A multi-label text classification method based on seq2seq

Abstract

The invention discloses a seq2seq-based multi-label text classification method in the field of tag text classification, comprising steps: S1: preprocessing the training corpus; S2: establishing a seq2seq-based multi-label text classification model, and training the parameters of the model; S3: Use the trained multi-label text classification model to perform text classification on the data to be predicted. The present invention does not need to manually extract features, and can use CNN to extract local semantic information of the text, which can improve the accuracy of text classification, and use the initialized fully connected layer, which can consider the correlation between labels, thereby improving the accuracy of text classification .

Description

technical field [0001] The invention relates to the field of label text classification, in particular to a multi-label text classification method based on seq2seq. Background technique [0002] Traditional text classification techniques mainly focus on single-label classification, that is, a text corresponds to only one category label. However, multi-label text classification is more common and more difficult in real life than single-label classification. Because in multi-label classification, the number of label subsets grows exponentially with the number of labels. Assuming that the number of labels in a multi-label classification problem is K, the theoretical total number of label subsets is 2^k-1, so from this exponential number of label subsets, there is no doubt that the correct label subset is selected is a huge challenge. To address this challenge, it is often necessary to exploit the correlation between labels to facilitate the learning process. For example: if ...

AI Technical Summary

Problems solved by technology

Using neural networks to process text classification makes up for many shortcomings of traditional machine learning methods, such as: no need to manually design features, etc., but the neural network methods proposed in recent years to process multi-label text classification rarely consider the correlation between labels and ignore the importance of local semantic features, which greatly affects the accuracy of classification

[0006] To sum up, the multi-label text classification method based on traditional machine learning requires manual design of features, which is very time-consuming and labor-intensive, and the quality of features has a great impact on the classification effect

At the same time, it is difficult for such methods to effectively consider the correlation between labels

However, the existing multi-label text classification method based on deep learning can automatically extract effective features, but it cannot effectively consider the correlation between labels and ignore the importance of local features.

Images