TWITTER SENTIMENT ANALYSIS

We present an approach of novel for automatically to classify the feeling of the messages of Twitter. These messages are classified as positive or negative with regard to a limit of question. It is useful for the consumers who want to seek the feeling of the products before purchase, or the companies which want to supervise the public feeling their marks. There is no preceding research on classifying the micro feeling of the messages to the services blogging like Twitter. We have the results of the algorithms of study of machine to classify the feeling of the messages of Twitter by using the distant monitoring. Our data of formation are composed of the messages of Twitter with the emoticons, which are employed as noisy labels. This type of data of formation is abundantly available and can be obtained by automated means. We prove that the algorithms of study of machine (naive Bayes, maximum entropy, and SVM) have exactitude above 80% once exerted with data of emoticon. This article also describes pretreatment necessary stages in order to carry out high degree of accuracy. The principal contribution of this article is the idea to employ beeps with emoticons for the distant directed study.