So sánh phương pháp

	Xử lý ngôn ngữ tự nhiên đa phương thức	Cơ chế chú ý (Attention Mechanism)	Transformer Thị giác
Lĩnh vực	Khai phá văn bản	Học sâu	Học sâu
Họ	Process / pipeline	Machine learning	Machine learning
Năm ra đời	2021 (modern era, CLIP onward)	2015	2021
Người khởi xướng	Radford et al. (OpenAI) — CLIP, 2021; Li et al. — BLIP-2, 2023	Bahdanau, D.; Luong, M.T.	Dosovitskiy, A. et al.
Loại	Cross-modal understanding and generation pipeline	Neural attention layer (encoder-decoder)	Transformer architecture for images (self-attention over patches)
Công trình gốc	Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., & Sutskever, I. (2021). Learning Transferable Visual Models From Natural Language Supervision. Proceedings of the 38th International Conference on Machine Learning (ICML), 8748–8763.	Bahdanau, D., Cho, K. & Bengio, Y. (2015). Neural Machine Translation by Jointly Learning to Align and Translate. ICLR.	Dosovitskiy, A. et al. (2021). An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. ICLR.
Tên gọi khác	Çok Kipli NLP (Multimodal NLP), vision-language models, multimodal learning	Dikkat Mekanizması (Bahdanau / Luong Attention), dikkat mekanizmasi, neural attention, additive attention	Görsel Transformer (ViT), görsel transformer, ViT, patch transformer for images
Liên quan	4	5	5
Tóm tắt	Multimodal NLP is a family of natural-language-processing pipelines that combine text with one or more additional data modalities — most commonly images, but also audio and video — to perform understanding and generation tasks such as visual question answering, image captioning, and multimodal sentiment recognition. The field gained its modern form with CLIP (Radford et al., 2021) and has since advanced through architectures such as BLIP-2 (Li et al., 2023) that bridge frozen image encoders and large language models.	The attention mechanism, introduced by Bahdanau, Cho and Bengio in 2015 and refined by Luong, Pham and Manning the same year, lets a sequence decoder dynamically learn which of the encoder's outputs to focus on at each step. Before the Transformer, it substantially improved machine-translation quality by freeing models from compressing an entire input into a single fixed vector.	The Vision Transformer (ViT), introduced by Dosovitskiy and colleagues in 2021, splits an image into fixed-size patches, treats those patches as a sequence, and applies the Transformer self-attention mechanism to image classification. Given enough training data, it surpasses convolutional neural networks (CNNs).