Twitter, as one of microblogs, has become a predominant media used in expanding the service quality to the customer. In Indonesia, for example, most of National airline companies employ Twitter as a media in order to notify their customers about up-to-date information as well as to monitor all real time comments from the customers. Varied opinions (positive, negative, or netral) expressed by netizens through micro-blogging are interesting to analyze. Analysis of netizens' responses to the services provided by the companies can be done with a study of Sentiment Analysis. In this recent study, data collection technique used was crawling technique available on Crawler4j. Furthermore, some processes would be performed to convert unstructured data into structured data including cleaning, case folding, parsing and filtering. The data were then extracted into a certain group of words and the words were then simplified into their basic form using Vega Algorithm. Each word’s weight will be calculated by using the TF-IDF method. Each comment then was classified into positive, negative, or netral opinion with Support Vector Machine method. Two open source library, those were LIBSVM and LIBLINEAR used in the experiment. The results obtained that the text can be classified with the accuracy of 93.7128%.

Keywords—Support Vector Machnine; sentiment classification; opinion; micro-blogging;