dual learning for machine translation

A framework that leverages the dual nature of two tasks (e.g., English to French translation and French to English translation) to provide feedback to each other.