Research on Emoji Sentiment Recognition Based on Convolutional Neural Networks

Emoji usage has become increasingly prevalent in social interactions, making emojis essential tools for communication and marketing. This study proposes an emoji recognition method based on convolutional neural networks (CNN). The approach involves grayscale processing of emoji images, feature extraction using convolutional kernels, dimensionality reduction through pooling methods, and finally, emoji classification using a Softmax classifier. The CNN-based recognition method achieved an accuracy rate of 83.54% in emoji recognition tasks.