Improving OOV handling in recommendation systems is crucial for better performance, especially in the inductive setting.