Study On The Application Of Hierarchical Bayesian In Emotional Classification

With the rapid development of the Internet,people's love for the Internet has also increased.In the process of the birth and development of the Internet,various platforms are produced,and users also express their inner thoughts and opinions through the Internet platform.With more and more users,various data such as society and commodities are involved in these data,and this information also forms a huge database.In the process of database formation,it also brings sensational problems in the field of natural language processing.Therefore,these data need to be processed and monitored.In order to achieve this goal,scholars who have studied in the field of natural language processing have been exploring it in recent years.In order to conduct emotional research on huge information,researchers in this field use lexicography,statistics,and machine learning methods to process them,and finally implement automated text processing methods.Based on the previous research results of the scholars,this paper adopts the improved combination of TF-IDF and occulty Bayesian to classify the text sentiment,and the combination of improved TF-IDF and naive Bayes and feature-weighted fusion.The naive Bayesian sentiment classification algorithm was compared.Then the extension of the sentiment dictionary is studied.The dictionary is expanded by the base word2 vec method,and then the implicit classification Bayesian is used as the classifier for emotion classification.This paper uses books,computers,and hotel review data as the object of experimental research.Through the analysis and processing of data,the related work of emotion classification and dictionary expansion is carried out.The main research contents of this paper are as follows:Under the research of the emotion classification related technology in this paper,the method of model fusion is used for the process of text sentiment classification.For text preprocessing,use the jieba word segmentation tool in Python for word segmentation,remove stop words,and so on.The feature extraction and sentiment classification models are studied,including mutual information,CHI statistics,TF-IDF,etc.The improved feature extraction process of TF-IDF algorithm is studied in detail.The sentiment classification algorithm studies the use of the classification process of the implicit Bayesian and analyzes the advantages of this sentiment classification algorithm.Emotional classification based on improved TF-IDF combined with the implicit naive Bayesian algorithm.The improved TF-IDF algorithm is used to extract the text features,change the text into vectorization,and use the implicit naive Bayesian classification algorithm as the classifier to classify the text emotions.The test data was used for the test,and the test results were analyzed using the evaluation indicators,and compared with the above two methods,the test results were superior to the other two methods.The expansion of the emotional dictionary.Emotional classification or sentiment analysis methods in the field of natural language processing include emotion-based dictionaries.At present,representative sentiment lexicons include: Hownet,the Taiwan University Dictionary(NTUSD),and the Dalian University of Science and Technology Chinese Emotional Vocabulary Ontology Library(without auxiliary emotion classification).However,new words in the network may be ignored.In this paper,the word2 vec algorithm is used to expand the dictionary.The content of the text is matched with the dictionary to select the emotional words.For the new words of the network,the word2 vec method is used to calculate between the words and the words.Relationship,and finally use the ambiguous Bayesian as a classifier to classify the text emotions.