Multilingual GAN

A Multilingual GAN pairs the generative adversarial framework with cross-lingual components — a shared encoder, language-conditioned generator, and a language discriminator — so that a single model can generate or align representations across multiple languages simultaneously. It is applied to cross-lingual text generation, machine translation, multilingual data augmentation, and language-invariant feature learning.