September 13, 2013
Listen to our Director of Research, Marsal Gavaldà, explain the mechanisms behind Google’s Knowledge Graph, and learn how the idea is not as new as one might think.
Click here to watch Gavaldà describe the enormity of different kinds of knowledge graphs.
I wanted to talk about a technology trend that is revolutionizing how computers understand language. And that is the Knowledge Graph, which is very relevant to the work we do here at Expect Labs. You can visualize the Knowledge Graph as a giant network where the nodes are objects, such as a particular place, an individual person, a specific movie, and these nodes have properties such as the date they were founded, born, or released, and they also have links, from one node to another, that encode relationships such as “similar to”, or “is-a”, like a dog is a mammal is an animal, or “part-of,” such as steering wheel is part of a vehicle.
What’s fascinating is that the idea of a Knowledge Graph is not totally new. In fact, you can argue that it’s an extension of the taxonomies and ontologies that have existed since the dawn of civilization, with, say Aristotle’s attempt at categorizing natural phenomena over 2,000 years ago, or the French enciclopédistes like Diderot and d’Alembert who during the Enlightenment in the 18th century attempted to collect and write down an organized compendium of all human knowledge. However, what is new today is the scale of the Knowledge Graph, also the amount of detail, and the fact that is being created in an automated way.