Okay

welcome back.

I'm happy you made it in the cold.

We are talking about a field

a subfield of logic-based AI called knowledge representation.

The idea in this is to...

There's multiple ideas

actually.

The main idea is that we want to describe the world in terms of concepts.

That's like

in natural language

we have nouns.

Dogs and cats and automobiles and persons and photons and numbers and whatever.

We want to describe the world not only in terms of classes of objects

but also of their

their relationships amongst each other and also at the level of individuals

objects in the world.

One way of doing that we've looked at is...

Semantic networks.

Semantic networks are a very simple way of doing this.

But it already gives us some of the stuff we want.

One of the things it gives us is the notion of a network of concepts and a corresponding network of individuals.

That's an important concept idea.

We divide them into things that only talk about classes of objects

the terminology.

And parts of the ontology that talk about individuals and their properties and what classes they belong to

and what the relationships of individuals are next to each other.

You can think of this a little bit like a database.

In a database

you can have different classes of objects like lectures and persons

and you can have a relation between a particular person

say me

who is in the teacher's class

relation to AI-1 in the winter semester 25-26.

The terminology you can think of as the schema of the database.

Think of all of the edges as very simple topples.

Triples, actually.

And the A-box you can think of as the data

schema and data in this.

And we're going to look at that a little bit more.

We've looked at a particular way of starting to do this using logic because logic gives us much

much more inference.

And we have mechanisms for that.

And we've started out with propositional logic only that we've given it an ontological interpretation

where the propositional variables basically act as concepts.

Sorry.

And the nice thing about this is that we can still use the tools from propositional logic.

If we have a world description, we can actually see whether it's satisfiable,

whether there may be a world that actually is represented by this.

And the central idea in which this logic is better than the semantic networks is that we can actually make world descriptions