Text is the major method that is used for communication now a days, each andevery day lots of text are created. In this paper the text data is used for theclassification of the emotions. Emotions are the way of expression of thepersons feelings which has an high influence on the decision making tasks.Datasets are collected which are available publically and combined togetherbased on the three emotions that are considered here positive, negative andneutral. In this paper we have proposed the text representation method TFIDFand keras embedding and then given to the classical machine learning algorithmsof which Logistics Regression gives the highest accuracy of about 75.6%, afterwhich it is passed to the deep learning algorithm which is the CNN which givesthe state of art accuracy of about 45.25%. For the research purpose thedatasets that has been collected are released.