Review for NeurIPS paper: Diverse Image Captioning with Context-Object Split Latent Spaces