Research on Military Multimodal Knowledge Fusion Representation Based on Retrieval-Augmented Generation
Jing Yao, Junlin Li, Bowen Zhao · 2024
There are various forms of military knowledge and intelligence, including text, audio, images, videos and other modalities. Traditional knowledge representation methods cannot effectively associate and fuse these knowledge and intelligence, but with the rapid development of large language models based on retrieval augmentation generation technology, they have been able to identify and process multimodal information, so on this basis, this paper aims to realize the association representation of knowledge through the retrieval and multimodal processing capabilities provided by retrieval augmentation generation technology, build a vector database for multimodal battlefield intelligence knowledge, provide efficient storage and retrieval capabilities, and effectively associate and fuse battlefield knowledge to achieve effective battlefield multimodal Knowledge fusion representation.