Mike Dillinger, PhD was Technical Lead for LinkedIn’s Knowledge Graph initiatives and founding Manager of the Taxonomy and Crowdsourcing teams. He also started the Machine Translation programs at LinkedIn and eBay and worked both as an independent consultant focusing on how to deploy AI systems and as a researcher in knowledge representation.
Talks and Events
2022 Talk: Making Friends with the Elephant in the Room: Validating Knowledge Graph Quality
In an environment where we created web-scale knowledge graphs that drove almost all of a huge company’s products for hundreds of millions of users, we were constantly at a loss to answer questions from top management like: how good is our knowledge graph? Is it better than what company Y has? How much of the whole domain does it cover? Which parts of the knowledge graph are least reliable? How good is the knowledge graph for this or that industry? What do you need to be able to answer these questions? This presentation describes the key blockers for validating knowledge graphs at scale and describes new methods for approaching this goal.