The Future of Knowledge Graphs

Highlights

"Classes are out, Shapes are in. Ontologies become complicated because everything is represented as a class, and the way we build classes largely involves inheritance. Shapes are more contextual and open-ended; you can describe the characteristics of classes that have different properties that change upon the context in which they are embedded. This becomes especially powerful when combined with SPARQL, which allows you to determine that context very robustly. "
"Dynamic Knowledge Graphs. The second approach is to either build models or use reification to create conditional knowledge graphs (KGs). The latter is better suited for Long graphs where property values (and even their associated entities) change over time, and you want to retain this value for analytics; however, it also makes for more complex queries. I expect that we’ll see additional abstraction layers that make this process less painful and more transparent."
"Event Driven KGs. Systems are event-driven. We receive and interpret signals (messages), then figure out what to do with them. A knowledge graph is a description of an environment, with entities interacting within it, and this shift is changing the way we think about knowledge graphs from a store of knowledge to a self-contained simulation system."
"SHACL will end up being a lingua franca for semantic interchange"

Backlinks