Conversational ability test.

[Robin Lee Powell]

tsani and I independently came up with this test within a couple of
hours of each other yesterday (!?), and it seems like it helps
people grok what "I can speak Lojban" means.

Stick two people in two adjacent rooms; they can hear each other but
not see each other. Present them with a co-operative puzzle to exit
the room. Shoot them :) if they speak anything but Lojban.

People who can get out of the room "speak Lojban". That doesn't
mean they're "fluent", whatever the hell that means. But they can
get shit done.

Now, here's the fun part: can you come up with a set of tasks that
we could implement on a couple of laptops so we could actually try
this at jbonunsla?

For bonus points, can you come up with tasks that have various
difficulty scales? How about tasks that require something other
than simple directional instructions (i.e. "OK, now try the button
on the right")?

-Robin

--
http://singinst.org/ : Our last, best hope for a fantastic future.
.i ko na cpedu lo nu stidi vau loi jbopre .i danfu lu na go'i li'u .e
lu go'i li'u .i ji'a go'i lu na'e go'i li'u .e lu go'i na'i li'u .e
lu no'e go'i li'u .e lu to'e go'i li'u .e lu lo mamta be do cu sofybakni li'u

[.arpis.]

Are "sentences" made up of lojban words which are incorrect
grammatically lojban? What about if both people misuse (by implicit
contract) certain words (I can't think of a good example now)? What if
they habitually forget to separate the main bridi from the sumti
before it? What about lojban with a lot of "zoi" and "zo'oi" and "me
la"? Where is the line drawn between incorrect lojban, bad lojban, and
not lojban?

My complaints aside, that's a really cool idea.

I may be colored by recent experience, but bootstrapping using already
made games might be effective.

I'm thinking specifically about Portal 2 and IloMilo (although both
have mechanisms for avoiding need for any conversation), but more
generally speaking, playing a cooperative puzzle game might do.

> --
> You received this message because you are subscribed to the Google Groups "lojban" group.
> To post to this group, send email to loj...@googlegroups.com.
> To unsubscribe from this group, send email to lojban+un...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/lojban?hl=en.
>

--
mu'o mi'e .arpis.

[MorphemeAddict]

I understand Robin's suggestion to be a test for just one person, even though two people are involved. So if the testee knows Lojban well enough to talk the 'tester' through solving whatever the puzzle or other task is, then he 'knows' Lojban. The 'tester', OTOH, should be a native Lojban speaker with no other language in common with the testee. 

I don't think this would be an adequate test, but it's a start. 

stevo

[Robin Lee Powell]

I actually had it in mind as a test for both people.

-Robin

[John E Clifford]

Or a doable one, in the absence of said native speaker (who could probably tell without all this apparatus).

---

From: MorphemeAddict <lyt...@gmail.com>
To: loj...@googlegroups.com
Sent: Wed, January 4, 2012 1:35:56 PM
Subject: Re: [lojban] Conversational ability test.

[Pierre Abbat]

On Wednesday 04 January 2012 13:40:36 .arpis. wrote:
> Are "sentences" made up of lojban words which are incorrect
> grammatically lojban? What about if both people misuse (by implicit
> contract) certain words (I can't think of a good example now)? What if
> they habitually forget to separate the main bridi from the sumti
> before it? What about lojban with a lot of "zoi" and "zo'oi" and "me
> la"? Where is the line drawn between incorrect lojban, bad lojban, and
> not lojban?

I think that using bare foreign words should be penalized more heavily than
quoting them with ZOI, which should be penalized more heavily than
grammatical errors. It may be difficult or impossible, in certain cases, to
avoid foreign words. For example:
(alis) .i ko lebna le .arxokuna le tricu
(bab) le moki'a
(alis) .arxokuna
(bab) lo .arxokuna cu mo
(alis) me la'o sy. mapache .sy.
(bab) ua .i mi punji le .arxokuna ma
I'd ignore, or score very lightly, the sorts of errors that everyone makes
when speaking, even in his native language. For example, I've heard a Spanish
speaker say "otro", then realize that the noun he's about to say is feminine,
then say "otra" and the noun. Of course, until we have some native Lojban
speakers with enough proficiency, we can only guess what sorts of errors
those will be , but here's one example. You mean to say "jmina", but get the
nasals mixed up, and half devoice the "j", resulting in something
between "cnima" and "jnima". You then say "si", but you don't know which the
listener heard. The preceding word is a CV cmavo like "mi". So you don't know
whether the hearer is erasing one word "cnima" or two "mi jnima".

Pierre
--
li fi'u vu'u fi'u fi'u du li pa

[MorphemeAddict]

Having a native speaker is preferable, but can be replaced by using the most fluent speakers available. Lojban has always been a bootstrap operation. 

stevo

[Robin Lee Powell]

For me the goal was to test *both* people, not just one of them. So
I find having one of them be particularily advanced to be counter to
my idea.

The test is the task, not the other person.

-Robin

On Wed, Jan 04, 2012 at 07:35:06PM -0500, MorphemeAddict wrote:
> Having a native speaker is preferable, but can be replaced by using the
> most fluent speakers available. Lojban has always been a bootstrap
> operation.
>
> stevo
>
> On Wed, Jan 4, 2012 at 4:17 PM, John E Clifford <kali9...@yahoo.com>wrote:
>
> > Or a doable one, in the absence of said native speaker (who could probably
> > tell without all this apparatus).
> >

> > ------------------------------
> > *From:* MorphemeAddict <lyt...@gmail.com>
> > *To:* loj...@googlegroups.com
> > *Sent:* Wed, January 4, 2012 1:35:56 PM
> > *Subject:* Re: [lojban] Conversational ability test.

[tsani nicte]

Of course, we can only guess at the type of mistakes that'll be made, but if the listener can't understand, the speaker can expect a [ki'a] and then start over with [sa] or [su], as required. This test seems like a good idea to me too. (I mean, I did come up with it practically simultaneously with Robin) 

From my point of view, zoi-quotes shouldn't be allowed in such a way that they quote another language. All test subjects should pretend to speak exclusively Lojban. This way, it'd be necessary to describe what is meant using only Lojban.

On a side-note, I'd expect zoi-quotes to be used only to quote non-linguistic material, such as hyperlinks (as clearly demonstrated on IRC) and onomatopoeia. 

Back on topic, it's clear that a certain amount of vocabulary is necessary, at least in terms of gismu. (In fact, knowing only gismu, it's possible, I believe, to avoid using all the TAG cmavo at the exception of [fi'o], should [fi'o] function exactly as does TAG itself in all contexts.) Otherwise, a already demonstrated grasp of the grammar would obviously be required. Typically, I observe, rather skilled Lojbanists do clearly understand the grammar, yet are simply lacking in the department of vocabulary. (I myself am a victim of vocab-laziness.) I estimate that knowing the 700-or-so most frequently used gismu is a sufficient amount when it comes to solving these various puzzles. 

As for the puzzles themselves, I'm not certain as to what could potentially be implemented. Throwing switches in a coordinated fashion could test for comprehension and correct use of ordinals, as I believe was pointed out by Robin in #lojban. To test for directions, perhaps, we could construct a sort of maze or obstacle course. The testee is required to escape, but is blindfolded and can hear nothing but the voice of the tester, who omnisciently observes the maze from above, or through a camera affixed to the head of the testee. Following only the tester's instructions, the testee would have to escape.

Any other thoughts for tests?

mu'o mi'e la tsani

[Bob LeChevalier, President and Founder - LLG]

tsani nicte wrote:
> As for the puzzles themselves, I'm not certain as to what could
> potentially be implemented. Throwing switches in a coordinated fashion
> could test for comprehension and correct use of ordinals, as I believe
> was pointed out by Robin in #lojban. To test for directions, perhaps, we
> could construct a sort of maze or obstacle course. The testee is
> required to escape, but is blindfolded and can hear nothing but the
> voice of the tester, who omnisciently observes the maze from above, or
> through a camera affixed to the head of the testee. Following only the
> tester's instructions, the testee would have to escape.
>
> Any other thoughts for tests?

Look to role-playing games. A frequent idea is a puzzle that involves
some amount of thinking and some amount of role playing. Planning and
then carrying out the plan may require good communication (though most
role playing puzzles would involve more than two interlocutors).

Psychologists use role-playing to work through social interaction
problems, and these also would be a source of "situations" to be worked
through, and these interactions tend to be more like everyday
communications.

Negotiation is probably one of the tougher communications problems.

The important thing is not that the puzzle or problem be successfully
solved, but that the two parties come up with and implement an approach
that demonstrates their skill in communicating.

--
Bob LeChevalier loj...@lojban.org www.lojban.org
President and Founder, The Logical Language Group, Inc.