If you’ve been in the SEO business, you’ve probably heard the term “knowledge graph”. But if you try to use Google to understand what one is, it’ll make your head spin unless you have a math or a computer science background. We’re going to unpack this term in a way you can understand and talk about why this concept is so important these days.
Google’s Confusion Of The Term
First, we have to clear up something. There are two ways that the term “knowledge graph” is used. We’ll be explaining it in the generic sense, but Google also has its own Knowledge Graph, which we’ll differentiate through capital letters.
Google’s Knowledge Graph is what creates those useful boxes of information to the side or just under your search results. It’s also what returns the answers when you ask a question to your Google smart speaker. It’s a giant database of facts and the connections between them that Google leverages to create answers to questions.
Only the most highly-trusted facts go into there and it’s unknown how much the boxes drive traffic. Google has been tight-lipped about how it gathers and categorizes information for its Knowledge Graph. It’s become so trusted that Wikipedia has noticed a big dip in how many people use their service.
What Is A Graph?
What, then, is a generic knowledge graph? To answer that, let’s break it apart. A graph, in this case, is not the pretty pictures you make in Excel or see on a dashboard, though there is some relation. In computer science, a graph contains two kinds of information:
- Data points (vertices)
- Connections between the points (edges)
Google Maps is an excellent example of this kind of graph. Each destination on your trip is a data point. Google also has a list of all the possible ways (the edges) you can get between each point. When you do a search for directions, Google calculates the quickest route through the graph between all the points, then tells you how to get there.
Google’s Knowledge Graph works in a similar way. When you ask a question, Google splits up your query into keywords, matches them to points in the graph, and traces the connections between them to arrive at the answer. However, to understand this we have to understand knowledge.
What Is Knowledge?
Heavy question! Here’s what we mean in this context. Knowledge can be divided into two kinds. The first is things. There’s probably a more technical term, but things will work. Apples are things. Barack Obama is a thing. AI is a thing.
The second thing is called meaning (semantic information). This is a kind of information about a thing. For instance, apples have colors including green, red, and yellow. Barack Obama was the 44th President of the United States. AI stands for artificial intelligence. In short, semantics tells us what a thing means.
If you’ve read our previous articles on Schema markup, it’s the same thing. Schema is intended to give semantic information to search engines. If I write “123 Sesame Street” into a page, Google has to guess the meaning. Is it a real address, or am I making something up? However, if I mark that section of text in my code using Schema as an address, we are telling search engines that “123 Sesame Street” is, indeed, an address and I want it to mean that.
What Is A Knowledge Graph?
Now we can put them together. A knowledge graph is a graph where the points are made up of things and are connected through semantic information. The thing “apples” is connected to the things “red”, “green”, and “yellow” through the meaning of “color”, and so forth.
Another term you might see for a knowledge graph is an ontology. Those of you who took philosophy know that ontology is the study of “what is”. As a noun, it is a complete formal knowledge graph about a particular subject or domain of discourse (say, business).
Isn’t This Just A Knowledge Base Or A Wiki?
No. A knowledge base might have all of the things, but it lacks the formal semantic information about the links between the things. You can link anything to anything else arbitrarily inside a knowledge base. While you can add semantic information indirectly, say by putting links in a table of contents or in a related subjects section, a knowledge graph has meanings about the links baked into it.
This is why you can search “When was McDonald’s founded” in Google and get the answer (April 15, 1955), but you can’t do the same thing in Wikipedia even though Wikipedia has that information. Google can take the thing “McDonald’s” and the semantic query “founded” and follow the link to “April 15, 1955”. Wikipedia only has the thing. You then have to read the page to get the information.
Smart And Flexible
The really cool thing about knowledge graphs is that you can create new meanings and things on the fly and the graph will adjust itself automatically. When your GPS app knows that a road is closed or blocked, it can recalculate a new route. When Google finds a new piece of information or a more accurate meaning, it can weigh it and add it to its Knowledge Graph.
Similarly, your business could use a knowledge graph to do things like create a more intelligent chatbot, add better search capabilities to your website, build recommendation engines, and create better content management systems.
Even if you have no interest in building your own knowledge graph for your business, knowing about how they work helps you understand why new SEO features like Schema are so important. We’re all quite used to putting content on our site for Google to find and letting the search engine figure out what it means. Schema lets us tell search engines what our content means.
Now that voice search is growing in a big way, providing semantic meaning along with our information is vital. We can’t afford to have Google guess what our content means for us, as smart as it is. If people are relying on Google’s Knowledge Graph for their factoids, we need to be proactive in giving Google correct information.