Re: language trainer - kind of thing
From: Alexei A. Frounze (alexfru_at_chat.ru)
Date: 02/15/05
- Next message: Stewart Gordon: "Re: Waistcoats and weskits"
- Previous message: Richard Herring: "Re: person specification (polite/familiar)"
- In reply to: Alexei A. Frounze: "language trainer - kind of thing"
- Next in thread: Des Small: "Re: language trainer - kind of thing"
- Reply: Des Small: "Re: language trainer - kind of thing"
- Reply: Lee Sau Dan: "Re: language trainer - kind of thing"
- Messages sorted by: [ date ] [ thread ]
Date: Tue, 15 Feb 2005 13:37:28 +0300
It seems to me that such a project has quite a lot in common with the famous
set of programs acting as interlocutor. I remember that that has been quite
a popular thing and there even was some kind of subjective test for the
operation of such an AI-aided program ... like the user had to conversate
with it for a few minutes or something like that and then guess whether or
not he has talked to a computer/program or another human (obviously using
another computer connected with the user's).
I mean, if we put aside the analysis of the user input, such a program
basically attempts to construct a sentense using the word banks, grammar
rules and whatever else information it may need for this.
Right?
Could anyone give links to such interlocutor AI-aided software? Perhaps that
may have some value to me as well.
Thanks,
Alex
"Alexei A. Frounze" <alexfru@chat.ru> wrote in message
news:375tm7F58fel7U1@individual.net...
> Hi,
>
> recently I've started learning yet another foreign language (I'm Russian)
> and came across the same old problem which I've almost forgotten about
> during the never-ending course of practcing my 1st foreign language,
> English...
>
> As usual, it's not just the vocabulary that must be learned from the
scratch
> (well, the reality is no so bad), but the grammar with all its rules,
> tenses, verb conjugations, pronouns and whatever comes in the middle of
> them... I'm learning Spanish now.
>
> To my surprise I've discovered that the language is indeed relatively
> simple, especially when put in right words, clear rules, and accompanied
> with comprehensive examples (I managed to find a great Spanish grammar
> reference book in English at one of the biggest bookstores in Moscow,
which
> could have been impossible just as well because normally none expects
anyone
> else to learn a foreign language through another foreign rather than
through
> the native one :).
>
> So, I started learning the language in its very core, the grammar. I'm
> leaving much of the vocabulary and the speech-pracrice for the future.
This
> works for me well. I'm very used to understand from the structured
> information rather than from just a collection of disjoint or not
otherwise
> categorized examples.
>
> But to me, w/o attending any courses/classes of the language (I'm a
fulltime
> software engineer), it's hard to make up the multitude of the Qs and
correct
> As to perfect my skills and memorize whatever is not memorized yet (not to
> mention the "simple" movement from the short-term to the long-term memory,
> which also needs some regular practice).
>
> Having read the grammar book about 5 times from the beginning till the
end,
> I came up with an idea... And the idea is not sort of grown up from
> nowhere... I have a few other Spanish textbooks, none of them presenting
the
> information suitable for learning the grammar in a structured way. I have
2
> different CD Spanish courses providing almost the same thing (except for
> colorful pix, audio and video) and a number of prepared, fixed,
unextendable
> questions/quizes/puzzles/whatever. It's just that. If I want to learn the
> most common irregular verbs in all their forms or just learn the entire
> conjugation pattern, I must look elsewhere or simply do everything by hand
> wasting lots of time to just get prepared for the actual process of
learning
> something -- I must structure things, prepare questions and answers and
then
> repeat them often with various modifications like using different verbs,
> pronouns, tenses, etc. I mean I didn't find any practical aid in those
many
> language sources.
>
> So... The idea... The idea is simple, though to some clever guys in here
it
> may look kinda stupid or maybe even someone may say that it's impossible
to
> do it, but... I want to make a language trainer application for the PC.
> The functionality is rather simple: within selected grammar area, at
random
> a sentense is being genrated according to the grammar rules being
currently
> studied by the user. The sentense is than shown on the screen in one
> language (native or English would do in my case) and program asks to enter
> the translation of it in the language being learned (e.g. Spanish).
>
> Of course, this implies that the sentenses be generated simple so as not
to
> make the whole thing being aka mission impossible. :) Personally, I think
> that there are several areas that can be trained separately:
> - word banks, common expressions, whatever else may fall here
> - verbs and their conjugations, tenses, pronouns, sentense structure (and
> the order of parts of speech in it), what else is missing?
>
> What I'm currently talking about is the 2nd of the two above... Here's a
> portion of my other post to a Spanish-related newsgroup:
>
> -----8<-----
> Besides, the program obviously should have the translations of the words
> being used in it, in all forms that are applicable...
>
> Example:
>
> infinitive: eat <-> comer
>
> simple present:
> (I) eat <-> (yo) como, (you) eat <-> (tú) comes, (he/she/it) eats <->
> (el/ella/Ud.) come
> (we) eat <-> (nosotros) comemos, (you) eat <-> (vosotros) coméis, (they)
eat
> <-> (ellos/ellas/Uds.) comen
>
>
> present participle / gerund: eating <->comiendo
>
> past participle: eaten <-> comid<o|a>[s]
>
> etc...
>
> So, for example if the program decides to use 3rd person plural, use the
> verb to eat and the present tense, then it asks for a translation of:
>
> They <eat|are eating>
>
> and expects:
>
> [Ellos|Ellas] <comen|estan comiendo>
>
> There should be no other words that are irrelevant to the tense, verb
itself
> and its conjugation, word order, gender and number agreements.
>
> I think in this simple way it's possible to organize the rules formally
and
> generate a particular example/problem just out of the rule and word banks.
>
> I'm not saying it's possible to make an ideal thing, but something that'll
> make it possible to train the basic stuff sounds like feasible.
>
> -----8<-----
>
>
> At the moment, conjugation isn't a problem -- it can all be easily ripped
> off from a big reference of irregular verbs. The main thing is perhaps the
> logic and the rules that will make the application generate sane sentenses
> (though, who cares if you occasionally eat/drink someone (or yourself :),
> provided the grammar is OK? :).
> Currently I'm thinking of managing a number of sets of different things
such
> as verbs of different types (transitive, intransitive, reflexive),
pronuons
> (all kinds of), some nouns, maybe something else to make up simple
> direct/indirect objects. The primary concern is to pick whatever can be
> picked at random to a degree not breaking the constraints imposed by the
> grammar. At the same time, I'd like to have the basic possibility to have
> multiple correct answers (because, as you might know, the pronouns can be
> dropped if the verb conjugation tells you the person)... and... I'd love
the
> thing to be extendable with new words and possibly new grammar rules --
> through a number of text files (maybe requiring some postprocessing like
> rebuilding an index for fast search within files, or just translate text
to
> binary form for the mere use during the app normal use).
> I'm not saying anything about statistics and scoring here and the
> organization of the things alike, it's perhaps the least important now...
>
> What do you think about it? And would be out there anyone willing to
> participate in such a project, to lend a helping hand?
>
> TIA,
> Alex
> P.S. of course, I'm not pretending to make something sutable for each and
> every language, but for English and Spanish this would probably work OK.
> P.P.S. I sure can program myself, but as noted, not too much of spare
time,
> and, two or more thinking heads are better than one :)
> P.P.P.S. Expect no money, no nothing for your participation :) But... Who
> knows... If I've never seen similar program yet, why not this be the first
> ever, why not make you famous? It might sell or become 2nd linux with
> thousands of hackers and users :))
>
>
>
- Next message: Stewart Gordon: "Re: Waistcoats and weskits"
- Previous message: Richard Herring: "Re: person specification (polite/familiar)"
- In reply to: Alexei A. Frounze: "language trainer - kind of thing"
- Next in thread: Des Small: "Re: language trainer - kind of thing"
- Reply: Des Small: "Re: language trainer - kind of thing"
- Reply: Lee Sau Dan: "Re: language trainer - kind of thing"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|