I'm switching to the...

Did I unblur my background?

Yes.

Yes.

Good.

So, we should be all set up, almost.

Okay, so we talked about in the last lab session, we talked about the grammatical framework.

Just a very quick reminder.

We have the setup in GF where we have abstract grammars for specifying which abstract syntax trees are allowed.

Essentially, in our pipeline, let me show you the pipeline here.

What we want to do is we want to start from natural language utterances.

Think of text strings.

That's what we're starting with.

We're not doing speech.

We create syntax trees, something like sentence, NP and VP and all of those kind of things.

Then we do semantics to instruction, which gives us a logic-like screen.

I'm sorry, can you share the screen online?

Pardon me?

Hello?

Hi, sorry, can you share the screen, please?

We can't see anything online.

Yes, absolutely.

That's what I planned, but didn't do.

Sorry.

That helps, I hope.

So we do semantics constructions into a logical expression, which we then in the pragmatic analysis do unspeakable things to make it complete by copying stuff.

The context in there, resolving, anaphora and all of those kind of things, that's kind of for a later place.

What we're really doing is this part now, but I would like to remind you that we are somehow looking at logic expressions as well.

That's what we want, eventually.

I need this idea.

Because I'm going to come to different styles of grammars.

And you've played with that in your homeworks, I hope.

And have had your successes and your failures.

The homework, as I said before, has been set up that you'll be partially successful and partially failing.

That's perfectly OK, because I want you to have thought about the problems first so that we can now see that you can appreciate the solutions that GF gives you.

We've done things like that.

We have abstract grammars.

Basically, we had categories.

They're slightly different than the ones we used.

We had transitive verbs, verbs that take two arguments.

We had noun phrases and we had only proper noun sentences.

We had lexical rules that basically classify certain...

And now it becomes interesting.

Not words, but syntactic concepts.

If you compare this with that, then really we have constants in a sentence.

We have a representation language here, whereas we have strings there.

These constants, Joan, Mary and love, are actually those constants here.

Why do we want to do that?

Well, the concept of Peter or the concept of love is something that is a syntactical constant, independent of the language.

This is the language-independent love, whereas this string is the language-dependent love.