Research And Application Of Sentiment Analysis Model For Small Sample Data Based On Causal Rules

Text sentiment analysis can analyze the sentiment tendency of sentimental texts,and has created great value for the development of economy and society in the era of big data.The methods of text sentiment analysis can be divided into dictionary-based methods,machine learningbased methods,and deep learning-based methods.Methods.Machine learning and deep learning have achieved good results in text sentiment analysis tasks with large data scales,but machine learning relies on highquality feature construction and selection,and deep learning methods to achieve good classification results rely on large sample training models.It is difficult to achieve good classification results on small sample data,and the classification results are difficult to interpret.In fact,in many scenarios,it is difficult to obtain sufficient sample size,such as lack of knowledge,small amount of labeled data,and high cost of labeled data,such as social topic outbreak data in the field of public opinion monitoring,and online fraud data in the field of information security.For sentiment analysis of such small sample data,deep learning does not perform well.The causality is the most effective relationship for predicting events,it can make full use of small sample data to solve the prediction problem.Aiming at the shortcomings of machine learning and deep learning in the sentiment analysis of small samples of text and the uninterpretability of the classification results,the paper uses a combination of causal inference and machine learning to study the issue of sentiment analysis of small samples of text,verifying that the methods in the paper can Improve the accuracy of sentiment analysis of small sample texts,and interpret the classification results.The main research work of this paper is as follows:(1)Introduce causal analysis theory into the field of text sentiment classification,and use causal inference algorithms to mine causal rules between things from Chinese texts.The form of the causal rule is defined as:Ki,Kj?Kk,which means that there is a direct causal relationship between the three variables Ki,Kj,and Kk,that is,the variables Ki and Kj are the causes of the variable Kk.The causal rules in this article are composed of keywords in the text.The TF-IDF algorithm is used to extract keywords from the training set text,and the keywords are formed into a keyword dictionary,which is used to vectorize the text;a causal inference algorithm is used The causal rules are mined from the vectorized data,and the causal rules are used as the classification features of the naive Bayes classifier.This paper uses the CCU causality inference algorithm to mine causal rules,which is a constraint-based causality inference algorithm.It uses variable association,independence and conditional independence tests to limit possible causal relationships between variables,and uses local search methods to discover causal relationships.(2)Propose a sentiment analysis model based on causal rules.The model combines causal rules and machine learning algorithm Naive Bayes for text sentiment classification.This method counts the emotional polarity of each causal rule mined from the text,calculates the probability of each causal rule appearing in the positive and negative text category,and uses the causal rule with emotional polarity as the naive Bayes algorithm Classification features,calculate the occurrence of positive and negative causal rules in the text to be classified,predict the text category,and use emotional causal rules to interpret the classification results.The performance of the model is verified through experiments.Using a small sample of financial news text data sets,the proposed model is compared with other benchmark models on the data set.The experimental results prove that the model proposed in this article is accurate and precise.Relatively good performance on other evaluation indicators has verified the effectiveness and stability of the model.(3)Aiming at the small sample data scenario of online course reviews,apply the sentiment analysis model based on causal rules proposed in this article to the sentiment analysis of course reviews on the MOOC platform,perform sentiment analysis on course reviews,and use sentiment causal rules to classify the results Explain,provide a higher level of course feedback information for learners and professors.The design and implementation of a prototype system for sentiment analysis of course reviews based on causal rules has demonstrated and tested the functions of the system,which verified the usability of the system.