An Approach to Disambiguate the Innate Sense of a Message with the Help of Emoji

Abstract

This project aims to understand the tonality of a text using emoticons as the predictor. The emoticons have transpired to be an increasingly significant and standardized part of modern textual inputs. The combination of visual and textual content in the same message builds a compact form of communication, sometimes which becomes tough to portray by the words only. Emoticon prediction and word-emoticon embedding could be used for any kind of Sentiment Analysis based works like Opinion Mining, User Profiling etc. This work focuses on identifying the tone of emotion behind a body of text which is used to analyze the sentiment of a user according to h(is)er comments or messages in online media. This experiment is carried out on the comments collected from some Twitter handles, Facebook walls and Whatsapp conversations. During the experiment, it is observed that in a few cases the senses of different comments are not possible to retrieve only from the words, but the texts which are combined with some emoticons are evaluated correctly. Challenges faced in this work are explained at the end of this report.