In the early days of search, results relied heavily on the matching of words. If someone was looking for information about “hero sandwiches”, then the resulting pages would most likely have a fairly high concentration of those words on the page. The trick with this is that this type of sandwich is called different things depending upon where you come from. They might be known as “hoagies” or “subs” or, here in New England, “grinders.” In some places they are simply called “Italian Sandwiches.”
This effect created many challenges for web site content creators who wanted to reach a large audience. We would either need to sprinkle each variation into the page which would then make our original keyword less relevant or we could create individual pages targeted at each variation. The problem with that technique is that it ends up creating multiple pages that are saying roughly the same time but with different words. Which of these pages is actually more relevant to what the person is looking for?
Enter the Search Entities
Around 2011 and the Google Panda Update, we started to first see “entities” coming into play. In more recent years the concept has grown and expanded considerably. Unlike keywords, these entities can be framed and have elements that can be understood. Google’s “submarine sandwich” entity, shown in the image, contains all sorts of facts and data about it – including the knowledge that a hoagie, hero, grinder, and Italian sandwich are roughly equivalent to each other in meaning. Once a semantic frame around an entity is created, Google can now present information to us which doesn’t even necessarily contain the keywords being searched for. If I create an article about the best grinders in Connecticut, someone may also be able to find that same article by searching for the best hero sandwiches in CT, as well.
Based upon all the articles about these types of sandwiches around the web, Google can build a knowledge graph around the subject which can then be displayed or used to help finding relevant resources to show in the results.
Affecting Entities: Creation, Expansion, and Influence
Many entities already exist, but others do not. Some need to be expanded so that Google’s understanding of the entity is more complete. And in other situations, we might want to influence Google’s understanding of an entity to alter what it means. For the examples below, let’s say that I just opened up a new Sandwich Shop called Sammies By Stock.
- Entity Creation: My first step is to begin trying to create an entity “Sammies By Stock” which contains the information about my business – where I’m located, the hours I’m open, and of course the fact that I sell sandwiches. I might also have invented a new kind of sandwich called the Wichburg – so we’ll want to create content which describes exactly what a Wichburg is so that Google can create an entity around that and then connect it to the fact that a Wichburg can be purchased at Sammies By Stock.
- Entity Expansion: We also want to make sure we’re expanding existing entities to include us. The “submarine sandwich” entity needs to grow to include the fact that my Sammies By Stock entity is a place where you can purchase one of these if you happen to be nearby. We might also have to expand Google’s knowledge of the word “sandwiches” to let it know that “sammies” are just another word for the same thing.
- Entity Influence: Sometimes we want to actually change Google’s understanding of an entity. For example, lets say that there are a bunch of reviews and web sites talking about our Wichburg Sandwich in a not so positive way. These sites can then give Google the notion that Wichburg Sandwiches really suck. My challenge, at that point, is to come up with a strategy to overcome that negative association and influence it to be a positive one. It may not be an easy task (especially if the sandwich were to actually suck), but it is a goal that might be worth pursuing.
Providing Information To Machines
Quite a few articles on this subject talk about AI or Artificial Intelligence as the way Google learns all these things. As an avid sci-fi fan, I like to reserve that name for Terminators and computers named Hal. We’re quite a long way from that sort of AI, but we do have Machine Learning. It can still be quite a challenge to do these things – and it also takes a lot of time and patience. As we talked about quite a bit over the past few weeks regarding Google’s June 2019 Core Update – one can’t just make a claim about something and hope that Google will consider it to be true. Often we need to cite sources, and make sure that others are out there confirming those claims. In other cases, though, we can do things to help the machines understand what we’re saying without too much trouble.
- Structured Data: By structuring information in a specific way as outlined by schema.org, a machine can more easily understand what we’re talking about. In the pictured example here, we see a structured data recipe for a Hero Sandwich as seen on Genius Kitchen. We have structured data templates for just about any sort of entity you can imagine, from products to people to articles and even things like medical conditions and procedures.
- NLP (Natural Language Processing): Computers can start to make sense of some of the things we use in natural language. This can be done through context clouds and co-occurance, as well as things like semantic triples. We’ll get into more of this in future installments.
- Links: Seeing how documents link to each other within the context of the content itself will always be a huge factor in helping computers see how things fit together. Linking out or getting links from other pages which are giving additional context to what we’re saying helps Google to validate claims and to be more certain that what we’re saying is accurate.
Are Keywords Dead?
Sort of – but not really. I try not to think in terms of keywords nowadays, but in the end, we do need language to communicate and using the right words to describe things are always going to be important. The big growth here isn’t about killing off keywords but in understanding that many words (in the right context) have other words which can be used to describe the same thing. Of course, keywords are also still important when it comes to paid search because, after all, we’re bidding on them.
In this day and age we need to look out beyond the words themselves and understand things like intent (what the searcher is actually looking for), consistency, and context. A good marketing team is going to be aware of the latest developments and help you create at web site that not only matches you with users searching for keywords, but matches you with potential new customers and those genuinely interested in what you have to offer or share.
If, on a scale of 1 to 10, keywords are a ten, then entities go to eleven.