Improving the Accuracy of Sarcasm Detection in Text Data Using a Smooth Support Vector Classification Model with Word-Emoji Embedding for News and Indian Indigenous Languages