wu :: forums (http://www.ocf.berkeley.edu/~wwu/cgi-bin/yabb/YaBB.cgi)
riddles >> hard >> Language Proficiency Verification
(Message started by: william wu on Jan 11th, 2003, 3:54am)

Title: Language Proficiency Verification
Post by william wu on Jan 11th, 2003, 3:54am
I'd like to salvage the following interesting puzzle such that it is solvable, but I don't know how to -- or even if the answer is supposed to be that it's not solvable. Exact quote:


"There is a Man A, who claims to know Tamil Language. You dont know Tamil language so you can not test him. You hire another Man B, who knows Tamil Language, to test this. Both A and B knows your language too. Though you dont trust both of these guys still you want to test that A knows B. Also you know that A and B dont know any other language too. How will you test man A knows Tamil ? "


I think it's supposed to be a good puzzle because the page I found it on had it next to sink the sub and single-file hat execution, as examples of problems the webmaster discusses with his CS faculty. Problem is, I don't know what entails not trusting B. Does it mean B will translate our English queries reliably, but won't reliably tell us if A responds in Tamil properly? I e-mailed the author for clarification and his answer was meaningless, and then I e-mailed again and he stopped responding (I think he's sick of me). I won't reprint his e-mail address here. Perhaps we can figure out what is necessary to fix the problem?

One idea I thought of was to grab some obscure English text and first record A's English --> Tamil translation of it, then ask B to translate A's recording back to English and check that against the text. If B's translation of A's recording matches the text, then A knows Tamil for sure. But if there's a false negative, that only tells us that A and/or B is lying.

Title: Re: Language Proficiency Verification
Post by redPEPPER on Jan 11th, 2003, 8:01am
I found the correct riddle's wording.  Here it is:

"There is a Man A who claims to know English language.  He's a CS student and likes logic riddles.  You like riddles too so you ask him for one.  Unfortunately the riddle is worded poorly, which is a problem for a logic riddle where every detail is important.  English is obviously not Man A's native language.  You can hire another Man B who knows Man A's language and also proper English, to translate the riddle.  Though you don't know what is Man A's language.  Maybe it's Tamil.  You can communicate with Man A in English through e-mail but you can only ask one question, after which any new e-mail will be ignored.  How will you gather a correctly worded riddle?"

Title: Re: Language Proficiency Verification
Post by william wu on Jan 11th, 2003, 8:04am
LoL  :D

Title: Re: Language Proficiency Verification
Post by prince on Jan 13th, 2003, 11:55am
--"One idea I thought of was to grab some obscure English text and first record A's English --> Tamil translation of it, then ask B to translate A's recording back to English and check that against the text. If B's translation of A's recording matches the text, then A knows Tamil for sure. But if there's a false negative, that only tells us that A and/or B is lying."--

I think that's the right idea.  You could add an incentive so B would not lie i.e. add "B, your pay will be doubled if the translations match" to the text to be translated.  You could reword the puzzle so that B is not trustworthy, but is greedy.

Title: Re: Language Proficiency Verification
Post by Garzahd on Jan 13th, 2003, 2:08pm
I was discussing this with willywutang before, and I tossed that answer out because if A and B are both untrustworthy, then it's possible they may conspire against you.

For instance, say B actually knows Tamil but he also knows the obscure Arctic dialect "Foo". And by a strange coincidence, so does A. You ask them each to write something down, and they deliver identical reports in Foo, which you cannot tell from Tamil.

Then B walks off licking his chops and maybe shares half of his doubled salary with A.

Hence the original comment that we're looking to "salvage" the problem such that it has an interesting solution.

Title: Re: Language Proficiency Verification
Post by towr on Jan 13th, 2003, 2:15pm

on 01/13/03 at 14:08:59, Garzahd wrote:
For instance, say B actually knows Tamil but he also knows the obscure Arctic dialect "Foo". And by a strange coincidence, so does A. You ask them each to write something down, and they deliver identical reports in Foo, which you cannot tell from Tamil.
But it is stated they don't know any other languages..

Title: Re: Language Proficiency Verification
Post by James Fingas on Jan 13th, 2003, 2:16pm
The other way to do a similar thing would be to ask Man A to say something in Tamil, and then get Man B to translate.

Here is another idea. Get Man B to record 1000 sound clips, preferably which all sound a little similar (so that Man A cannot remember all of them). Half are to be meaningful Tamil sentences and half are to be meaningless sounds in the Tamil language. Here are some English examples:

"Brock snooby addleplum" (meaningless ... or is it?)
"Bring me a new chicken" (meaningful ... or is it?)

Now play them to Man A in random order (with replacement), and get Man A to rate each sound clip as "meaningful" or "meaningless".

1) If Man A marks half of them as meaningful, and consistently gives the same rating for each sound clip, then Man A knows Tamil and Man B was trustworthy.

2) If Man A is not consistent in his ratings, then he does not know Tamil.

3) If Man A rates all of them as "meaningful" (or all as "meaningless"), then either Man B was not trustworthy, or Man A does not know Tamil, or both.

There are some other possible outcomes, with similar sorts of conclusions. It is possible for Man B (or Man A) to prevent you from learning whether or not Man A knows Tamil (by not giving you a 50/50 mix of meaninful/meaningless sound clips, or by always answering one way), but you will know if this has happened.

Title: Re: Language Proficiency Verification
Post by Icarus on Jan 13th, 2003, 4:45pm
Jame's idea is a lot sounder than the double translation method for another reason: Anyone who has compared an original text to one that has been translated into another language and then back can tell you that even when the translators are really good and honest, the result often varies considerably from the original.

Title: Re: Language Proficiency Verification
Post by Garzahd on Jan 13th, 2003, 4:59pm
Alright, my mistake.

So maybe Foo isn't a true language, so maybe they're doing the whole wink-wink-nudge-nudge thing. My point is that if they cooperate to work against you, you're hosed.

Title: Re: Language Proficiency Verification
Post by redPEPPER on Jan 14th, 2003, 3:37am
I think the part of the riddle that says they don't know any other language is one of the clearest part of the riddle and is meant to be understood that they can't communicate by any other mean than English or Tamil (if they both know it).  No secret code involved.  If they want to work against you they'll have to do it in English or Tamil.  As you can hear what they would say in English, you have to find a way to know if they conspire in Tamil.

The most cryptic part for me is when the riddle says "you want to test that A knows B".  Was that a typo that was supposed to mean we want to test that A know Tamil, as stated at the end of the riddle?  Or are we supposed to test both?  How is it relevant if A knows B?  Would that mean they could prepare a fraud beforehand?

Also, supposing A and B are potential accomplices, what is their goal?  Is it to prevent you from verifying A's claim to know Tamil, or is it to make you believe A knows Tamil even if he doesn't?  The latter would be too easy to disprove, but the former doesn't really make sense...

Title: Re: Language Proficiency Verification
Post by James Fingas on Jan 14th, 2003, 9:55am
I think the problem with this lies in Man B. When we say he is "untrustworthy", what do we mean?

I think it is possible to prove that if Man B wants to screw up your system, then he can prevent you from ever learning whether or not Man A knows Tamil. For instance, he could completely refuse to speak in Tamil, and refuse to properly translate anything spoken in Tamil into English. In this way, he could act as if he doesn't know Tamil, and therefore be completely non-helpful in this problem.

For this reason, the concept that Man B is "not trustworthy" must be clarified for us to propose a foolproof solution.

Here is another thought, based on prince's incentive idea:

1) Tell Man B to go sit in a chair in a room. This chair is located directly underneath an anvil ("big rock" in Tamil) which will fall on Man B in precisely 15 minutes (but Man B does not know this).

2) Tell Man A the situation, and add "To convince me that you actually know the Tamil language, you must speak through an intercom into that room, and convey to Man B the danger that he is in, but without speaking English. When he jumps out of his chair (and looks up at the ceiling, worried), I will consider that proof that you can speak Tamil."

The operative idea is that although Man B may be untrustworthy (he might lie to you about whether or not Man A knows Tamil), he would look out for his own skin if he could.

One more thought: if this is actually a crypto-puzzle in disguise, then we might want to add another prop to the story: a voice-scrambler. The output of the voice scrambler is still recognizeable as speech, and can still be translated, but it is not possible to identify the speaker (even if it is yourself).

Title: Re: Language Proficiency Verification
Post by Icarus on Jan 14th, 2003, 6:16pm
James - you have a real mean streak! And what happens if it turns out that A does not speak Tamil?  Let's hope you weren't looking for a Tamil speaker to translate for you to the local cops!

Title: Re: Language Proficiency Verification
Post by James Fingas on Jan 15th, 2003, 5:30am
But what you don't tell Man A is that it is in fact an inflatable anvil ...

[/backpedal]

Title: Re: Language Proficiency Verification
Post by BNC on Jan 15th, 2003, 9:29am
It reminds me of a story by Asimov I read as a child (I can’t remember the name). In that story the hero had to identify a robot with special characteristics (modified 2nd rule for those interested).
One of the methods he tried was similar – exposing the robots to danger. The robot escaped recognition by realizing the investigator won’t kill him unless he was sure.
Similarly, our Man B may think you would not kill him if Man A can’t speak Tamil – thus he would not move!

Title: Re: Language Proficiency Verification
Post by James Fingas on Jan 15th, 2003, 9:42am
I guess I'm just a little more ruthless than that ...

Okay, here's another idea that doesn't even use Man B. Originally, I thought of this as a way to make sure Man B was actually speaking Tamil, but then I realized it doesn't matter.

Construct 1000 English sentences, all relatively close in meaning (so that Man A could never remember all 1000 of them). Record Man A speaking the Tamil translation for each of these sentences. After that has been done, play back the Tamil translations, in random order, and get Man A to translate them back to English. If the translations are fairly close to the originals, then Man A does, in fact, know Tamil.

This neatly gets around the fact that you can't trust Man B, because you can trust Man A (to try and prove that he knows Tamil). Furthermore, since it is impossible for Man A to remember all 1000 sentences, never mind trying to remember 1000 phony translations and their correlations to the 1000 sentences, Man A can only correctly give the English sentences back if he encodes the English translation into his "Tamil" translation. It is a given that Man A knows no other languages (than English and possibly Tamil), so if the "Tamil" translations are not in English, then they must be in Tamil.

Title: Re: Language Proficiency Verification
Post by towr on Jan 15th, 2003, 10:02am
There are people that can memorize whole telephone books.. So I wouldn't count on him not being able to remember 1000 sentences and their supposed translation..

Title: Re: Language Proficiency Verification
Post by James Fingas on Jan 15th, 2003, 11:50am
towr,

That's a good point, but you could just make enough sentences that you're sure he couldn't memorize them all. How many you have to make just depends on how sure you want to be.

Note the the number of possible sentences in a language is infinite, but the storage capacity of the human brain is finite, so there is some sort of limit to his possible memory abilities.

Title: Re: Language Proficiency Verification
Post by towr on Jan 15th, 2003, 12:23pm
but you might be dead before that limit is reached..
I would assume you want the test completed in a day, or a week, or at most within a year..

Title: Re: Language Proficiency Verification
Post by James Fingas on Jan 15th, 2003, 1:10pm
Now don't get all tied up in the details! What's a couple million years in the grand scheme of things? You're just jealous 'cause I didn't even need to use Man B!

Title: Re: Language Proficiency Verification
Post by Johno-G on Jan 15th, 2003, 3:02pm

on 01/15/03 at 10:02:10, towr wrote:
There are people that can memorize whole telephone books..


towr: I'm sure memorising phone books is a magic trick! I can't remember how it's done, but I remember being shown by my scout leader! No-one could possibly memorise an entire phone directory!!!
Plus, even if you could, this would have to be done over a LONG period of time - the world record for remembering a number sequence is 100, and that was exceptional! (if my memory serves me correctly, that man is now barred from every casino in Britain!) If it's a stretch for a human to remember 100 integers (that are familiar) in a set order, it's sureley gonna be impossible to memorise, on the spot, 1000 sentences in a languages you've never heard before and their appropriate responses!!

Title: Re: Language Proficiency Verification
Post by Icarus on Jan 15th, 2003, 8:06pm

on 01/15/03 at 09:29:57, BNC wrote:
It reminds me of a story by Asimov I read as a child (I can’t remember the name). In that story the hero had to identify a robot with special characteristics (modified 2nd rule for those interested).
One of the methods he tried was similar – exposing the robots to danger. The robot escaped recognition by realizing the investigator won’t kill him unless he was sure.


The story is "Little Lost Robot", and it can be found most easily in Asimov's "I, Robot" anthology. It was the 1st Law of robotics, "A robot may not injure a human being, or, through inaction, allow a human being to come to harm", that was modified (to drop the "inaction" part). The investigator, Dr. Susan Calvin, the main character of Asimov's Robot novels, pretended to place a person in jeopardy to try and spot which robot would hesitate in rescuing him. The robot escaped recognition by mimicing the behavior of normal robots.

(When you don't have much money to buy books, and don't care for the selections of the local library, you end up reading the ones you like several times. :D)

Title: Re: Language Proficiency Verification
Post by BNC on Jan 16th, 2003, 1:16am
Icarus: Yes, it was the 1st rule. I remember now. I even remember that solving the problem involved using a seemingly irrelevant talent of the robot (radiation detection). Too bad we can’t use it here.

James: Man A doesn’t need to memorize the sentences. He can just invent a “pseudo-language”. For instance, take the English sentence, and replace each letter with the letter that is 18 places after in the Alpha-Bet (cyclic). Place an “a” as every other letter letter. E.g., the word “hello” would become “zawadadag”. As far as you're concerned, it may be a valid Tamil word. And Man A can easily translate it back.

Note: I used written text, which is simpler to explain in writing, but it may be done just the same (although it may be more difficult) with spoken language.

Title: Re: Language Proficiency Verification
Post by redPEPPER on Jan 16th, 2003, 3:50am

on 01/15/03 at 20:06:15, Icarus wrote:
The robot escaped recognition by mimicing the behavior of normal robots.

I recall at one time it was quite the opposite: the robot escaped recognition by convincing the other robots to mimic its own behavior :)


Title: Re: Language Proficiency Verification
Post by James Fingas on Jan 16th, 2003, 6:44am
BNC,

That's where I'm relying on the precise definition of the problem. We know that Man A knows no other language, and I would count anything that lets you verbally encode meaning as a language. If we assume that Man A can't invent a language fast enough to encode the sentences properly, then the solution holds.

I know, it's somewhat of a lame answer, but I think it's fundamentally unsolvable if you try to use Man B.

Title: Re: Language Proficiency Verification
Post by Icarus on Jan 16th, 2003, 7:35pm

on 01/16/03 at 03:50:24, redPEPPER wrote:
I recall at one time it was quite the opposite: the robot escaped recognition by convincing the other robots to mimic its own behavior :)


I wasn't trying to give the whole story! Asimov may gone, but his heirs still have lawyers looking for copyright infringement! :o

Title: Re: Language Proficiency Verification
Post by udippel on Jan 20th, 2003, 9:28am
To redPEPPER:
You wrote: "The most cryptic part for me is when the riddle says "you want to test that A knows B".  Was that a typo that was supposed to mean we want to test that A know Tamil, as stated at the end of the riddle?  Or are we supposed to test both?  How is it relevant if A knows B?  Would that mean they could prepare a fraud beforehand?"

I'd read it straightaway as "you want to test if A knows B", meaning "you want to find out if A knows B". Let's get on with this: If they don't know each other, we can assume that any conspiration would have to be done in Tamil. We can trust A, let's assume, that he won't falsify anything B says, since he wants to convince us that he knows Tamil. At least, anything he says will be to his best ability to convince us. B definitively knows Tamil. So, logically, the sequence starts with B. We have to request B to say something to A that gives A the chance to prove to us (by speaking English to us) that he understands what B has said (or asked), irrespective if B tells the truth to A or not.
I'm 'only' stuck what such could be. Both don't know any other language, any conspiracy would only be possible in Tamil. So if we can find that they conspire (and didn't know each other beforehand), A knows Tamil. Even if they conspire and we ask B to say something to A that is impossible to prepare, this holds true as well: A will only be able to get add-on info through Tamil, which would prove A.
To me everything runs down to our clever request to untrustworthy B to extract something in Tamil from him that A could use this way or another to prove to us that he knows Tamil.

No solution, but: is this logic correct?

Title: Re: Language Proficiency Verification
Post by Phil on Jan 29th, 2003, 6:04am
Let me take a stab at this. You hire A to translate a book you wrote into Tamil, but he could just write jibberish and you wouldn't know. You put an ad in the paper for a Tamil speaker and B shows up. He could be conspiring with A. If A doesn't know B then any conspiring would have to be in Tamil, but as long as they actually speak Tamil and want you to know they speak Tamil, what would be the point of conspiring?
You know they speak no other language (how you'd know that I don't have a clue, but it's a given). The only possible conspiracy would be a coded message, prepared beforehand.
I say a word to A, he translates it to B in Tamil. B tells me the word. How would you tell the difference between a code and a real foreign language?
My solution, give words that are very similar in English, like hit, hat, hip. If the Tamil words are also very similar, it's a code. I would guess that any code complex enough to hide the similarity would also be too complex to work in your head.
Other tests: give words with shades of meaning. The double translation should not be perfect. If B always gives you back the exact word you spoke, they're cheating.
But how do you test Tamil grammar? If A knows a few Tamil words but his grammar is horrible, B could still figure out what he means and fix up his grammar in the English translation. If you tell A to translate some bad English grammar into Tamil, A can tell B in his bad Tamil grammar what you're doing.

Title: Re: Language Proficiency Verification
Post by BNC on Jan 29th, 2003, 9:56am
I think one of the problems here is that you don't trust B at all. He may get valid Tamil words, but lie and say they're not.

Title: Re: Language Proficiency Verification
Post by udippel on Jan 30th, 2003, 10:16am
for Phil:
You hire A to translate a book you wrote into Tamil, but he could just write jibberish and you wouldn't know.

check

You put an ad in the paper for a Tamil speaker and B shows up. He could be conspiring with A. If A doesn't know B then any conspiring would have to be in Tamil, but as long as they actually speak Tamil and want you to know they speak Tamil, what would be the point of conspiring?

here I don't follow. The task is to find out, if A knows Tamil. So, once you get them conspiring in Tamil, the riddle is done

I say a word to A, he translates it to B in Tamil. B tells me the word. How would you tell the difference between a code and a real foreign language?

here I'd adhere to what was said earlier: a code to communicate more than basics is 'foo'; a third language; which is excluded

I would guess that any code complex enough to hide the similarity would also be too complex to work in your head.

see above

Other tests: give words with shades of meaning. The double translation should not be perfect. If B always gives you back the exact word you spoke, they're cheating.

let's negate your phrase: if B never gives back the word you said, we're lost: either A doesn't speak Tamil or B is cheating. But what we want to find out: does A speak Tamil ?

But how do you test Tamil grammar? If A knows a few Tamil words but his grammar is horrible, B could still figure out what he means and fix up his grammar in the English translation. If you tell A to translate some bad English grammar into Tamil, A can tell B in his bad Tamil grammar what you're doing.

I assume the riddle wants a boolean solution; not evaluating the level of proficiency.

I do agree with BNC: we can not trust B to tell us the truth; we do only trust that he speaks Tamil.
Can we assume that A wants to convince us that he speaks Tamil ? I'd read a 'yes' from the text of the riddle

Title: Re: Language Proficiency Verification
Post by Phil on Jan 30th, 2003, 11:04am
Okay, I missed that B definitely knows Tamil. I was thinking it was just a general riddle about how to test the knowledge of two strangers in a subject you know nothing about. Instead its a riddle of passing trustworthy information through an untrustworthy B.

Title: Re: Language Proficiency Verification
Post by johnp. on Feb 18th, 2003, 1:27am
Rephrase of the problem: "There is a Man A, who claims to know the Tamil Language. You dont know the Tamil language so you can't test him. You hire another Man B, who knows the Tamil Language, to test Man A. Both A and B knows your language. Though you dont trust any of these guys you still  want to test that Man A knows the Tamil language. You also know that neither A nor B know any other languages besides Tamil and your own language. How will you test whether or not Man A knows Tamil?"

Hm.

The problem says nothing about A and B having any form of contact, so I'll assume the two guys know nothing about each other. Since you trust neither, it's also important to not let these guys have any form of communication. Man A might know Man B in real life and vice versa, but you'll never tell them about each other so there's no way they can plot against you.

Get Man A to write a passage in Tamil on a piece of paper, telling someone (Man B, unknown to Man A) to perform a few simple tasks, for example "tug your left ear, turn around three times counterclockwise, say Abra Kadabra and sit down on the floor". It's in A's interest to make you believe he knows Tamil, so he won't pull any tricks here if he really knows Tamil. If he doesn't know Tamil, he'll either write gibberish or try to communicate in some encoded form of your own language, not obvious to you when you read the Tamil passage thoroughly. And if it's not obvious to you, why would it be obvious to Man B? Hmmm... it could be an encoding scheme common to the Tamil-speaking population? Likely enough to even be considered? I don't think so.

Give the sheet of paper to Man B, telling him to do the tasks written on the paper. Man B doesn't know who's written these tasks (he knows nothing about Man A). If the passage is written in Tamil, Man B will either perform the tasks (alternatively tell you "hey, this is silly, I'm not going to turn around, say these silly words and sit down on your dirty floor"), thus confirming that A knows Tamil, or he'll claim it's gibberish. If he confirms, great for Man A. If Man B says it's gibberish, you've got another problem. It could be gibberish, but you can't trust Man B, so you can't really know for sure. You could ask Man B to write a similar set of tasks in Tamil for Man A to perform, but if Man B is hostile he would write gibberish, so Man A won't be able to make sense of it.

This all depends on the non-trustworthy Man B, which makes the problem unsolvable.

You can try to do tricks like buying a book or a newspaper in Tamil, pulling out words and asking both to translate into English, but as long as B is not trustworthy and could be lying / making up words you can't use him to control A.

Or?

Title: Re: Language Proficiency Verification
Post by udippel on Mar 1st, 2003, 8:45am
What do the others think ?? To me at least it seems we have an agreement on the foundations of the riddle as confirmed by johnp.

Question remains: can we somehow exploit the knowledge of B to solve the riddle or is johnp right with his suggestion 'unsolvable' ??

Title: Re: Language Proficiency Verification
Post by Quetzycoatl on Dec 31st, 2003, 9:35am

If A knows Tamil but pretends not to, then the puzzle is immpossible since he can always utter gibberish no matter what you present him with. I think we have to interpret the puzzle to mean that, whether he knows Tamil or not, he wants you to think he does.

So now the real task is to get your hands on some text that you know to be in Tamil. Given that, there are probably multiple ways to test A. Here's an example:


  • Take two Tamil texts (or one tamil and another in a third language or gibberish) and mix up the pages as much as you like.

  • Give it to A and have him translate it into English, writing the English words on the original pages above their Tamil counterparts.

  • Based on the breaks in subject matter, sentences etc. you should be able to tell if A knows Tamil.


The task now is to get B to provide you with Tamil text since he is the only source of Tamil (you can't just go to a bookstore and buy some because you think the publishing industry is out to get you and you can't tell if the books they say are in Tamil really are).

If you give B some English text to translate, he can give you back either a correct translation, an incorrect translation (though still in Tamil), or gibberish (a part Tamil part gibberish translation works just as well as an incorrect translation). We can eliminate the possibility of B translating it into some sort of code because the problem clearly states that neither A nor B know a third language.

If you run B's texts by A using the method described above then:

  • If the breaks are in the right place, then B translated into Tamil (though not necessarilly accurately) and you know A knows Tamil.

  • If the breaks are in the wrong places or don't exist then you know A is lying and therefore does not know Tamil (Of course B could deliberatley add breaks into his own translation, though I think you could get around this by creating your breaks at random places, not just mixing pages, and just check for those ignoring other breaks).

  • If A tells you the text is gibberish then he is either guessing or B is a jerk and A knows Tamil. The only thing I can think to do in this situation is to keep repeating everything, alternating who you give the English text to in the first place, until one of them doesn't say its gibberish...

Bleah




Title: Re: Language Proficiency Verification
Post by TenaliRaman on Dec 31st, 2003, 11:33pm
*somehow i missed this convo before*
William,
I am a tamil speaker.If you could direct me to the site or give the email address of the author, i could contact him to get the details of the problem.Then i can translate back in english that is ofcourse if you trust my ability in english ;).

Title: Re: Language Proficiency Verification
Post by william wu on Jan 1st, 2004, 2:28am
OK, check your IM (instant messages) box.

Title: Re: Language Proficiency Verification
Post by rmsgrey on Jan 2nd, 2004, 8:02am
I think that getting A to write a series of instructions to B (in Tamil) and telling B that he gets paid a large bonus for carrying out the instructions given has a pretty good chance of working. If A and B conspire somehow, then you've got the problem of a third language (particularly since you can trivially prevent them communicating what's in the original instructions other than by the "Tamil" instructions). Of course, if B is a complete jerk and wants to get A in trouble more than he wants money, then there's nothing you can do (assuming he's at least as smart as you are).

Title: Re: Language Proficiency Verification
Post by Quetzycoatl on Jan 2nd, 2004, 11:34am

on 01/02/04 at 08:02:08, rmsgrey wrote:
I think that getting A to write a series of instructions to B (in Tamil) and telling B that he gets paid a large bonus for carrying out the instructions given has a pretty good chance of working. If A and B conspire somehow, then you've got the problem of a third language (particularly since you can trivially prevent them communicating what's in the original instructions other than by the "Tamil" instructions). Of course, if B is a complete jerk and wants to get A in trouble more than he wants money, then there's nothing you can do (assuming he's at least as smart as you are).


I dont think there is any reason to believe that we can't easily prevent A and B from conspiring, we never have to make one known to the other after all. And as far the monetary incentive, that is a highly unpredictable method for producing results. B could already be rich for example, or a monk who has taken a vow of poverty.

Title: Re: Language Proficiency Verification
Post by King_T on May 9th, 2004, 1:31am
If the assumption can be made that the characters respond to some reward or threat and that they will abide by the rules of the discussion (no extra talking), then the puzzle is not trivially unsolvable.  If so, I don't understand why this is considered to be unsolvable.

You tell them that hey both only get paid (or live or whatever) for correctly matching answers.  Also tell them exactly what you're going to do with each person.  Ensure both know that one of the questions will be "Does A know B?", with simple confirmation if affirmative.

If you're allowed to separate them, then the method simply is:

- talk to B (in English) about some key facts about his life.  
- get B to translate a few phrases of yours into alleged Tamil and point out the nouns to you.  

- go to A and ask him (in English) to verify B's facts; you'll know if A knows B.
- change some of the nouns in your translations and get A to translate them back.  You now know if A knows Tamil.

QED

Maybe the question was originally that you meet both A and B at the same time, all signals and clues are available to all, and you have to  figure out both answers while standing there.  Then, I believe the answer to be beyond my ken...at least at 2:30am.

Title: Re: Language Proficiency Verification
Post by Three Hands on May 9th, 2004, 4:20am
Again, it is impossible through your method, King T, to ensure that the language is indeed Tamil that has been written down. It could be some other language that they both know and you don't.

Also, you are assuming a structure to the Tamil language which means that it would be grammatically correct for you to swap nouns around (either that, or I have *no* idea where you are getting these nouns from...) while other non-English languages work on the basis of nouns having genders - and gender-based terms around them - which may lead to your swapping around of the nouns creating a grammatically incorrect sentence or two, which would then provide your friend with ample reason for claiming that the statements are invalid (assuming he's remember the cipher he came up with to use with Man B...)

Perhaps the easiest solution would just to be to learn Tamil yourself from some independent person/book, and then test your friend yourself. Granted, it's a solution which takes longer, but I would imagine carries with it a higher chance of guaranteeing the claims of Man A  :)

Title: Re: Language Proficiency Verification
Post by King_T on May 9th, 2004, 10:10am
Hi Three Hands,

re: It could be some other language that they both know and you don't.

The puzzle states that neither of them knows another language other than yours and possibly Tamil.


re: Assuming a structure...

One doesn't need to use proper gender-based nouns to explain which sentences make sense, which don't, and what the closest meaning of the sentence is.  Here's a short excerpt from the conversation:

"B, please translate the following and identify the nouns:"
- It's raining, let's go inside the building.
- The camel is the fastest way to travel.
- The Tamil Tigers are going to take the World Series.

"A, please retranslate the following (Tamil sentences) into English:"
- It's raining, let's go inside the camel.
- The is the fastest way to travel.
- The Tamil Tigers are going to take the World Series.


If he doesn't know Tamil, then you'll get "Those sentences look okay to me" or a huge load of gibberish.


If he does know Tamil, then he'll tell you something like this:

"Firstly, I wouldn't suggest going into a camel if it's raining, a camel is an animal.  That word can also mean stick...either way it doesn't make sense.  And your grammar sucks."

"The *what* is the fastest way to travel?  You've left out a word or you're trying to say something else...either way the sentence doesn't make sense."

"As if...the New Delhi Yankees just brought in two closers for 200 billion rupees and have the deepest batting order since the Blue Jays' WAMCO lineup."

Title: Re: Language Proficiency Verification
Post by Three Hands on May 9th, 2004, 4:21pm
How well does your solution work if it turns out A knows B? Also, they could decide it's more fun to wind you up, and so A deliberately tells lies about B when you are trying to verify.

Then, they could also use some kind of code to make the sentences look foreign, but in fact be English in a specific code, which B then (correctly) identifies the nouns, and A correctly decodes the sentences, and so correctly identifies what the sentences are after you've changed them.

If, however, you also did a thorough check for any codes which may have been used, and could be remembered in a human mind, then you could check to see if the sentences do genuinely appear to be foreign. This may want a computer to scan the "translated" sentences in order to quickly check for codes, but that's about it.

Besides, something like an Internet translation site may prove pretty good for generating some appropriate sentences, and bypass the need for Man B...

Title: Re: Language Proficiency Verification
Post by King_T on May 9th, 2004, 4:46pm
I feel that it's clear that if A tells lies about B, that they won't get their money (or they both get killed or whatever) as their answers don't match...which is why I detailed the necessary assumption in the first place.  As I said...without it, the riddle is trivially unsolvable.

The complicated code wouldn't work because the second translator doesn't know which phrases or nouns were switched.  Even if the code was so complex and complete that it did, then it would just count as a different language which isn't allowed in the riddle.

The problem, as stated, is solved.

Title: Re: Language Proficiency Verification
Post by Three Hands on May 9th, 2004, 4:52pm
If you encoded a sentence using a letter-substitution code, then the words remain English, just disguised. Hence, the second translator would be able to decode the sentences into English, and then be confused by the words that are produced due to your moving around of nouns...

And also, they may decide that money couldn't replace the amusement that is produced by maintaining the charade that A knows Tamil, in spite of your best attempts to prove that he doesn't. And killing them because they don't know each other seems a little extreme, and wouldn't prove whether A knew Tamil or not  :)

Title: Re: Language Proficiency Verification
Post by pedronunezmd on Jun 5th, 2004, 1:59pm
Here is another way to solve this riddle without using Man B at all.

Man A does not know that you don't have a trustworthy third person who knows Tamil. So you just tell A that you have finally found a 3rd person who is trustworthy who knows Tamil.

Now you tell A that either he admits that he does not know Tamil now and leave your presence, or else he will have to translate a message from english to Tamil, at which point you will leave the room and ask your imaginary trustworthy Tamil-speaker if the message is correctly translated, at which point if it is not, the room that A is in will be flooded with poison gas.

Assuming you pull off the bluff, A will either admit he has been lying and leave, or else he will write something on a piece of paper that is in Tamil.

Then again, maybe A is suicidal.

Title: Re: Language Proficiency Verification
Post by Three Hands on Jun 5th, 2004, 7:26pm
Either that or knows you well enough that he considers the bluff worth calling...

Title: Re: Language Proficiency Verification
Post by Sean Hogan on Jul 14th, 2004, 9:52pm
"There is a Man A, who claims to know Tamil Language. You dont know Tamil language so you can not test him. You hire another Man B, who knows Tamil Language, to test this. Both A and B knows your language too. Though you dont trust both of these guys still you want to test that A knows B. Also you know that A and B dont know any other language too. How will you test man A knows Tamil ? "

Any other language besides yours and Tamil, or any language besides Tamil?

Title: Re: Language Proficiency Verification
Post by Random Lack of Squiggily Lines on Dec 7th, 2008, 12:54pm
I truly think this Problem could be solved, so I'm bringing it back from the dead. Me and my brother figured out a good idea, or a frame that could propel this, in about 5 minutes.

Anyways,

I asumed that A Is trustworthy, except for his "Tamali", And B is UNtrustworthy, EXCEPT for his Tamali. By untrustworthy , he is random.

We chose 2 questions, One that the correct answer is "Yes", one "No". Qy, and Qn.

We tell Qy to A, and he gives At, A's verson of the question.
We tell Qn to B, and he gives Bt, B's verson of the question.

We then have A tell At to B, and B tell Bt to A.

This would leave us with ABt, and BAt, either yes or no. If they both are what they were in the first place, A speaks Tamali. If not, we have room for improvement.



Powered by YaBB 1 Gold - SP 1.4!
Forum software copyright © 2000-2004 Yet another Bulletin Board