Research And Implementation Of Emotion Analysis System Of Online Commodity User Comments

In these times,people's living standard and lifestyle have been continuously improved while the Internet technology developed rapidly.Internet technology has penetrated all aspects of people's life,and online shopping has become a new mainstream way of shopping.According to statistics,more than half of consumers will comment on the goods they buy online.These comments provide especially important reference information for producers and consumers.In the face of large-scale user comment data,how to design an efficient,accurate,and easy-to-use comment analysis system,and display the analyzed data in an intuitive perspective,develop into a hot research field.To display the emotional tendency of user comments in a fine-grained way,the user comments emotional analysis algorithm designed in this paper transforms the emotional analysis problem into entity tagging problem and emotional classification problem.Through the improvement of BERT pre-training model,activation function and optimization function,the effect of BiLSTM+CRF entity labeling algorithm is improved.Then through the rule-based matching algorithm to match the subject words and emotional words,K-means to classify the subject words and emotional classification based on emotional dictionary,a set of feasible fine-grained emotional analysis algorithm flow is integrated,which can analyze the emotional tendency of commodity users' comments according to different attributes of commodities.The emotion analysis system designed in this paper is based on Python language,uses Django to design web framework,implements BERT+BiLSTM+CRF model and K-means algorithm through scikit-learn and Keras,and uses PyEcharts to show the analysis results of comment data.Finally,through the system test,the system meets the design requirements.The research,design and implementation process of the system are as follows:1.System requirements analysis stage: the first task is to determine the goal of system realization,analyze the feasibility of the system according to the realization goal,then analyze the function of the system in detail from the perspective of top-level use case analysis,and finally analyze the non-functional requirements of the system from the perspective of non-functional requirements such as security,ease of use,maintainability,etc.2.System overall design and detailed design phase: the first step is to design for the overall architecture of the system,then use HIPO mode to design the overall functional diagram of the system,and use IPO table to carry out detailed design of each functional module,and finally design the database of the system through entity attribute diagram and data table.3.Design stage of emotion analysis algorithm: This paper proposes a label model based on BERT-Bi LSTM-CRF,which improves the accuracy of tagging subject words and emotion words,and designs a set of emotion analysis algorithm flow,which can extract fine-grained emotion tendency of comments.4.System implementation and testing stage: the system writing platform is PyCharm,the main part of the algorithm is written in Python language combined with machine learning tool Scikit-learn and advanced neural network Keras API,and the front-end functions are implemented in JavaScript and other languages.Finally,we use the designed test cases to test the functionality of the system.