Research On Fine-grained Text Sentiment Classification For Social Internet Of Things

Text sentiment classification is important to improve the ability of autonomous decision-making and communication among object peers in Social Internet of Things.Fine-grained text sentiment classification,that is,aspect-level sentiment classification,has been a research hotspot in the field of natural language processing in recent years.However,the existing solutions often fail to fully capture and efficiently use the information between sentences and their respective aspect terms,and there are difficulties in realizing the information sharing between sentences and their respective aspect terms.Therefore,the classification results of the models are not ideal.This paper combines the neural network and attention mechanism to conduct research on fine-grained text sentiment classification methods.Through model design and structure optimization,two fine-grained text sentiment classification models are proposed for the Social Internet of Things to automatically extract the internal meanings that users want to express for their networks.Deploy or enhance services to make informed decisions,thereby enhancing the connectivity and user-friendliness of the Internet,and ultimately achieving the goal of improving autonomous decision-making and communication capabilities between objects,while providing valuable information to humans or smart devices,and services make automated decisions.The main research content and innovation of this paper are as follows:(1)The basic principles of text sentiment classification model based on long short-term memory networks are briefly analyzed.In this paper,eight typical long short-term memory network-based text sentiment classification models are briefly analyzed,including their model structures and workflows,and their shortcomings are summarized,while effective solutions are proposed for the shortcomings.These eight models are used as the effect comparison model of the subsequent proposed models.(2)A fine-grained text sentiment classification model based on mutual attention mechanism(MAN)is proposed.Aiming at the problem that the existing text sentiment classification models cannot fully obtain and efficiently use the information between sentences and their respective aspect terms,oriented to the field of Social Internet of Things,based on bi-directional long short-term memory network and mutual attention mechanism,the MAN model is proposed in this paper.The experimental results on the three datasets of REST,LAPTOP and TWITTER show that the MAN model can effectively obtain the mutual attention information between sentences and their respective aspect terms,and efficiently use the information to serve the fine-grained text sentiment classification task,thus improving the accuracy of fine-grained text sentiment classification in Social Internet of things.(3)A fine-grained text sentiment classification model based on collective attention mechanism(CAN)is proposed.Aiming at the problem that the existing text sentiment classification models,which cannot achieve information sharing between sentences and their respective aspect terms,oriented to the field of Social Internet of Things,based on BERT pre-trained model and collective attention mechanism,the CAN model is proposed in this paper.The experimental results on the three datasets of REST,LAPTOP and TWITTER show that the CAN model can effectively obtain the collective attention information between the sentences and their respective aspect terms,and realize the information sharing between them,so as to efficiently serve the fine-grained text sentiment classification task.This model can solve the problem of fine-grained text sentiment classification in Social Internet of things with relatively high accuracy.