Conditional Generative Adversarial Networks for Emoji Synthesis with Word Embedding Manipulation

Emojis have become a very popular part of daily digital communication. Their appeal comes largely in part due to their ability to capture and elicit emotions in a more subtle and nuanced way than just plain text is able to. In line with recent advances in the field of deep learning, there are far reaching implications and applications that generative adversarial networks (GANs) can have for image generation. In this paper, we present a novel application of deep convolutional GANs (DC-GANs) with an optimized training procedure. We show that via incorporation of word embeddings conditioned on Google's word2vec model into the network, the generator is able to synthesize highly realistic emojis that are virtually identical to the real ones.