequency in the data set) in the data set through the data enhancement method and then improving the performance of extreme multi-label classification based on existing methods is an effective solution to the "long tail" Pathways with Poor Performance in Label Classification

Images