Hello everybody and welcome to today's video nugget.

After we've seen in the introduction to natural language processing that we have both pattern-based

methods, as well as language technologies that try to compute the meaning of natural language

utterances.

I would like to kind of just oppose meaning representations and language just to see what

we're up against and generally understand what the meaning of natural language might

be.

And it's actually a difficult question.

So we have a long history of trying to understand what it might be that language means.

We see, you may remember this from school, we have with Plato, we have the cave allegory

where they try to understand what that difference is.

We've seen with Aristotle, we have this idea of syllogisms, which became modern logic and

calculee in an effort to understand the meaning of natural language texts.

We have the discussion at the beginning of the last century where philosophers like Frege

and Russell and mathematicians maybe, that kind of try to understand the semantics of

words.

So if you will, you can contrast the meaning of the words Michael Kohlhase and the word

Odysseus.

A reasonable theory of meaning might be to say, well, Michael Kohlhase, that actually

refers to this object in the world, in this case, the very object that's giving a lecture

to you.

So that's the idea of a referent.

Whereas you can't really do the same thing when you hear the name Odysseus, right?

This is a guy, if anything, it's not an object in the world, not anymore.

Somebody who is probably, if he existed at all, is probably a made up figure.

It's somebody who is at least 3000 years dead.

So you can't kind of say that the semantics of the word Odysseus is by a referent in the

real world, but maybe a referent into a collective narrative or imagination.

Kind of that is considerably different.

So we have a lot of problems to solve when it comes to the meaning of natural language.

I just want to give you a couple of things that can go wrong if you don't have the meaning

of natural language.

Say you kind of have a relatively simple translation, you might translate the German as very metaphorical,

which is why I'm using it.

The ghost is willing, but the flesh is weak.

You translate it to English, you translate it back and then things go wrong because ghost

in German, which could be translated into spirit, but spirit would also can be translated

back and is in this translation from something spirit as in the soul goes back to alcoholic

spirits, which kind of gargles the meaning.

The same thing here, the German word fleisch, which just means flesh, which just basically

means in this sentence, my mind is willing, but my body is weak, just basically means

that you've probably not resisted some temptation that you at the cognitive level wanted to

resist.

It has nothing to do with eating and drinking.

So the meaning counts if you want to translate well.

So the answers of what semantics might be are different from the area you're coming

from.

So in psychology and cognition, semantics is just basically what is in our brains, what

is the neural activation in our brains and what are the mental models here.

In mathematics, we've seen that we basically use logic as the meaning of, in this case,