Re: [MMUSIC] Update on Orlando time for human language draft discussion

[Randall Gellens <[e-mail]>]

At 7:47 AM +0100 3/11/13, Gunnar Hellstrom wrote:

>  On 2013-03-11 00:30, Randall Gellens wrote:
>>  At 10:31 PM +0100 3/10/13, Gunnar Hellstrom wrote:
>>
>>>   On 2013-03-10 23:02, Randall Gellens wrote:
>>>>   If possible, I think it may be better to keep service 
>>>> negotiation distinct from language negotiation.  Otherwise 
>>>> things get very complex.
>>>>
>>>   That was just a use case.
>>>
>>>   Do you mean that we should divide it like this:
>>>
>>>    Imagined example in a call from A to B.
>>>    A prefer to talk English, and A has some capability to talk French
>>>    A Prefers to talk, but A is capable of typing
>>>    A Needs to get text in English or French
>>>    A has some benefit of getting spoken English or French, but A 
>>> cannot accept that as the only channel for language perception.
>>>
>>>    A wants to have a conversation with B who declares only an 
>>> ability to talk and understand spoken French.
>>>
>>>   An evaluation of the needs, preferences and capabilities of A 
>>> and B would result in that the call is not possible because we 
>>> have common capabilities only in the direction from A to the B.
>>>   In the return path, from B to A, the only option it to receive 
>>> spoken French from the B, while A is depending on getting written 
>>> English or Written French.
>>>
>>>   -----
>>>
>>>   This information can then be used in a separate service request 
>>> for getting a service C that adds rapid conversion of spoken 
>>> French to typed French in the path from A to B, still letting the 
>>> spoken French also pass over to A.
>>
>>
>>  It's very easy to add complexity in order to solve more aspects of 
>> the problem.  I think we need to work very hard to keep this 
>> mechanism, at least initially, very simple.
>>
>  Yes, so I suggest that we now add a number of use cases, both 
> simple and complex and then sort out which cases can be covered 
> without getting lost in complexity.

I agree that the document needs to be more detailed and more clear on 
the problem it is solving and the use cases it is targeting.

-- 
Randall Gellens
Opinions are personal;    facts are suspect;    I speak for myself only
-------------- Randomly selected tag: ---------------
1.  If reproducibility may be a problem, conduct the test only once.
2.  If a straight line fit is required, obtain only two data points.