Art Murray demonstrates tools for enhancing legacy knowledge bases at KMWorld Connect 2021

Many knowledge bases and repositories are built using traditional content management systems with an underlying content database.

While useful, however, content databases are limited, especially for users seeking a “birds-eye” view of how the many concepts and topics are interrelated—a key requirement for KM. The recent widespread interest in knowledge graphs has been partly in response to this limitation, as they provide a visualization of the topics and their interrelationships. The challenge is how to effectively integrate both technologies.

In a presentation at KMWorld Connect 2021, Art Murray, CEO, Applied Knowledge Sciences, showed an approach that was successfully applied using a number of technologies including automated entity-relationship extraction, graph databases (Neo4j), and knowledge graphs (neovis.js). It also emphasizes the importance of incorporating a standard ontology (SUMO/KIF) into the data model as a means of achieving improved disambiguation and reduction of conflicting terminology and meaning. The end result was not a replacement of the original content database, but rather a significant enhancement, as each component in the new system addressed a different user need. 

The presentation also addressed challenges faced when attempting to integrate graph databases with legacy data more suited for the relational model. In addition, the importance of engaging subject matter experts (SMEs) for data model validation and fine-tuning the results, including usability testing, was demonstrated.

Applied Knowledge Sciences' “bread and butter business” is knowledge transfer to help companies "pull knowledge from people’s heads" in order to document it and make it available for everyone in the organization. With Applied Knowledge Sciences now past its 25^th year, Murray said he decided it was time to practice what he preaches and put the company’s information into a knowledge base.

Murray walked attendees through the stages it took to create the knowledge base, as well as the individual processes in each stage and the tools and technologies used.

If attendees were to take anything away from this presentation, Murray said, he hoped it would be that there are a lot of great tools out there for performing text analytics and extracting knowledge from text, and that it doesn't have to be expensive, since there are many effective free and low-cost tools. 

Murray’s observations and lessons learned

Relational model and graph model each have their place

• e.g., relational model is well suited for time-series data meant to be entered/read every 10 seconds and you want to refer to x years of data
• Graphs can represent tables
  • Node property is a URL link to a RESTful resource (e.g., GET/POST data from a table, even a specific row)

however…

• Combinatorics, O(n2) can still eat you alive
• Going through these steps is a great way to ferret out and formally establish best practices!

Conclusion…

• When it comes to relational vs. graph, it’s not an “either/or,” rather, think “AND” (both can work together in harmony)
• Same goes for proprietary vs. open source
• Proprietary: rich collection of “out-of-the-box” features and functions; highly modularized, API-driven
• Open source: “DIY,” “sweat equity” intensive; highly customizable; practical solution for individual consultants/SMEs, small teams; good for prototyping, customization

Future approach: the way forward

• Add SUMO/KIF logic functionality to identify and reduce conflicting information, redundancy
• Incorporate:
• More quantitative performance metrics (precision vs. recall, etc.)
• Cyclomatic complexity metrics (improve performance by removing redundant paths, etc.
• Knowledge flow models
• Human sensemaking tools
• Semantic authoring

KMWorld Connect 2021 is going on this week, November 15-18, with workshops on Friday, November 19.

On-demand replays of sessions will be available for a limited time to registered attendees and many presenters—including Murray—are also making their slide decks available through the conference portal.

For more information about KMWorld Connect 2021, go to www.kmworld.com/conference/2021.

Access to session archives will be available on or about November 29, 2021, so be sure to check back for on-demand replays.