Using Freebase to Fix Your Company's Google Knowledge Graph

“Who is the CEO of Fedex?”

“Starbucks Customer Service”

 “Parent Company of NBC”

Searches like that are now being answered by Google’s Answer Box and Knowledge Graph. These and other questions are part of Google’s efforts to return answers to questions, not just links to sites. Companies are among the “entities” (people, places, things) that Google is attempting to develop a set of knowledge around.

Of course, Google is not always correct with their answers. We’ve seen wrong phone numbers appearing in answer boxes and company description boxes (see below) associated with the wrong websites. So we have to know where Google’s answers come from in order to correct those mistakes.

One way Google is answering these questions is by using the data in Wikipedia and a lesser-known site called Freebase. Freebase (the website, not the other thing) is a database that Google purchased in 2010. For brands, Freebase contains corporate info in well-categorized format. Official website, phone numbers, names of executives, related companies, and much more are included.

Company info in Freebase appears to be populated in 2 ways:

Users log in, add and edit content, Wikipedia-style. This is the way that Freebase started, and the way I’ve updated info myself.

Google adds info directly from its own interpretation of crawling the brand’s site. This one is my conjecture, not something they’ve stated.

#1 is interesting and straightforward. If you find some incorrect or incomplete information about your company, it is worth your 5 minutes to fix.

#2 is a lot more interesting (not like, “Mom, I gotta tell you what I found out today!” interesting, but still interesting). If you know where to look, Google is giving us a bit of a behind the scenes in their interpretation of some site content:

For any large company, Google tells you what it thinks your contact page is, support page, and various phone numbers.

It may be the Google reviewed the link text to the page (“contact us”) and thought this page was likely to contain important organization phone numbers.They then mapped the phone numbers to the nearest logical text.

How do you see this information:

Its actually in a hidden on your company page in Freebase.  The entries are locked as well – not the editable versions that you normally see.

Using your Chrome/Firefox developer tools, unselect the “display none” nonsense and you’ll see a box like the one above appear.

Notice that this is all information found in the schema.org protocols for organizations that Google wants everyone to add to their site. In other words, if you don’t fill it out – Google is going to do their own guesswork.

So what can I do with this information?

Impress Your Friends

Fix incorrect answer boxes:

Update Freebase’s editable sections about your organization or another person, place, or thing you want to correct.

Looking at the hidden div, you can learn which page Google thinks is your “contact page.” Add schema tags to improve their understanding of that page, add other useful content to the page, or correct outdated info.

On pages with charts and tables of information, it helps to use old-school HTML tables, ordered lists, and definition tables as well (to easily show how the data should be read).

Freebase Closing Down

The call to action here is easy, Google announced recently that they are shutting down Freebase and basically just migrating the data to Wikidata. Its unclear exactly how that will affect updating your data, and even less likely that this hidden data trick will remain available.