Earth Science Knowledge Graph

PIS: LEWIS MCGIBBNEY, NASA JPL | YONGYAO JIANG, ESRI

This project focused on creating an automatic approach to building a dynamic knowledge graph for Earth science, which would improve data discovery by leveraging implicit, latent existing knowledge present within the NASA DAAC websites.

Through the ESIP Lab project, the team:

• Created a prototype that improves search results based on user intent.
  • For example, when a user searches “sea surface temperature level 3”, the search engine in the data portal cannot understand the user’s real intent. By default, the search engine will convert the query to “sea” or “surface” or “temperature” or “level” or “3” to retrieve metadata. Several irrelevant medatada will be returned although they contain one or two words in the query, e.g. sea surface wind data. However, with the prototype, the query will be converted to “sea surface temperature”  or “SST” and “level 3”. In addition, the search engine knows “sea surface temperature” is a variable and “level 3” is processing level.
  • Tested geo-parsing options like Google Maps API, CLAVIN, geoparser.io, etc to extract location from a query on an AWS virtual machine.

PRESENTATIONS:

• Automatic Approach to Building Earth Science Knowledge Graph to Improve Data Discovery (ESKG)
• ESIP’s Earth Science Knowledge Graph (ESKG) Testbed Project: An Automatic Approach to Building Interdisciplinary Earth Science Knowledge Graphs to Improve Data Discovery

FOLLOW-ON EFFORTS:

ESKG has been considered within a larger NASA EOSDIS review of the use of knowledge graphs for representing EOSDIS data.