Emotion Detection using Online Machine Learning Method and TLBO on Mixed Script

Due to rapid modernization of our societies, most people, if not all, have access to online social media and mobile communication devices. These people hail from diverse cultures and ethnicity, and interact with each other more often on these social media sites. Moreover, due to their distinct backgrounds, they all have an influence on the common language in which they communicate. Also, many users employ a myriad of shorthand, emoticons and abbreviations in their statements to reduce their effort. This calls for a means to assist in better communications through social media. In our work, we have researched on understanding the underlying emotions and sentiments of these interactions and communications. Our focus was on analyzing the conversations by Indians in the code-mix of English and Hindi languages and identifying the usage patterns of various words and parts of speech. We have categorized statements into 6 groups based on emotions and improved the model using TLBO technique and online learning algorithms. These features were integrated in our application to assist the mobile device users in quickly sort and prioritize their messages based on the emotions attached with the statements and provide much more immersive communications with their friends and family.