Research On Sentiment Classification Based On BiGRU And Aspect Attention Module

Aspect level sentiment classification is a fine-grained task in emotional analysis,which can provide more complete and deeper results for a given aspect.The emotion classification at the aspect level is in the development stage.The neural network model based on deep learning has been widely used in the field of emotion analysis.The deep learning algorithm based on attention mechanism has been proved to be able to deal with the emotion classification tasks at the aspect level because it can focus on the specific aspects of the review text.This paper proposes a variety of emotional classification models for research.The research contents are as follows:Firstly,after analyzing and researching the relevant algorithms of the current aspect level emotion classification task,it is found that the current neural network model has many problems such as too long training time and insufficient text information extraction when dealing with the aspect level emotion classification.Therefore,an aspect level emotion classification model based on bi directional gate recurrent unity(BiGRU)is proposed.BiGRU neural network has higher efficiency and performance,which can significantly improve the speed of model training.The experiment verifies the effectiveness of BiGRU in emotion classification task.Secondly,the current attention mechanism based model can only deal with one aspect at a time and use the same set of attention parameters for different aspects,so there is a fuzzy definition of different emotional tendencies.This paper proposes a hierarchical emotion classification model based on attention mechanism.The model uses different attention modules for different aspects.Each attention module has its own attention parameters,which can deal with different emotional tendencies in different aspects,and can deal with multiple aspects simultaneously by using the parallel computing ability of attention machine system,which can significantly optimize the model processing of aspect level emotion The performance of the sense classification task.Finally,considering the advantages of BiGRU and attention mechanism,the final model and the attention module model of bi-directional gate control cycle neural network are proposed.The experimental results show that,for the emotion classification task at the aspect level,BiGRU combines the aspect attention module model to obtain the optimal results in the accuracy evaluation index,and the iteration time required by the model is greatly reduced.