IDENTIFY SARCASTIC TWEETS WITH COMPARISON OF TEXT AND EMOJI

Abstract

The objective of this project is to identify the sarcastic twitter tweets with the comparison oftext and emoticons. The report investigates various techniques used for the identification of sarcastic tweets. Different stages involving twitter tweets processing which includes pre-processing stage, segmentation of text and emoticon and further more feature extraction will also be deliberate and discussed. Lastly the end result ofthe algorithms will identify the sarcastic tweets. The system first proceeds with the pre-processing ofthe dataset with 9 thousand plus oftweets. We have selected the tweets which are having only happy and sad emoticons. After that we have separated the text and emoji. Using the Text-Blob python library, have to calculate the polarity ofthe text (i.e. positive or negative or neutral). We then cleaned the text (i.e. removing stop words, punctuation etc.) and by using count vectorizing library we created a bag of words model from which we will obtain matrix, after that splitting the dataset into training and testing, and applying we a sparse Gaussian Naive Bayes, Multinomial Naive Bayes algorithm to create a model for predicting polarity and emoticon ofthe unseen text.