Calling all linguaphiles

[gusl]

http://gus.editme.com/OntologiesLanguage

I am making a list of concepts to learn in a new language. Not just single word concepts, but also phrases or "speech actions", basically anything you might want to express in your new language. This is a Wiki, so feel free to add your own.

Text manipulation: I haven't found a quick way to hack a Windows script. I got Linux installed yesterday. I just need to figure out how to install g++ to start writing some text-manipulation scripts. Then I need to figure out how to make the network/Internet work.

Once that's done, I'll run the scripts.

apple
banana
carrot

will become

^apple: apple
^banana: banana
^carrot: carrot


When seen through a translation site such as babelfish, these pages will become bilingual glossaries.

^apple: maçã
^banana: banana
^carrot: cenoura


Possible Features:
Left-hand side: either English or New Language.

By caching the translation:
Sorting alphabetically in new language
Human edition for better concept / phrase translations (creates the problem of version control)
See several languages side-by-side, for linguistic interest


As you can see, learning Dutch is inspiring me in many directions...

Comments

[mathemajician.livejournal.com]

First question: Why the hell are you learning Dutch? That must be about the least useful language you could possibly learn! Seriously now. Not many people in the world speak Dutch and every Dutch person I've ever met speaks English so well that it's almost scary.

Anyway. Learning a language. I don't really know anything about learning languages but I'm trying to learn Italian and for better of worse I'm coming up with a lot of theories about how best to do this as I go. I even wrote a tiny java program to help with learning vocab a few weeks back. It words ok.

So, what's my latest and greatest theory? I think that the way to learn a language is by "growing" a model of the language in my mind. What you would do is have some software that tests you on the languages. The development would be broken up into a large number of levels, perhaps 100 or 200. Each level would be described by a formal "model" that the language program could read in and use. The language used would have to be reasonably complex to allow the proper expression of the models. Something like prolog I think would be best.

So to start with you'd have to learn some nouns and some basic articles:

it("gatto", noun, masculine) = en("cat", noun)
it("casa", noun, feminine) = en("house", noun )
etc...
Along with some grammar substitution rule like

noun_clause( "il " + it(noun,masculine) = "the" + en(noun) )
noun_clause( "la " + it(noun,feminine) = "the" + en(noun) )

I sure that the above is done all wrong but hopefully you get the idea... I'd need to study formal grammar descriptions some more and revise Prolog to really work out the best way to do this.

At each level you would build up the formal model of the language a little bit more. Starting with the simple and most general structures and working towards more complex structures and exceptions to the more simple rules etc. Some levels would add new nouns, some new verbs, some new tenses for verbs you already know, some new sentence connecting words and so on... The system would test the person at each level and when the person felt that they really understand one level they would then go on to the next. Each level would be a reasonably small step so you are never stuck working on any one level for very long. Perhaps one model per day and 300 models in total to learn the basics of the language.

So you sort of learn simple short patterns of words in order to learn the grammar. Over time you build up the number of these simple patterns and start to learn how to link them together to make more complex patterns.

Ok, enough ranting!!!!

ciao

[mathemajician.livejournal.com]

Oh I see, so you're actually going there to live for a while? Why? ... if I keep reading your blog I'll probably find the reason soon...:)

[gustavolacerda.livejournal.com]

I think that the way to learn a language is by "growing" a model of the language in my mind. What you would do is have some software that tests you on the languages.

I am optimistic about machines that can tell whether a simple sentence is correct: basic grammar isn't that complicated. But for advanced learners, I'm not so sure.

One of the things I want to do with this Wiki is a collection of sample sentences, some of which are "extensions" of each other.

I helped him.
I helped him stand up.
I tried to help him stand up.
Today, I tried to help him stand up.
He is thankful because today I tried to help him stand up.

(this particular set is really hard in Dutch: words change place at every addition. English has a closer mapping to the "real" semantic space.)

This way, I would like to cover all the major grammatical patterns in universal language. i.e. my set of sentences would exhibit all the major grammatical patterns in every language it's translated to. Thanks to universal grammar, human languages only differ so much. By adding sentences that express many varieties of concepts and semantic structures, I hope to achieve a universal ontology of descriptive language. Well, most novels probably have enough diversity of sentences to accomplish what I want, except for the fact that they lack their construction: so a reader who does not know the language will be a loss trying to understand the sentence by itself; whereas a reader who does not know the language can decypher the sentence step-by-step by looking up the translations of the subsentences.

So maybe what I need is add a corpus of sentences to my Wiki, and then add all their subsentences. Then native speakers of the target language can translate everything.

I think that working with examples such as these is all one needs in order to learn the grammar of descriptive language (e.g. descriptive Dutch) (there are uses of language other than informative utterances, but maybe their grammar is not so different afterall).