{"title":"A sentiment classification algorithm of Bi-LSTM model fused with weighted word vectors","authors":"Chaohui Chai, Dong-Ru Ruan","doi":"10.1109/ICCEAI52939.2021.00049","DOIUrl":null,"url":null,"abstract":"As a hot topic in the field of natural language processing, sentiment classification has always attracted much attention. With more and more comments from users in different fields, it is necessary to build a more accurate and efficient sentiment classification model. The traditional distributed word vector representation method cannot well represent the ability of words to distinguish text and cannot capture the emotional information in sentences. We use Word2vec to obtain semantic information between words, obtains word distribution representation characteristics, and then combines emotional dictionary to judge word emotional information, uses TF -IDF algorithm to construct the word distribution characteristics of weighted word vectors, so as to effectively capture the emotional information of contextual sentences. Finally, Combining the bi-directionallong short-term memory network (Bi-LSTM) model can get more accurate sentiment classification results. The experimental results show that after selecting appropriate model parameters, the weighted word vector method combined with emotional information and the distributed feature vector method based on semantic relations have improved accuracy and other indicators; through the feature representation of the weighted word vector Methods Compared the Bi-LSTM model with other text classification models, it is concluded that the feature representation method of the weighted word vector has improved the classification results in each model, and the classification effect is the best in the Bi-LSTM model.","PeriodicalId":331409,"journal":{"name":"2021 International Conference on Computer Engineering and Artificial Intelligence (ICCEAI)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2021-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 International Conference on Computer Engineering and Artificial Intelligence (ICCEAI)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICCEAI52939.2021.00049","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
As a hot topic in the field of natural language processing, sentiment classification has always attracted much attention. With more and more comments from users in different fields, it is necessary to build a more accurate and efficient sentiment classification model. The traditional distributed word vector representation method cannot well represent the ability of words to distinguish text and cannot capture the emotional information in sentences. We use Word2vec to obtain semantic information between words, obtains word distribution representation characteristics, and then combines emotional dictionary to judge word emotional information, uses TF -IDF algorithm to construct the word distribution characteristics of weighted word vectors, so as to effectively capture the emotional information of contextual sentences. Finally, Combining the bi-directionallong short-term memory network (Bi-LSTM) model can get more accurate sentiment classification results. The experimental results show that after selecting appropriate model parameters, the weighted word vector method combined with emotional information and the distributed feature vector method based on semantic relations have improved accuracy and other indicators; through the feature representation of the weighted word vector Methods Compared the Bi-LSTM model with other text classification models, it is concluded that the feature representation method of the weighted word vector has improved the classification results in each model, and the classification effect is the best in the Bi-LSTM model.