Across numerous fan created and curated websites, there exists a wealth of semantically rich descriptive data for a variety of Japanese visual media, such as anime, manga, and video games. The amount and granularity of these data makes it valuable for domain researchers, but the semantic heterogeneity and lack of interconnectedness makes analysis cumbersome. Seeking to address this issue, the Japanese Visual Media Graph ( JVMG ) project aims to create a type of global database built using RDF and external ontologies to better enable data-driven research in the domain. We discuss the benefits this approach has when compared to the local relational databases used at the fan-sites, such as enabling the easy creation of aggregate resources using data from multiple providers, and the ability to constantly update and alter a schema over time. This encourages the incorporation of additional data in the future, while still maintaining rich data semantics and provenance. While many of the benefits are discussed in the context of the JVMG project or the Japanese visual media domain, we discuss how this database approach may be similarly advantageous to other projects that seek to create aggregate resources, or collect heterogeneous data from diverse sources.