Sentiment analysis of micro-blogs, such as twitter, is one of the hottest topics. The speed of information propagation is becoming faster and faster. We cannot control the flow of information on tweets. So, we need to know the characteristics of such communication tools. The present paper extracts the features of emotional tweets, based on feature selection by SVM. An attention is paid to part of speech, particularly to the particles. Copyright © 2012 by International Society of Computers and Their Applications (ISCA).