Direct Entity Dissection: How Google's Knowledge Graph Has Advanced
In the summer of 2019, I wrote a piece outlining how Google had made a giant leap forward in its ability to understand entities. Fast forward but a year and a half later and things have changed dramatically.
What I'd like to do is revisit some of the very same entities in order to see what's changed. I think it will be evident that Google's made some striking advancements in how it can pick an entity apart and break it down into smaller pieces.
Here's how Google can parse entities to a far greater degree than it used to.
How Google Used to Break Down Entities
Let's start with what once was. Back in 2019, what caught my eye was how Google became so adept at breaking down an entity category into finer and highly-relevant parts.
What do I mean?
Google became able to take one aspect of an entity and further refine it by breaking that aspect into multiple layers or subtopics.
To showcase this, I focused on sports entities (aka sports teams). Though not the only instance, this entity category reflected the most striking examples of Google's ability to understand things at the subtopical level.
Here's a good example from my post in 2019:
What you have here is Google breaking down the entity, the Pittsburgh Steelers, into various aspects, one of which is "players." Google then further refined this by breaking down "players" into the most relevant player position for the entity. In this case, it was the team's historically good collection of great Wide Receivers (the fellows who catch the ball).
At the time, I noted how amazing this was. Google understood that out of all the various positions on a football team, the wide receiver position was the most relevant to the Pittsburgh Steelers (which in this case is quite accurate).
The Limitation of This Entity Refinement
Time is a funny thing. Back in 2019, I was ecstatic to see this level of entity refinement. It was quite an advancement. Prior to this, for these entities, Google would show a list of players without any further entity refinement. Meaning, all you would get is a list of current players not a special list of the team's all-time great at a given position.
The limitation I didn't recognize at the time is that this sort of entity refinement doesn't take the entity head-on. What you had here was Google breaking down the entity into its obvious parts and then refining individual aspects. In the case of sports teams that meant further refining subcategories such as players as was done in the panel for the Steelers.
That's no small thing to do. What made it more amazing was that Google did a decent job of showing the most relevant subtopical breakdown.
Still, this was not Google tackling the entity head-on (no pun intended). This was not Google redefining what the most relevant topical and sub-topical breakdown of an entity was.
In other words, this was not Google redefining the most relevant way to dissect an entity. This was not Google understanding the most relevant and the most granular way to dissect an entity overall.
Again, this was Google taking one subtopic and refining it further. But what if Google could look at the entire entity and further refine it as a whole? Wouldn't that be something to take note of?
Direct Entity Dissection: Google's Way of Breaking the Entity Overall Into Smaller Pieces
This all brings us to the present because what we get now for the same sports entities is a bit different and a whole lot more advanced.
Have a look at what Google now shows for the Pittsburgh Steelers in the Knowledge Panel:
As Jason Barnard talked about in-depth, Google is now showing expandable Knowledge Panel tabs that operate almost like the People Also Ask (PAA) box does.
However, the tabs in the Knowledge Panel are far more than PAA questions. They are a direct and overall refinement of the entity at the sub-topical level.
Simply, they are subtopics that speak to the entity as a whole (as opposed to the further refinement of an existing subcategory). That's a major step forward.
Well, to better refine and compartmentalize the broader entity means you need to understand that entity better. It's "easier" to refine the entity from a single angle than it is to refine it as a whole.
But, Mordy, did Google refine the entity as a whole with the expandable Knowledge Panel tabs here?
Breaking Down Google's Advanced Knowledge Graph Entity Understanding
Let's dive into exactly how well Google is getting "into" an entity with the expandable Knowledge Panel tabs.
This is where my knowledge of SEO and sports collide!
Let's take a look that the tabs Google shows for the Steelers again and go through them one-by-one:
Records: That's a pretty generic tab. Google shows it often for sports teams as many of these franchises hold all sorts of records in their given sport. That speaks more to the category of entity than the specific entity per se.
Super Bowl Rings: This is a great example of Google understanding the entity as the Steelers are tied with the New England Patriots for having the most Super Bowl wins of all time. (As you'll see below, Google shows the same for the Dallas Cowboys, who also have had much success in the Super Bowl, but doesn't in cases where the team in question has not done as such.)
Coaches History: This is literally the tab that made me think Google took a big step forward in understanding how to refine entities as a whole. If we expand the tab we can see that since 1966 the Steelers have had just four head coaches.
That is unheard of.
The Steelers are famous for sticking with their head coaches for years upon years. If you don't believe me, here's what ESPN says on the matter:
It's literally one of the defining characteristics of this entity and it's a pretty nuanced one at that. Google getting this right and being aware of this very specific aspect of the entity needs to be appreciated.
Rivals: This too might be construed as a general entity focus (all teams have rivals). However, the Steelers are famous for their rivalry with the Baltimore Ravens:
It's not just here in the case of this specific entity. If you dive into the world of sports entities there are a plethora of examples showing how Google has advanced (though, some entities do lag behind). Often enough, Google shows tab headers in the Knowledge Panel that reflect an improved ability to parse an entity as a whole.
Here's what we get for the NY Yankees:
Again, the 'Rivals' tab is a bit generic as is the 'Symbol' tab. However, the 'Retired Numbers' tab is spot on. No team, perhaps in the history of sports, certainly not in the history of professional American sports, has retired more player numbers than the Yankees:
Having a tab that discusses the numbers of retired players reflects a defining attribute of the entity. The entity here is defined by its historic players and the sheer number of them. This is a tab that speaks to the entity's inner-identity.
Take the Dallas Cowboys, their tab on 'Worth' is a major part of their identity as well:
The team's owner, Jerry Jones, revolutionized the earning power of NFL teams. It's part of what got him into the Pro Football Hall of Fame and is well-known among football fans.
I could go on, but I think you can see the point I am trying to make. Google has moved beyond the smaller task of breaking down one aspect of an entity and is able to parse those things that define the entity as a whole.
In other words, Google is better able to parse and speak to the entity's overall identity. Not only can it speak to identity from a single vantage point, but it can determine those aspects that allude to the entity's core identity.
This is a major step forward in entity understanding.
More Specific Understanding Beyond Traditional Entities
It's not just about "entities" - at least not as we might think of entities. Google's ever-increasing ability to parse entities and to dig into them in all-new ways goes beyond teams, people, corporations, objects, etc. Rather, it extends to topics as well.
A topic, much like a physical object, is a thing... is an entity. Google's ability to better parse entities means it can better dissect a topic into finer parts.
Case in point, the subtopic rankings Googled launched in November 2020. This should highlight the practical implications of Google being better to parse entities.
This really speaks to a larger trend of Google trying to energetically offer users a heck of a lot more "specific" content and less "broadly reaching" content. Whether it's subtopical rankings, Passages, a better ability to determine relevancy, Google is showing users more highly-specific content than ever before. It's quite an important trend to take notice of.