Emotion Classification on Twitter Data Using Word Embedding and Lexicon Based Approach

Twitter is one of the leading social media platforms for its users for tweeting 280 characters in a single tweet. Social media influence the users to share data, promoting advertisement and posting useful information to the followers. The social media data helps the individuals and business personals to take decision based on the analysis with the data. Twitter sentimental analysis is important for identifying the similar text patterns present in the given input text. The analysis further classified with emotional, positive and negative tweets. Existing approaches limited in accuracy, therefore, the word embedding and lexicon based approach has been introduced for increasing the accuracy. The twitter data stream is taken as an input and preprocessed by removing stop words, hash tags and urls. Further, the data is tokenized and applied with the word embedding method to detect the location and lexicon based approach to segregate the sentimental and emotional tweets. The system has been tested with the live data set as well as offline dataset, and the result shows very promising.