Case-related news viewpoint sentence sentiment classification method based on CNN-BiLSTM + action model

Abstract

The invention discloses a case-related news viewpoint sentence sentiment classification method based on a CNN-BiLSTM + action model. The model comprises a word embedding layer, a convolution layer, apooling layer, a BiLSTM layer, an Attention attention layer and a Softmax classification layer. The method comprises the following steps: step 1, preprocessing viewpoint sentences of case-related news, and then encoding all words into word vectors through a word embedding layer; 2, inputting the word vector obtained in the step 1 into a convolution layer, and carrying out convolution operation; 3,the feature vectors obtained after convolution are input into a pooling layer to be subjected to maximum pooling operation; 4, inputting the feature vector obtained after the maximum pooling into a BiLSTM layer; step 5, performing Attention operation on each hidden state vector obtained by the BiLSTM layer; and step 6, classifying the output obtained by the Attention operation through a softmax classification layer to obtain a target emotion category probability. According to the method, complex feature engineering and redundant manual work can be effectively solved, and the workload and complexity can be reduced to a great extent.

Description

Technical field [0001] The present invention relates to the technical fields of natural language processing and deep learning, in particular to a method for categorizing the sentiment of a news opinion sentence involved in a case based on a CNN-BiLSTM+attention model. Background technique [0002] The sentiment classification task of the news opinion sentence involved in the case can be regarded as sentence-level sentiment classification. In the past, the sentiment classification methods were mainly based on dictionary and based on machine learning. Dictionary-based methods generally have an sentiment dictionary. The sentimental words in the dictionary are matched with the words in the sentence, and then rules are designed to analyze the sentiment of the entire sentence; the method based on machine learning usually uses a classification model to perform the sentence Sentiment classification, first extract text features from the data set, and then train a machine learning classifi...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a kind of sentiment classification method based on CNN-BiLSTM+attention model involved in news opinion sentences, to solve the problem of complex feature engineering and artificial engineering based on rules and statistical machine learning methods

Images