•  
  •  
 

Scientific Information Research

Keywords

vector database; multimodal data fusion; vector data retrieval; vector data indexing; AI application ecosystem

Abstract

[Purpose/significance]The article reveals the theoretical systems, technological systems, and applied systems of vector databases, aiming to promote innovation in the research and practice of multimodal AI related theories, technologies, and applications. [Method/process]This article elaborates on the evolution of vector databases and defines its core concepts through literatures tracing and content analyzing. Subsequently, it compares and analyzes their characteristics and values, and based on this, sorts out their application mechanisms, functions, corresponding key technologies and application modes. Simultaneously, it discusses the challenges and countermeasures faced by vector databases, and looks forward to their development trends from theoretical, technical, and application perspectives. [Result/conclusion]Vector databases originate from the construction of the vector index method system, develop in vector data retrieval engine construction, and mature in vector database management system construction. Compared to relational databases and graph databases, vector databases exhibit obvious characteristics in data models, indexing mechanisms.They hold various value for users, data managers, developers and researchers. The key technologies are divided into

Digital Object Identifier (DOI)

10.19809/j.cnki.kjqbyj.2024.04.002

Download

Share

COinS