Sentiment analysis is natural language processing technique to mine the valuable information from large unstructured data. There were numerous studies had been made on different domain for sentiment analysis e.g. movie reviews, health care, etc. We are applying sentiment analysis in educational domain for assessment of teacher feedback. In this we are creating automatic teacher assessment model by applying mutual information for extraction of related features form text. This extracted feature is applied as input to support vector machine algorithm. We are using SVM for sentiment classification. In this will provide multiclass classification of text into more positive, positive, more negative, negative and neutral classes.