Knowledge graphs are a valuable tool that organizations can use to manage the vast amounts of data they collect, store, and analyze.
At Data Summit 2023, Joseph Hilger, COO, Enterprise Knowledge LLC and Sara Nash, senior consultant, data and information management, Enterprise Knowledge, LLC discussed “Building Knowledge Graphs” during their workshop session.
The annual Data Summit conference returned to Boston, May 10-11, 2023, with pre-conference workshops on May 9.
Enterprise knowledge graphs’ representation of an organization’s content and data creates a model that integrates structured and unstructured data. Hilger and Nash described what a knowledge graph is, how it is implemented, and how it can be used to increase the value of data.
“Information is just exploding, there’s more information in different formats than ever before,” Hilger said. “We have everything from sensors to open data. You can create so much data now.”
The expectation now with AI is dealing with unstructured data in a timely manner to uncover and protect insights, whereas before, companies only really had to deal with databases.
“We have natural silos within an organization,” Hilger said.
Knowledge graphs are a part of a data fabric, they frequently become the map that says how” X and Y” fit together, Hilger noted.
Knowledge graphs can store information in a machine readable and human understandable way. It can discover facts and patterns, add knowledge to data through how things fit together, search for data and content at the same time, and aggregate information from multiple disparate solutions.
“Graphs understand what something is,” Hilger said. “It can point at a document and say, ‘this document is a description of this product.’ You can query to get answers.”
Foundation of Knowledge Graphs
Nash broke down the building blocks that make up knowledge graphs. They consist of:
- Folksonomy
- Controlled list
- Taxonomy
- Ontology
- Knowledge graph
- AI
Taxonomies supports capturing the synonyms of things, Nash explained. They are powerful in helping to structure information across an organization.
“Things not strings,” Nash said. “When you get into taxonomy, you can describe the thing itself.”
Taxonomies work together with ontologies to describe what relationships exist between things. Linking the relationships between things and the structured data relationships between the information, that’s when a knowledge graph is created.
“[Taxonomies] take metadata and support a hierarchal structure,” Nash said. “It’s the best place to start to make semblance of the relationships between things.”
A common example of a knowledge graph people encounter everyday is Google’s knowledge graph, Nash noted. Schema.org is a good place to also start looking at how to construct a knowledge graph.
“It brings people into a space of searching information to the information itself versus finding a document that directs you then to the information,” Nash said.
Graph is not a one-size-fits-all solution, however, preferred use cases include highly interrelated data, sparse data, and flexible schema.
It is not recommended for data that isn’t interconnected, data that’s dominated by attributes rather than relationships, and rigid schema.
“Ultimately, there is a way to bring those things together,” Nash said. “A data fabric can help.”
Hilger described a use case involving Capital One. There was a lack of alignment around the meaning, format, and intent of data elements across organizational divisions, reducing the ability of data producers and consumers to find, use and trust data.
Enterprise Knowledge created a data fabric to underlie the knowledge graph, which enables data federation and virtualization of semantic labels or rules to capture and connect data based on business or domain meaning and value.
“Data fabric is what makes it all make sense,” Hilger said.
Enterprise Knowledge’s knowledge graph technologies include a graph database that provides a persistent data storage location for the knowledge graph and ontology data.
The data orchestration tool facilitates the exchange and transformation of data between systems and the graph database.
Then, the taxonomy/ontology management tool provides a user interface to allow business users to design build and govern their ontology and taxonomy models.
Many Data Summit 2023 presentations are available for review at https://www.dbta.com/DataSummit/2023/Presentations.aspx.