Introducing a novel codebook transfer framework with part-of-speech enhances image modeling by leveraging pretrained language models.