GeoPlatform Knowledge Graph

From Earth Science Information Partners (ESIP)

Chris MacDermaid Xentity, Lucas Scharenbroich Pro-West & Associates, Nathan McEachen TerraFrame and Joel Schlagel DOI,

Discovery Cluster Telecon: November 17, 2022

Agenda:

  • GeoPlatform Knowledge Graph Background - Joel Schlagel
    • Goals
    • Geospatial Data Act
    • 10 years of Knowledge Graph
  • GeoPlatform Background - Chris MacDermaid
    • GeoPlatform
    • GeoPlatform FAIR Insights
    • STAC collection search
  • GeoPlatform Geospatial Knowledge Infrastructure - Nathan McEachen
    • A Common Geo-Registry for Sharing Authoritative Geographic Data as They Change Over Time for Knowledge Graph Interoperability
    • Data Mesh Architecture
  • GeoPlatform Generated Knowledge Graph - Lucas Scharenbroich
    • Generative Pipeline
    • Extracting Knowledge Graphs from Datasets
    • Integrating with External Knowledge Graphs
  • GeoPlatform Knowledge Graph future / Questions - Team: Chris/Joel/Lucas/Nathan

Notes:

The mission of the Geoplatform is to provide data discovery of federal geospatial data.

Geospatial data act of 2018 re-ups this mandate, and it says that data need to be accessible not just discoverable using geoPlatform.

Geoplatform harvests geospatial metadata from Data.Gov, and it provides insights on data findability, accessibility, and reusability.

This geospatial infrastructure also allows sharing agencies' knowledge about geospatial objects, so agencies can derive the knowledge about geographic objects from authoritative sources.

The geography is also a basis of dataset interoperability.

One way to do this is using a geo-registry - Geographic objects are the common link between data sources.

Agencies should publish knowledge graphs allowing other agencies to pull the data and merge it with their own knowledge graphs.

This allows agencies to find relationships between geographic objects without having to join multiple features (moderator's note: can query geospatial relationships without the computational expense of loading feature data and performing geospatial operations).

Demos provided by Geoplatform shows Geoplatform's early implementations of these approaches.

Questions:

Doug Newman: Graphs should be stewarded by people who have the knowledge about them. This is an argument for many graphs, rather than one uber-graph. Question about traversal across distributed graphs - can we use virtual links? This could save overhead used to maintain a geo registry.

Dalia Varanka: Knowledge graphs that link to siloed data within agencies could resolve to one another when agencies identify the relationship across the agencies key knowledge concepts?

Josh Schlagel: This is still proof of concept - still trying to demonstrate how this could work.