Improving Sentiment Analysis on Twitter with Intention Classification
Since the emergence of Web 2.0 or Social Web, microblogging websites such as Twitter have become very popular among users on the Web. Microblog users often have different intentions when expressing their thoughts by posting messages on the websites. The intentions typically range from updating their current activities, sharing their opinions on various topics to advertising on products and services. Current opinion mining and sentiment analysis approaches have difficulty in handling tremendous quantity of posts. One of the reasons is due to all posts, some of which do not contain any sentiment or opinion, are included in the analysis process. This paper proposes a content filtering module based on text classification technique for screening irrelevant posts. From the experimental results, using Support Vector Machines (SVM) with the feature selection technique, Information Gain (IG), yielded the highest accuracy of 84.5％.
component Sentiment Analysis, Opinion Mining, Social Media, Microblog, Text Classification
Wilas Chamlertwat Pattarasinee hattarakosol Choochart Haruechaiyasak
Technopreneurship and Innovation Management Programt Chulalongkorn University Bangkok, Thailand Department of Mathematics, Faculty of Science Chulalongkorn University Bangkok, Thailand Human Language Technology Laboratory National Electronics and Computer Technology Center Pathumthani
国际会议
海口
英文
254-258
2011-07-15（万方平台首次上网日期，不代表论文的发表时间）