Emotional data classification method and system

Abstract

The present invention provides a method and system for classifying emotional data, the method comprising: constructing a document-document graph and a word-word graph corresponding to a training data set, in the document-document graph, a node represents a word in the training data set document, the geometric information of the edge represents the correlation between the documents, and in the word-word graph, the node represents the word in the training data set, and the geometric information of the edge represents the correlation between the words; according to the document-document The geometric information of the graph and word-word graph constructs a graph-based regularization item in the objective function; optimizes the objective function and outputs a document-sentiment matrix; obtains documents in the test data set, according to the document-sentiment matrix The sentiment tendency corresponding to the documents in the test data set is obtained. By adopting the method and system, the accuracy of emotion classification can be improved.

Description

technical field [0001] The invention relates to natural language processing technology, in particular to an emotion data classification method and system. Background technique [0002] With the development of Web 2.0, more and more users generate emotional data in web pages, which usually exist in the form of comments and blog data in the network. Sentiment classification refers to automatically predicting the sentiment orientation of sentiment data generated by users, for example, predicting whether a comment is positive or negative. [0003] Recently, sentiment classification has gained widespread attention in natural language processing, and sentiment classification methods can be divided into supervised sentiment analysis and unsupervised sentiment analysis. Supervised sentiment analysis relies on human-annotated training data, and in some cases, the labeling work is time-consuming and expensive, which motivates unsupervised or semi-supervised sentiment analysis. [00...

AI Technical Summary

Problems solved by technology

However, it is difficult to define a universally optimal emotional vocabulary to cover all words from different domains

Furthermore, most semi-automatic dictionary-based methods do not yield satisfactory results

The traditional and more advanced dictionary-based method is based on the constrained non-negative matrix tri-factorization (ConstrainedNon-negative Matrix Tri-factorization, referred to as CNMTF) sentiment classification method, which uses domain-independent emotional vocabulary as prior knowledge for emotional classification. However, the experiments show that the sentiment classification accuracy of CNMTF-based sentiment classification method still needs to be improved

Images