Building a Knowledge Graph for Food, Energy, and Water Systems
Metadata[+] Show full item record
A knowledge graph represents millions of facts and reliable information about people, places, and things. Several companies like Microsoft, Amazon, and Google have developed knowledge graphs to better customer experience. These knowledge graphs have proven their reliability and their usage for providing better search results; answering ambiguous questions regarding entities; and training semantic parsers to enhance the semantic relationships over the Semantic Web. Motivated by these reasons, in this thesis, we develop an approach to build a knowledge graph for the Food, Energy, and Water (FEW) systems given the vast amount of data that is available from federal agencies like the United States Department of Agriculture (USDA), the National Oceanic and Atmospheric Administration (NOAA), the U.S. Geological Survey (USGS), and the National Drought Mitigation Center (NDMC). Our goal is to facilitate better analytics for FEW and enable domain experts to conduct data-driven research. To construct the knowledge graph, we employ Semantic Web technologies, namely, the Resource Description Framework (RDF), the Web Ontology Language (OWL), and SPARQL. Starting with raw data (e.g., CSV files), we construct entities and relationships and extend them semantically using a tool called Karma. We enhance this initial knowledge graph by adding new relationships across entities by extracting information from ConceptNet via an efficient similarity searching algorithm. We show initial performance results and discuss the quality of the knowledge graph on several datasets from the USDA.
Table of Contents
Introduction -- Challenges -- Background and related work -- Approach -- Evaluation -- Conclusion and future work