Emotion Classification and Respective Sticker Mapping Using Convolutional Neural Networks
Sayan Ghosh, Raja Karmakar, Twinkle Chatterjee, Sandipan Ghosal · 2023
The utilization of social media is aggravating and so is the demand of using messaging apps, to stay connected. People now not only send text messages but also prefer using stickers which is the new alternative to emojis. While an emoji can only portray emotion, stickers are used to graphically illustrate and capture the human mind’s attention by being more relatable and accurate. One thing to consider here is that men and women have a different choice of emoji to express their emotion so a random emoji may be ambiguous, i.e., confusing to the other gender. Now, emotion analysis has gained significant importance through online media in recent years. Thus, we have tried to primarily focus on two steps, namely emotion classification and gender analysis for sticker generation. Therefore, this paper employs machine learning models to classify emotions and predict gender to generate accurate stickers. Convolutional Neural Networks (CNNs) are highly productive in feature extraction and object classification on images which is why we have used a CNN for emotion analysis and another CNN for gender analysis combining both we can map the stickers from a prebuilt directory.