Research On Emotional Analysis And Visualization Of Online Comments On Agricultural Products Based On RoBERTa Model

In recent years,with the vigorous development of the national economy,China ’s emphasis on agricultural development is also increasing.The e-commerce platform provides a new sales model for agricultural products,so that agricultural products can enter the market more conveniently and quickly.The large number of online reviews generated by e-commerce platforms is the most important source of information that affects consumers ’ purchasing decisions.Product review information not only carries users ’ views and emotions on products,but also implies important information such as users ’ concerns and purchase suggestions.Therefore,the in-depth study of user reviews on e-commerce platforms can not only help operators more accurately grasp the needs of users,thereby improving their products and operations,but also give other users more reference value.This paper focuses on the research of online reviews of agricultural products,constructs an online review data set of agricultural products,and analyzes and processes the data samples in a deep learning way.Then use the model of visual data analysis to enhance the accuracy of information transmission,and design an interactive agricultural product online review sentiment analysis visualization system to help operators master and analyze data more effectively.The specific research contents are as follows :(1)In the part of information collection,the web crawler program of simulated login is written to achieve the purpose of information collection.Based on the collected text data,the online review text data set of agricultural products is constructed to provide complete and sufficient data for subsequent research.(2)For the online review data set of agricultural products,firstly,the data enhancement method is used to solve the problem of less neutral comments and negative comments in the data set.Through the results of comparative experiments of multiple pre-training models,the model Ro BERTa-wwm-ext with better sentiment classification of agricultural product text is obtained.Based on the pre-training model,based on the enhanced agricultural product online review text data set,adjust different loss functions and optimizers for model training,and finally obtain a high-accuracy agricultural product text sentiment classification model.(3)This paper designs and implements a visual analysis system for sentiment analysis of online reviews of agricultural products.Using graphics and other visualization methods,the conclusions of sentiment analysis are presented through five main visualization modules,as well as changes in sentiment trends under different time dimension data.Through the system interface interaction,users can also view the results of sentiment analysis of interest,as well as the display chart of fine-grained attribute sentiment information.