Opinion extracting and classification from questionnaire comments using HMM-POS Tagger and machine learning techniques

Measurement of academic services using questionnaires with multiple choice answers generally provide comments and advice columns. In results data analysis, comments and suggestions given by the thousands of student cannot be utilized due to the lack of analysis tools. Whereas comments and suggestions actually contain student opinions on various things, such as facilities, faculties, library and others. Opinion mining and sentiment analysis as a new tool in text mining can be applied to utilize comments and suggestions. This research has applied HMM-POS Tagger providing automatic TAG POS to the comments data on the training data using Hidden Markov techniques. By implementing the HMM-POS Tagger, opinion can be extracted from the comments. Furthermore if the comment is opinion, by using rule-based, it can be determined the target of the opinion and also the orientation of the opinion whether it is positive or negative. The data used was 1,000 comments given POS-TAG manually and 500 comments set as test data. Sentiment analysis is applied using four methods of classification, namely SVM, NBC, ME and KM-Clustering. The results showed that HMM-POS Tagger get precision of 0.95 for the detection of opinion and 0.91 for target detection. In the opinion classification results showed accuracy of SVM, NBC, ME and KM-Clustering are 0.84; 0.83; 0.84 and 0.88 respectively.