The Knowledge Panel Course: Educating the Child That Is Google
Script from the lesson The Knowledge Panel Course
Jason Barnard speaking: Hi, and welcome. Google is a child thirsty for knowledge, and you need to educate it. In the introduction to this course, I talked about this idea very briefly. We have an entire lesson about this because the concept is so fundamental to Knowledge Panels. Google wants to understand the entire world, and that’s a massive task.
Jason Barnard speaking: Google has built an algorithm that is capable of using the information it finds on the world wide web to learn about the world. And it’s increasingly able to learn on its own. Of course, it started with human curated databases, including Wikipedia. Google fed the algorithm with information from Wikipedia, Freebase, Wikidata, IMDb, MusicBrainz, and other sources that Google’s engineers considered to be reliable.
Jason Barnard speaking: The machine learning algorithms have been trained on this reliable human curated data. In recent years, the Google engineers have tested the child algorithms capacity to deduce facts from the information it finds on the web. The simple trick has been to compare the information it deduces from the world wide web to the information in Freebase, Wikipedia, and Wikidata.
Jason Barnard speaking: Since 2020, the algorithms have become increasingly reliable. And little by little, the child is learning to learn from the fragmented information it finds on the web. And that is huge. So, the algorithm increasingly learns on its own, but that doesn’t mean to say we should allow it free reign. We can help it learn by actively educating it. We can teach Google. Now, that is super huge.
Jason Barnard speaking: Basically, there are just three simple steps.
Jason Barnard speaking: Number one is the Entity Home. It’s the web page that Google uses as a reference for information about the entity from that entity, your version of the facts, if you like. Number two, it looks at corroboration.
Jason Barnard speaking: It wants to be sure that you are telling the truth and that your version is an accurate representation. To do that, it compares the information you provided on the Entity Home to all the information around the web.
Jason Barnard speaking: Number three, you need to guide the child. Ideally, it will start at the Entity Home, see your explanation, then you will link to the other corroborative sources that confirm that explanation and link back to the Entity Home from those. The child is then in an eternal cycle of self corroboration. It might seem strange to you that Google is looking actively for an Entity Home you own. Google calls this the point of reconciliation. The child algorithm wants your version of the facts in order to be able to reconcile the fragmented information it finds around the web into a version that makes sense and is reliable.
Jason Barnard speaking: Since everybody is not notable enough to have a Wikipedia article or Wikidata page, Google cannot rely on those over time. It can’t rely on IMDb or MusicBrainz either. Even if these platforms would like to be a complete source of information about their topics, they cannot be exhaustive since human editors cannot cover the entire topic exhaustively. So if it is to understand the entire world, Google has to build that understanding by finding the Entity Home for every entity and cross-check the information provided there with authoritative sources, simple.
Jason Barnard speaking: We can look at it this way with an analogy that really makes sense. Google is a child that wants to understand everything. In this scenario, you are the parent teaching the child about your little corner of the world. The child looks to the parent for a clear explanation about a fact. The parent explains to the child in simple terms the explanation you provide on the Entity Home, so the child has a clear understanding of the fact about you, but that’s of course not enough.
Jason Barnard speaking: The child has other influences, siblings, school teachers, grandparents, friends. Hopefully, these people confirm the facts you have provided and that corroboration supports the child’s learning and builds confidence in the facts in the child’s mind.
Jason Barnard speaking: So, your role as a parent is to explain on the Entity Home and then indicate which sources the child should look to for corroboration. It’s your job to push the child algorithm to the reliable, relevant, and authoritative sources that confirm the facts.
Jason Barnard speaking: In our analogy, if the parent is teaching the child about history, then you would want to push the child towards the history teacher at school since that person is expert, authoritative, and trustworthy. You might also want to push the child towards a trusted older sibling or a grandparent for corroboration. The child could also ask the baker, but that baker has very little expertise or authority on the topic of history compared to the history teacher at school. That said, if the topic was how to bake a cake, the baker would have more authority than the history teacher.
Jason Barnard speaking: So, the key to educating Google is that you identify the corroborative sources that have topical authority, expertise, and trustworthiness. Obviously, this number is not absolute, but the child is looking for about 30 corroborations from trusted, authoritative sources. Close family is great, but preferably a relevant, authoritative, and trusted expert, as in the case of the history lesson from the history teacher or the cake baking lesson from the baker.
Jason Barnard speaking: After those repetitions, assuming all of these different people repeat more or less the same thing, in more or less the same words, the child will not only understand, but it will be incredibly confident in its understanding. But the child still always needs that go-to source for when the information changes or when it has a doubt. And that source is always the parent. That is the Entity Home, the point of reconciliation. It needs that point of reference from you.
Jason Barnard speaking: When you say on the Entity Home the facts are ABC, if the corroboration around the world wide web says the information is indeed ABC, then the child will understand and be confident in its understanding and trigger a Knowledge Panel, which is what we want from this course. However, if you say on the Entity Home that the facts are ABC and the information around the web are sometimes CDEF and sometimes says BFD, and then the child would be confused. It won’t fully understand. It might believe the parent, but it certainly won’t be confident in any of its understanding if everybody else is contradicting you.
Jason Barnard speaking: The huge question here is what corroboration will this child trust and believe for your entity? You probably immediately jump to the big hitters, Wikipedia, the New York Times, the BBC. But coming back to our example with history, the History Channel is a great source. For baking cakes, sallysbakingaddiction.com is a super helpful source. So for corroboration, you need to look for sources that are expert, authoritative, and trustworthy within your specialisation, your geo region, and for your entity type.
Jason Barnard speaking: At Kalicube, we call this Entity Equivalents. So if I want to feed Google factual information about a person in the UK in gardening, I would need to look where Google is getting its information about other people in the UK in the gardening industry. If I want to feed Google factual information about a company in the US pet food industry, I would need to look where Google is getting its information about other companies in the US in the pet food industry.
Jason Barnard speaking: Now, this seems obvious, but the trick is to find those sources Google is looking at for that combination of entity type, geo region, and industry. And that is what the Kalicube Pro SaaS platform is designed for. That said, there is no need to pay for Kalicube Pro. If you know your industry, you probably instinctively already know which the power sources are that you need to focus on.
Jason Barnard speaking: For a gardener in Leeds in the UK, the gardeners’ association of the UK, the local newspaper, garden focused shopping guides in the UK, local fetes and events. In short, the trick here is to find where Google is getting the information for its Knowledge Graph and Knowledge Panels for your entity type within your industry and in your geo region.
Jason Barnard speaking: Google’s algorithms are very, very specific and hyperfocused. An expert, authoritative, and trustworthy Entity Equivalent sources will often be more effective than the traditional mass media sources, such as Wikipedia, the New York Times, and so on. That is Kalicube’s Entity Equivalent trick, and it works every single time.
Jason Barnard speaking: Now, once you have these sources of corroboration, you need to join the dots. And joining the dots can take multiple forms, including Schema Markup, simple hyperlinks, and semantic triples. I’ll explain each of these in the following lessons.
Jason Barnard speaking: Schema Markup is geeky, links are semi-geeky, and semantic triples are as simple as pie. You can use them all or just those you’re comfortable with. Applying all three will be far more effective than just using one or two. But if you aren’t technically minded, you can still educate Google and make this work for you.
Jason Barnard speaking: Always bear in mind, Google is a child and it simply wants to understand. It’s up to you to educate Google about yourself so that it understands and is sufficiently confident in its understanding to give you a Knowledge Panel.
Jason Barnard speaking: Thank you, and I’ll see you soon.
