A text sentiment classification method and system

Abstract

The present invention provides a text emotion classification method, comprising: S1, based on the preset weight matrix set in the restricted recursive neural tensor network model, extracting words whose weights are greater than the preset threshold in the text as semantic content words; S2, based on training The final restricted recurrent neural tensor network model extracts the emotional features of the semantic content words; S3, based on the emotional features of the semantic content words, performs emotional classification on the text. The text emotion classification method and system provided by the present invention, by adding a weight matrix set on the basis of the recursive neural tensor network model, reduces the weight of function words in model training, so that text emotion feature detection can focus more on content words and reduce information redundancy Interference, improve the accuracy of text sentiment classification.

Description

technical field [0001] The present invention relates to the field of text information processing, and more specifically, to a text emotion classification method and system. Background technique [0002] According to the survey results of the "Statistical Report on Internet Development in China", as of December 2016, the number of Internet users in China reached 731 million, equivalent to the total population of Europe, and the Internet penetration rate reached 53.2%. The Internet has made remarkable achievements in the fields of the overall environment, popularization of applications, and development of hot industries. With the development and popularization of network technology, the network not only provides netizens with a new way of information dissemination, but also provides users with fast and convenient interaction methods, serving as a bridge of communication between readers and authors, and between readers. The increasingly prominent role of the Internet in the fi...

AI Technical Summary

Problems solved by technology

[0004] However, although the RNTN model used in the prior art double-marks words by constructing word vectors and word matrices, there will be a lot of information redundancy in this process. For example, some articles such as: the, a, etc. will also are included in the semantic information with the same importance, but when performing text classification, the emotional features of semantic content words contained in the text are the discriminative features required in the classification process, so when using this model for text sentiment classification, Classification accuracy will be disturbed by redundancy, resulting in low classification accuracy

Images