A Sentiment Classification Method

Abstract

The present invention provides an emotion classification method, comprising: obtaining a word embedding matrix corresponding to a context and a word embedding matrix corresponding to a target word; according to the word embedding matrix corresponding to a context, a word embedding matrix corresponding to a target word, and the first semantic activation model, obtaining The context representation of target word semantic enhancement and the target word representation of context semantic enhancement; according to the context representation of target word semantic enhancement, the target word representation of context semantic enhancement and the semantic selection model, the context representation after semantic selection is obtained; according to the semantic integration model, Extract the syntactic representation in the syntactic dependency tree corresponding to the target sentence; obtain the sentiment classification result corresponding to the target word according to the context representation, syntactic representation and the second semantic activation model after semantic selection. Compared with the prior art, the present invention fully captures the semantic information related to the target word in the context, and comprehensively considers the relationship between the context, the target word and syntax, thereby improving the accuracy of emotion classification.

Description

technical field [0001] The invention relates to the technical field of natural language processing, in particular to an emotion classification method. Background technique [0002] Since the comments left by users on forums or e-commerce platforms are of great significance to merchants in analyzing user opinions, sentiment analysis has received more and more attention. Sentiment analysis is an important task in Natural Language Processing (NLP), and its purpose is to analyze subjective text with emotional color. [0003] At present, there are many methods for classifying the sentiment polarity of a sentence or a document as a whole. However, there are usually different target words in a sentence or a document, and the sentiment polarity of the target words may be different. If the overall emotional polarity classification is carried out directly, it will lead to errors in the judgment of the emotional polarity of the target word. Therefore, attribute-level sentiment classi...

AI Technical Summary

Problems solved by technology

[0004] In the prior art, there are some methods that combine neural networks and attention mechanisms for attribute-level sentiment classification. Although these methods can overcome the defects of shallow learning models, they still have the following problems: On the one hand, they cannot fully Capturing the semantic information related to the target word in the context, it is easy to cause misjudgment when the semantic relationship is far away or the word order changes; on the other hand, it does not consider the context and the relationship between the target word and syntax, and due to the lack of syntax constraints technology, which may identify syntactically irrelevant context words as clues to judge the sentiment classification of the target word, reducing the accuracy of sentiment classification

Images