Object-level sentiment classification method based on segmented convolutional neural network

Abstract

The invention discloses an object-level sentiment classification method based on a segmented convolutional neural network, and the method comprises the steps: carrying out the feature extraction of ato-be-classified text, and carrying out the context feature division of the to-be-classified text, and obtaining a first text-above feature and a first text-below feature; inputting the data into a preset segmented convolutional neural network; carrying out the convolution operation on the convolution layer to obtain local features of the preceding text and the following text; enabling the segmented pooling layer to perform maximum pooling operation on the local features of the preceding text and the following text to obtain the pooling features of the preceding text and the following text; enabling the feature fusion layer to perform feature fusion on the pooling features of the preceding text and the following text and the attention feature representation extracted by the attention module; enabling a softmax layer to process the fused features; and outputting sentiment classification results. According to the invention, the technical problems that the context feature with finer granularity cannot be obtained, and the emotion classification result accuracy is not high due to the facts that the existing emotion classification method does not consider the influence of the context divided by the object on the classification effect, and the maximum pooling operation is generally adopted in the pooling layer of the convolutional neural network.

Description

technical field [0001] The present application relates to the technical field of emotion classification, in particular to an object-level emotion classification method based on a segmented convolutional neural network. Background technique [0002] Sentiment classification aims to mine the categories of emotional tendencies expressed by people in texts. It mainly focuses on the research of sentiment classification tasks at the text level, sentence level and object level. Among them, object-level sentiment classification is a relative comparison. Fine-grained classification problem, which aims to judge the sentiment category of a specific object in a sentence, for example, "This laptop is very powerful, but the price is too high!" This task requires judgment on the object "functionality" Identify the positive emotional category, and judge the negative emotional category for the object "price". [0003] In the prior art, the convolutional neural network is used for sentiment ...

AI Technical Summary

Problems solved by technology

[0004] This application provides an object-level emotion classification method based on a segmented convolutional neural network, which is used to solve the problem that the existing emotion classification method does not take into account the influence of the context divided by the object on the classification effect. In the convolutional neural network The pooling layer usually adopts the maximum pooling operation, resulting in the inability to obtain finer-grained context features, which makes the accuracy of sentiment classification results not high.

Images