A cross-domain emotion classification method and system based on feature representation learning

Abstract

The present invention relates to a cross-domain sentiment classification method and system based on feature representation learning, comprising the steps of: performing characterization processing on source domain text and target domain text to obtain an initial text vector in the source domain and an initial text vector in the target domain; The initial text vector and the target domain initial text vector are respectively in the feature representation learning module, and the source domain text feature vector and the target domain text feature vector are obtained; the source domain text feature vector and the target domain text feature vector are sent to the feature representation strengthening module, and Carry out training; use the trained feature representation enhancement module to classify and predict the text feature vector of the target field. The invention can improve the effect of text sentiment classification in the target field.

Description

A cross-domain sentiment classification method and system based on feature representation learning technical field The present invention relates to the technical field of data processing, particularly a kind of cross-domain sentiment classification based on feature representation learning method and system. Background technique At present, there are many deep learning-based methods that can be used for text sentiment classification, and have achieved certain success. However, these methods are highly dependent on manually labeled data, especially the training corpus and the test corpus need to have the same characteristics distributed. For some emerging fields, there are scarce annotated training corpora, and manual annotation of corpus requires a lot of time and energy. Therefore, cross-domain text sentiment analysis methods are born. In domains with rich corpora through transfer learning Learn knowledge and transfer the learned knowledge to new fields, thereby...

AI Technical Summary

Problems solved by technology

However, this method lacks interpretability and cannot fully prove whether the network has fully learned the text features of domain adaptation, and there is still a lot of room for exploration.

Images