19 - Logic-Based Natural Language Semantics (WS 23/24) [ID:51103]
50 von 1703 angezeigt

So that's all for now.

See you next time.

Okay, so back to theory.

So I've been doing some Tableau machine implementation on the lab.

And I would like to kind of get to the next fragment so I want to remind you of fragment three.

So if you remember, we had complex verb phrases so complex utterances that behave like a verb.

We had verb phrases.

Howled and screamed, kicked the dog and poisoned the cat, kicked the cat and laughed, kicked and scratched.

So that's the kind of complexity we're adding with and and or and so on.

And the other kind of complexity was for verb phrases was given by negation, with the kind of interesting wrinkle that

negation via didn't

behave differently than the kind of sentence level negation, which we tried with

This is not the case.

And in particular, we looked at didn't laugh.

The solution to this problem was that we have infinite verbs, infinite verbs, sorry, finite verbs, the finite verbs

The ones that have tense, whereas the infinitive verbs, I kind of think the stem is more.

You know all this from probably your English classes in school, where when you looked up a verb, the infinitive is

In the lexicon.

With all the inflections.

Right.

The solution to this was that we use lambdas. Verb phrases were always of the form essentially lambda x

And what was it?

How?

Something like that.

The good thing about this

was

Once we shut down to the sentence level, we apply this to say,

And all the lambdas go away.

In a way, the lambdas are kind of a functional programming trick so that we can actually

give VP a compositional meaning.

So rather mild use of lambdas.

Today,

And the good thing about this is that if the lambdas go away, we don't have to worry about them in our grammar.

I want to

Go back even further.

Wow.

Probably not worth it. Let me go back.

In fragment two,

We went to

We went to introduce variables for pronouns.

Before we start on fragment four,

I would like to kind of let you in on a dirty little secret.

There's a reason we never did fragment two in MMT.

The reason is because we can't.

If you think about MMT,

Then we have two kinds of atomic identifiers.

We can say something like,

Peter is an iota.

We can declare a type for individuals, and then we can make Peter an individual.

Nice, Sandy, no problem.

Zugänglich über

Offener Zugang

Dauer

01:30:07 Min

Aufnahmedatum

2023-12-14

Hochgeladen am

2023-12-14 19:16:04

Sprache

en-US

Einbetten
Wordpress FAU Plugin
iFrame
Teilen