{"title":"Sentiment Analysis Using Naive Bayes Approach with Weighted Reviews - A Case Study","authors":"Brandon Joyce, Jing Deng","doi":"10.1109/GLOBECOM38437.2019.9013588","DOIUrl":null,"url":null,"abstract":"Online reviews are critical in many aspects, for business as well as customers. Yet the accuracy and trustworthiness of these reviews are usually unsubstantiated and little research has been performed to investigate them. In this work, we use a set of Yelp reviews on various topics (food, hotel, etc.) as an example to perform sentiment analysis and investigate the correlation between review comment sentiment and its numeric rating. We use feature selection techniques to statistically remove redundant words from reviews, thus improving run time and accuracy. Our method gives higher weight to those terms/words appearing in reviews with more useful votes. These techniques combined with Naive Bayes approach achieves an overall accuracy of 75%. More interestingly, our method is shown to perform well in 1-star and 5-star reviews, with a 92% accuracy for the latter. With such a strong accuracy, we argue that the proposed sentiment analysis technique can be used to shed light on all online comments, especially those without numerical ratings.","PeriodicalId":6868,"journal":{"name":"2019 IEEE Global Communications Conference (GLOBECOM)","volume":"30 1","pages":"1-6"},"PeriodicalIF":0.0000,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2019 IEEE Global Communications Conference (GLOBECOM)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/GLOBECOM38437.2019.9013588","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 2
Abstract
Online reviews are critical in many aspects, for business as well as customers. Yet the accuracy and trustworthiness of these reviews are usually unsubstantiated and little research has been performed to investigate them. In this work, we use a set of Yelp reviews on various topics (food, hotel, etc.) as an example to perform sentiment analysis and investigate the correlation between review comment sentiment and its numeric rating. We use feature selection techniques to statistically remove redundant words from reviews, thus improving run time and accuracy. Our method gives higher weight to those terms/words appearing in reviews with more useful votes. These techniques combined with Naive Bayes approach achieves an overall accuracy of 75%. More interestingly, our method is shown to perform well in 1-star and 5-star reviews, with a 92% accuracy for the latter. With such a strong accuracy, we argue that the proposed sentiment analysis technique can be used to shed light on all online comments, especially those without numerical ratings.