2 - Logic-Based Natural Languate Semantics (LBS WS2024/25) [ID:55062]
50 von 1064 angezeigt

Record. If you don't want to be recorded, mostly we can hear you from time to time.

Then better leave now if you really don't want to be recorded. So anybody who stays

to be maybe hearable at some point. Righto. And I'll share the slides. Good. We're almost good to go.

So you've experienced the dry run for one of these prep quizzes. These quizzes we use to

give you bonus points and keep you on your toes with preparing for the lectures.

And they're on time. I can extend them, but I usually won't.

You should know the URI of this. I apologize for not having a button for this yet.

That's something to do with the build problems we're experiencing.

Currently we might have a first understanding why this error is there. So I hope things get better.

And the quizzes will be basically like that. Multiple choice questions,

and so on. And we can look at the results. So what I've been showing is essentially

this bar graph here where we kind of see how many problems have been attempted by how many students.

The high bar is on the left. And in the end it's supposed to look a bit like this.

The stragglers here are all the latecomers. Okay. Right. These are the interactions per second.

How much has been attempted? And what are the results? Red is the

the all fail. Then there's a partially correct sometimes. And the green is the correct answers.

This test does not count to any bonus points. The pretest is giving me an understanding where you

know what should my prerequisites be? Where should I slow up? And apparently ambiguity in

natural language is something that is not a good prerequisite. It's actually not a prerequisite.

We're going to talk at length about these things. Whereas other things like part of speech tagging

is something many of you have some acquaintance with or are good at guessing. And so I'll basically

look at that. Okay. So back to the course. Are there any questions?

Yes. Do we get the results for our test? Yes. You can look at the results for your tests.

I think I always do it at 11. So that's I have your attention now. Maybe if we can actually get this

the kind of from 10 to 10 roster going, I might just basically make them available

immediately. Now it's not that way. We can negotiate about anything.

Okay. It's just that if we lose the first half hour of the lecture because you're kind of fizzily

discussing everything and only thinking about why this is true and not and so on. And if you have

doubts about our grading, which can be reasonable because you can imagine that we're always

scrambling to get the quiz questions ready at the last minute and errors have been there before.

And so the matrix channel is probably the best place and then we will kind of retroactively

do the necessary thing. And in the next year, this course will have much nicer quiz questions

and we can use them. You can use them for practice and those kinds of things.

Okay. Any more questions?

And if you don't like the ALEA system, it's a research prototype. Come and help us improve it.

Yes. Will the quizzes be available later? Like can we use it to revise later? Yes. We're going to

basically make the quiz problems public and then they will basically appear in the right place

after the section first. If we get the build problems under control.

Okay. So in this course, just very quick recap. I'm going to use this picture for that.

So the recap is we want to go from natural language utterances,

essentially bases of spoken or written language, into a semantic representation

that we could do something with. The typical thing is if you think about a

system, an information system at the Numbagh airport, you pick up the phone

and ask it when is the next plane to Mumbai. And then the system will translate this somehow.

And then we'll make a SQL query out of that typically and then go the pipeline back,

essentially from the answer of the inquiry into natural language. And then it says, yes, tomorrow

at 7.30. Oh, that is too early for me. Already what you see, oh, that's too early for me. It's a

relatively under specified sentence. Something like it starts at 11, what we talked about

last time. And we looked at basically, we're taking the utterance, we do a syntactical analysis,

how do the pieces fit together? What's the subject, the predicate, the object? Is there anything else?

Zugänglich über

Offener Zugang

Dauer

01:39:52 Min

Aufnahmedatum

2024-10-23

Hochgeladen am

2024-10-23 13:46:05

Sprache

en-US

Tags

language computational logic
Einbetten
Wordpress FAU Plugin
iFrame
Teilen