Statistical text analysis and sentiment classification in social media

In this paper, we propose a new method of classifying tendencies and opinions in texts of multiple sentence length extracted from social media and covering both formal and informal vocabularies. To extract contextual information from the texts, we carry out computations based on keywords, the position of the sentence and the flow of sentiments in the multiple texts. A feature vector for the given text is constructed from the contextual information, and is then classified with a Support Vector Machine (SVM) classifier as positive, negative or neutral. Our method performs well in classifying the gradient of sentiments expressed in social media.