In this video nugget, I would like to introduce the semantic web or the basic ideas of the

semantic web. Remember, we are in the chapter on knowledge representation and the idea is

can we actually make good models of the knowledge about some world here. And in this nugget,

we are going to look at a very particular world, namely the web itself. Remember, we

talked about knowledge representation and the semantic web is actually the idea of using

knowledge representation or semantic network ideas to talk about the knowledge that is

actually on the web. And by that, the idea is to of knowledge representation is to add

inference is that you want to go from the current web, which is kind of dominated by

unstructured or semi-structured documents, web pages, PDFs, web applications, all of

those kind of things. We want to go into a machine understandable web of data from which

machines, agents can actually draw conclusions. And basically, if you want to understand what

the semantic web wants, I think I would like to give you an example that kind of gives

you this, we have to add reasoning to the web. So say you are interested in who was

US President when Barack Obama was born. And with a little bit of Googling, you can find

out that Obama was born on August 4th, 1961 and then you would naturally think of ask

Google who was US President in 1961 and you might find out that it was either Dwight Eisenhower

or John F. Kennedy. And they basically changed over on January 20th. So now we basically

want to combine these two things and say well, August 4th is after January 20th when John

F. Kennedy was inaugurated. And we kind of still know that John F. Kennedy was still

alive on August 4th even though he had problems with staying alive later. And then something

interesting in this reasoning, that is what at the moment you as humans do, you come up

with the idea, yes, the answer to this question of who was US President when Barack Obama

was born is actually John F. Kennedy. Now, wouldn't it be nice if we could just basically

directly ask some kind of an agent this question and it would come back after a while and say

John F. Kennedy without us having to Google for information on the web and then kind of

putting it together. Other things are you might want to know which are even more interesting

is what's the number of COVID, new COVID infections over the holidays. And you might not actually

have over the holidays means there is, depending on how you define it, say a span of seven

days and then you might not have all the numbers in one place so you would have to kind of

get them all together and also know things like if you have a number for the new infections

in the EU and in Germany and in Spain, then you might want to not count EU plus Germany

plus Spain because Germany is part of the EU and stuff like that. So if you think about

advanced ways of handling data or information or knowledge, by the way, on the web you have

an awful lot of reasoning that you do but Google doesn't. Wouldn't it be nice if we

had a way to make the web into a web of data which is machine understandable so that machine

or agents can actually do what we are doing otherwise. That would save us a lot of work

and give us much more information at our fingertips and that's exactly what the semantic web wants.

Now there's a problem. We want to go from the current web to a web of data. We have

to kind of go all of those steps in this information ladder and to show you that all of these steps

are non-trivial, I would like to give you an example. Say we see a web page. Here's

one. It's kind of old but that doesn't matter. And if you actually look at it as a human

person, you can find out, yeah, there's an international conference. It's in a very nice

place. It's at a time where I may be actually free. There's a high reputation because people

like Tim Berners-Lee are speaking and so on. You kind of think about this and say, well,

wouldn't it be nice if I could go and the university would pay me for going to Hawaii

and I can go swimming because May in Hawaii is probably great and I would like to learn

to surf. So those are all the inference you draw as a human. Wouldn't it be nice if your

calendar agent could actually do the same thing and just put the date into your calendar,

book your or apply to the university for you to go and to fund you going to this conference

because it's actually relevant to what your work is and then actually booking the flight