Knowledge Graph Advances From Google

Query Sequences and the Knowledge Graph

Recently we showed that Google is ahead of the competition when it comes to showing answer boxes in search. Google is not one to rest on its laurels, and they continuously work on improving their Knowledge Graph capabilities. In today’s post, I will explore a number of the more novel variants that they have implemented.

Stephen Colbert

Google recently became the target of an attack by Stephen Colbert when he noticed that Google was shoing his height as 5’10”. He went on air to mount a Satirical attack on Google, “threatening” to sue them if they did not adjust his height to 5’11”. He also specifically called out Larry Page. Google responded with an adjustment, as follows:

I thought this was hilarious! As you can see, Google made 3 changes to the result:

1. They adjusted Stephen Colbert’s Height to 5’10.5″
2. They included Larry Page in the related results bar – because he’s taller
3. They noted that Jon Stewart is shorter

While this shows that Google has a sense of humor, it actually also shows something much more significant: Google can, and will, manually tweak knowledge panel results. I don’t believe there has ever been a prior instance of this happening before. Since the beginning of Google they have steadfastly maintained that they do not have the ability to manually tweak their results, and here we can see it happening in living color. Note: there are cases where Google tweaks presentation of their results. Try the query “barrel roll” or “recursion” for a couple of fun examples of this.

To be fair, the prior statements from Google were focused on regular web search results, but here we can see that the knowledge panels are a different beast.

Also of interest, is what happens when you click on the Conan O’Brien link shown. When you click through on this you see a similar knowledge box, but above it you get a carousel showing the height of many other popular television personalities.

This carousel only comes up if you click through on this related result picture. If you go to Google and manually enter in the search “conan obrien height”, you don’t get it, as shonw here:

Another Carousel Example

This carousel behavior is not limited to the height of entertainment personalities. You can also trigger it as follows:

If you now follow this query by clicking on the burj khalifa link:

As with the Conan O’Brien example we showed earlier, this does not come up if you manually search on “burj kahlifa height”:

Movies and Menus

Another interesting example is show if you search on your local movie theater by name. Here is an example showing the local theater here in Framingham Massachusetts:

Notice how you get the entire list of movies currently showing along with date tabs on the top. The only thing that is a little disappointing here is that you don’t get the number for the movie theater right there with it, which would make a ton of sense for Google to add there.

However, if you click on “The Book of Life” you also get a further interesting result.

This shows information on where the movie is playing in your area. This result also comes up if you simple search on “the book of life” in Google directly.

This new type of result comes up if you search on the name of a local restaurant

These types of results are fantastic for local businesses, however, as I noted before, would be greatly enhanced if the phone number was alos included.

Query Sequences

One last sequence to show for today, which is an extended query sequence. This shows how Google can maintain the context of a conversation, and also how it can understand the context of a response. First, we start by asking the height ot the empire state building:

Then we request “pictures”

Next, we ask for the name of the person who built it:

While we are there, we might want to eat something:

Throughout this sequence we see that I mentioned the Empire State Building only once, so Google remembered that context during “the dialog”. This concept of a dialog is an important one.

Note that I have also tested query sequences using sports stars, such as Tom Brady. You can easily test a sequence for yourself using Tom Brady, his spouse, children, what football team he plays for, when his next game is, etc.

Summary

Google is continuing to find ways to extract structured data from web sites and present that information in the SERPs. In this post, I have shown four important areas of exploration:

1. The movie and menu examples above show how they are extracting structured data from web sites. Thexe examples bring significant value to the business owners.
2. The query sequences involving height show how they are mining related search query data to see what users might request next. Then, if you do select a second person or building, Google recognizes the possibility that you want a lot of related data and shows you a carousel.
3. The empire state building example shows how Google can maintain conversational context in 2 ways.
4. Lastly, with my first example, we see how Google can actually go in and manually tweak the data set when theh choose. Presumably, they have this in place for certain classes of facts, so that they can correct factual errors in the structured data sources they use (although they used it for humor here)

You can expect these types of advances to keep on coming!