A text emotion analysis method based on dual-channel model

Abstract

The invention provides a text emotion analysis method based on a dual-channel model aiming at the problem that the single-channel neural network model has a single structure and can not fully extractthe text information. Includes such steps as training word vector by Word2Vec, representing text as word vector matrix, Then, using and sending the word vector matrix as input data to convolution neural network (CNN) and long-short-term memory (LSTM) network for feature extraction. After that, attention model is introduced to extract the important feature information of the text. Finally, the textfeatures extracted from the two channels are merged, and emotion classification is carried out by using the classification layer. The method provided by the invention has feasibility and superiority,and the performance of the method is obviously superior to other single-channel neural network models.

Description

technical field [0001] The invention proposes a text sentiment analysis method based on a dual-channel model, which relates to the field of text sentiment analysis. Background technique [0002] In recent years, with the rapid development of the Internet industry, many new media have emerged, and these new media are constantly impacting and changing people's way of life. The rise of various e-commerce platforms has made online shopping easy and popular without leaving home. The only feedback this shopping method gets is the consumer experience reviews left by consumers. These real reviews determine potential consumption. The only basis for consumers to consume. Therefore, performing sentiment analysis on these large amounts of emotional texts is beneficial to both e-commerce platforms and consumer groups. [0003] The main task of text sentiment analysis is to analyze text information with emotional color, extract features, and make polarity judgments. At present, there a...

AI Technical Summary

Problems solved by technology

However, with the development of society, the effect of this method is not ideal in the face of increasingly diverse texts

In addition, the use of machine learning methods for text sentiment analysis requires a large number of manually designed data features. With the increase of text data sets to be processed, traditional machine learning methods have been unable to learn the deep information features of text quickly and well.

[0004] When using traditional methods to extract text information features, the text is extracted without distinction, but it is worth noting that each word in the text contributes differently to the emotional polarity of the entire text. The commonly used neural network , such as convolutional neural network (Convolutional Neural Network, CNN) and long short-term memory network (LongShort-Term Memory, LSTM), when extracting features, it is impossible to distinguish the position of important information in the sentence, resulting in text classification results being greatly affected by irrelevant information. Big

The existing sentiment analysis methods basically extract text features by constructing a single-channel neural network model, and the performance of the single-channel model will be affected as the number of network layers increases, and cannot fully extract text features.

Images