h2g2 The Hitchhiker's Guide to the Galaxy: Earth Edition

Hello [joanna]. Do you make cheddar donuts?
How's everything going?

Hi!

*hands over cheddar donut*

Things are going fine with me... I'm just off for a meeting for my coursework - a symptom of general busyness, but term ends on Friday and I can't wait...

How're you?

Very well indeed! I'm doing an awful lot of research for my master's paper and getting lost in it, which is quite exciting . And I'm getting late to a boring course on African languages by staying over here. (actually the subvject's interesting but the lecturer speaks so slowly that you could solve half of the world's problem before she finished her sentence. I'm also passively looking for a job, which means it doesn't really work. I may have to become active in that field some day, otherwise my banker's going to kill me!

*savours *

Oops, that reminds me I should have a lunch before I go.

I was just walking up the beach when I smelled fresh doughnuts -- got any apple spice ones? I'd like half a dozen to take back to the atelier, please.

Typolifi, I think the stuff you're studying, semiology and all, is absolutely fascinating. What Joanna studies I don't understand at all. I just depend on it to work when I'm driving or baking or watching the stars.

Yeah, it's a good thing physics work when you're watching the stars. It somehow makes me feel reassured.
Actually I wanted to be a physician when I was a kid, but I wasn't working enough. Later I started learning more and more about sciences, especially about mathematics, but I was already on literary studies and now I end studying the mathematical language from a linguistic point of view. I seem to progress in a kind of 'don't know where you belong' way....

I get pretty talkative when I speak about myself.

*hands over a box of donuts to Lil*

Typolifi, you wouldn't by any chance have any experience with parsing sentences? My dissertation project is on building a parsing tool for linguistics teachers/students...

Lil, do you have as much suspicion of computer science as astrophysics? Even the s got too much for me after a while, and I switched to something, er, easier.

Parsing sentences? You mean, gramatically? That's what I do all the time, actually... But there's the methodologic issue. You wouldn't believe how much human sciences are split into schools and doctrines. I suppose you parse them the Chomsky 'government and binding' way, whereas we do it the descriptive functionnalist way. It's ridiculous, I know.

Anyway, I do also know some of the other ways, and would be glad to help. Only I'm going on holiday tomorrow and won't be back until next week.

Well I haven't really started it yet, and I haven't written any code or technically parsed anything, but I'm collecting linguists to potentially ask their opinions possibly at some later (or earlier) stage.

I think that my project will basically split into two ideas: one is a parser, which can be programmed with a grammar, so that students could say design a grammar which would be able to process certain sentences, and the other a derivation tool, so that one could step through the derivation of a sentence from a grammar.

I'm going to be using a categorial grammar, I think (not too hot on the terminology yet) I daresay you're right about my doing it the 'government and binding way', but I don't know yet as I don't know enough about the area yet. I do know that there's a different way of doing things in practically every single book I've got out of the library so far, and of course I am doing this from a kind of computational linguistics point of view; I think this teaching tool would probably be a starting point for students who wanted to know more about natural language processing, possibly, and computing students who study formal languages, like programming languages.

Wow, look at me go when I get on to a pet subject

Basically, I see two major categories of grammars (or syntactical theories, as we like to call them in order to avoidthe confusion with traditional primary school grammars). Either you assume there are sentences, and you describe rules that allow you to transform the inner form of this sentence to match a concrete one, either you analyse the words first to categorise them into classes. Each class has its embedded rules of combination with other classes and by doing whatever you want within these rules you get a sentence. It's more 'object-oriented' as it were.
From what you say about your derivation tool, I'll suppose it's the former category of grammars, ie the transformational grammars, that you are going to use. It's a good choice in Anglo-Saxon countries because that's what teachers and students will expect. They actually seldom know about the other kind. Anyway I don't think it changes much the programmation bit. Indeed, both ways you have to create objects corresponding to the 'end-of-transformation' classes (the letters you have on the bottom of a tree), with a list of actual words attached to it if you want your derivation tool to generate real sentences.
Then for the transformational rules themselves, I think it's possible to see them as kinda bijective operations, so that you need only one set of routines you can process both ways depending if you're in the parser or in the derivation module. But you can expect to have problems with the splitting of the sentences into groups corresponding to the output of the transformational rule (operation). It's called the ambiguity issue (A single sentence can be seen as the output of two or more different transformational procedures, that may result in differnt meanings). Chomskyan linguists have tried to take on this issue a fairly strict point of view, so you may find books with detailed studies of each ambiguity situation. This would be a good start for the laws of an algorithm you'd have to write for the sequencing of the sentence into groups.
Anyway, these ambiguities are not *that* frequent and usually the groups won't be too hard to split. Once you have them, you only apply the transformational rules backwards (from output to input, that why they should be bijective - if that's possible to program) to get one level higher.

PS: Perhaps if ambiguity's still an issue your parser could come out with the different possibilities. Now that I think of it I think it would be an asset. Teachers could tell their students 'you see, ambiguity's really an issue, even the computer can't cope with it'.

Pet subject too...

And also about the students being able to feed their rules in. I forgot it in my previous post, but this implies a third module in the program which would transform the rules they give (in a formalised way you'd have to specify, because the formalisation may vary on details from school to school and you have to adopt only one) into those bijective strings of operation I spoke off. This third module would come out with a library of routines the other two would use.


I don't know if all that seems to make sense, because in both these fields (transformational grammars and programming) I know stuff without being a totally reliable...

I ate this whole box of doughnuts while I was listening to you both. May I have another half-dozen, please?

I did some course credits on linguistic logic when I was at university back in the mid-70's, when transformational syntax was a sort of new thing, so this is all most interesting.

Good thing about virtual s is that the calories are all virtual too...

I was thinking about representing rules and lexical entries in xml. The idea is that the parser should be web-based, so it can be used by a whole class of students, and be programmed with exercises for them to do by a lecturer. But this is problematic - what technology to use? It has to be written in java, and it would be possible to use a servlet, or maybe an applet. But these have their problems because for example you cannot read files into an applet, and I don't know how easily one could draw trees with a servlet. I think though that if trees could be output in xml then one could try using xslt to display them.

it probably wouldn't be too hard for people to learn to write their own xml for lexicon/rules input, because of course all the markup would be meaningful.

Alternatively, one could design some form of input using boxes or something, which one could input the data into and would spit out xml at the other end to feed to the parser.

Meanwhile, my tutor suggests that I first parse strings by themselves, and then generalise the routine to parse more complex objects.

*hands lil more s*

Mmmmm doughnuts... and hmmmm... programming. I am no specialist in this field, but I suppose a structure translating the data into xml would be more readily used, whereas you can't expect the students, who'll probably have many things to actually do with the sentences after they're split, to spend time learning how to feed the parser.
But the general idea of a java applet seems very practical.

I don't really understand the notion of 'parsing strings' being simpler than something else. What's that more complicated thing to parse, actually?

*Allows himself to take another cheddar doughnut, and takes a tomato out of his pocket to slice over the cheese. A little salt, ready!*

I'd love a donut, J'au-æmne Please may I have...2

I'd love a donut, J'au-æmne Please may I have a DNA-shaped MCG with a hint of Javascript? Add a little uncertanty to the jam, please. Thanx!

*Ampton enters, walking tall. He looks at the spot where he learned to tap dance, where he ate his first donut, and at the snowball cannon just outside. He grins as much as his boom-box face allows him to grin, and confidently strides to the counter where he pulls out a note and hands it to whoever is on duty at the moment*


Hey there, remember me? Long time, no see!

I decided to take a vacation from CLI and tour around the old sites that I've not seen in a while. Anyway, I'd like a few donuts for sentimental reasons:
DNA
Moebius Strip
S3
S5
S7
IWA

Thanks.

BTW, if you're interested, Affy seems to be missing, or something. Some weird inter-dimensional guy's holding a contest to see who can find him, offering a magic stone as a prize. Just FYI.


*While waiting for the donuts to be delivered, Ampton begins break dancing, just for old times' sake*

Hi Ampton! Long time no see!

*serves multitude of *

I'm sorry that Affy's gone missing

*Ampton shrugs as he ingests an S3 and an IWA, as if to indicate that he always just assumed that Affy would eventually land himself in trouble like this*

*Ampton continues to break dance, but somehow manages to eat the S5 and S7 while doing so*