A Model and Query Language for Multi-modal Hybrid Query
Chuan Hu, Zihao Zhao, Along Mao, Zhihong Shen · 2024
As data grows exponentially, its diversity also increases, including both structured forms and unstructured forms like audio, images, and videos. Advances in AI have improved our ability to analyze unstructured data, leading to the use of multimodal hybrid queries that blend structured and unstructured data. However, database systems struggle due to the lack of adequate data models for multimodal data and languages for these hybrid queries. This paper extends the property graph model to represent multimodal data and their semantic information, introducing essential functions for hybrid graph queries. A high-level graph query language, CypherPlus, is presented, capable of expressing hybrid queries like “Give me the friends of the friends of Mary, who have blond hair and are younger than 30 years old.” A Neo4j-based implementation and experiments over synthetic and real-world datasets demonstrate the approach’s plausibility.