Some people are idiots.  Looky at this.
----- Original Message -----
From: Nick G
Sent: Monday, October 02, 2006 4:01 PM
Subject: Re: [MT-discuss] STOP IT RIGHT KNOW

If you did, in fact, not subscribe to anything, then options for unsubscribing are at the bottom of this message.  Thanks for reading, or, as the case may be, not reading.
----- Original Message -----
Sent: Monday, October 02, 2006 2:10 PM
Subject: [MT-discuss] STOP IT RIGHT KNOW

STOP SEND ING ME THESE MEIAL I DID NT SUBSCRIDE TO ANYTHING

Tim Bunnell <[EMAIL PROTECTED]> wrote:
Jayson is right on both counts: InvTool only uploads "fresh" material;
and, setting an utterance to "unrecorded" in the .inv file will cause
InvTool to offer it for recording again. Each line in the .inv file
pertains to one utterance. Lines start with a number that indicates the
status of the utterance and if that number is '0', it means the
utterance has not been recorded. So, if you are facile with a good text
editor (emacs or vi), you simply search for the line containing the name
of the file you want to rerecord (that's the second item in each line)
and set the number to 0.

t

Jayson Smith wrote:
> Hi,
> When you open up an inventory in Invtool, it starts with the first utterance
> it doesn't think you've recorded. The point of this is, rather then opening
> up the inv file and then hitting Save/Back/whatever a bunch of times
> searching for one particular utterance, you can make it think you never
> recorded it, so if you've in fact done the whole inventory or have recorded
> up to later than that one utterance, it will start you out right where you
> want to rerecord. Quite nice, actually.
> Jayson.
>
> ----- Original Message -----
> From: "snowman" <[EMAIL PROTECTED]>
> To: "Jayson Smith" <[EMAIL PROTECTED]>; "Tim Bunnell"
> <[EMAIL PROTECTED]>
> Cc: <[EMAIL PROTECTED]>
> Sent: Sunday, October 01, 2006 7:37 PM
> Subject: Re: [MT-discuss] ;Six Sucks
>
>
>> Interesting. Why is it that you would want to remove an utterance,
> instead
>> of simply re-recording it?
>>
>> ----- Original Message -----
>> From: "Jayson Smith" <[EMAIL PROTECTED]>
>> To: "snowman" <[EMAIL PROTECTED]>; "Tim Bunnell" <[EMAIL PROTECTED]>
>> Cc: <[EMAIL PROTECTED]>
>> Sent: Sunday, October 01, 2006 5:59 PM
>> Subject: Re: [MT-discuss] ;Six Sucks
>>
>>
>>> Hi,
>>> Actually, from what I understand, the system will only upload what
> hasn't
>>> already been uploaded. So if you rerecord that one utterance, that's
> all
>>> that will be uploaded. Of course if you add your own utterances those
>>> would
>>> be uploaded too.
>>> On the topic of finding that one particular utterance, there's a way to
>>> make
>>> Invtool think you never recorded it, but it's a bit complicated and care
>>> must be taken to avoid changing anything else. To do this, you need to
>>> edit
>>> the .inv file by hand, with a program that does not word wrap! I've
> done
>>> this before. If you want to know how, send me a note and I'll tell you.
>>> If
>>> others want to know how to do this, let me know also. This is
> especially
>>> useful when you give up on recording an utterance and later want to try
>>> again and see if the meters will accept it this time around.
>>> Jayson.
>>>
>>> ----- Original Message -----
>>> From: "snowman" <[EMAIL PROTECTED]>
>>> To: "Tim Bunnell" <[EMAIL PROTECTED]>
>>> Cc: <[EMAIL PROTECTED]>
>>> Sent: Sunday, October 01, 2006 6:48 PM
>>> Subject: Re: [MT-discuss] ;Six Sucks
>>>
>>>
>>>> Ah ha!!! Very interesting. You don't know you are a defective talker
>>> until
>>>> you bring the microscope near your lips. That makes a lot of
>>> sense.
>>>> Of course, I won't re-submit at this point, even though that's a pretty
>>>> important word. :-).
>>>> But, when I do re-submit, I presume I would just open my old invintory
> in
>>>> INVTool, find and rerecord that phrase, then Upload the entire thing
>>> again.
>>>> is that correct. Or, is their any way just to upload the new phrases?
>>>>
>>>>
>>>> ----- Original Message -----
>>>> From: "Tim Bunnell" <[EMAIL PROTECTED]>
>>>> To: "snowman" <[EMAIL PROTECTED]>
>>>> Cc: <[EMAIL PROTECTED]>
>>>> Sent: Sunday, October 01, 2006 8:50 AM
>>>> Subject: Re: [MT-discuss] ;Six Sucks
>>>>
>>>>
>>>>> You didn't know we editorialize? :-)
>>>>>
>>>>> Unfortunately, it's not a matter of ModelTalker doing a bad job of
>>>>> transcribing the word--that would be easy to fix. Looking at the
>>>>> biphone
>>>>> tracking output when your voice is supposed to say sucks, I see this:
>>>>>
>>>>>> 00SS from par202
>>>>>> SSAH from par508
>>>>>> AHKK from par508
>>>>>> KKSS from par508
>>>>>> SS00 from par336
>>>>> This shows where ModelTalker found the each of the pairs of phonemes
> it
>>>>> pulls concatenation data from. The initial silence and initial part
> of
>>> /s/
>>>>> (00SS) came from par202.wav. Then then entire syllable up to the end
> of
>>>>> the final /s/ and trailing silence came from the utterance par508.
> That
>>>>> utterance was "Most successful". We expected (and ModelTalker
>>>>> transcribes/pronounces) the initial part of the word "successful" to
> be
>>>>> just like "sucks"; it could be spelled sucksessful as far as our
>>>>> pronunciation is concerned. Unfortunately, when I listen to this word
>>>>> as
>>>>> recorded by Jim, the pronunciation is pretty much like sixsessful.
> Our
>>>>> pronunciation meter in InvTool is supposed to detect differences
>>>>> between
>>>>> our transcription and the talker's pronunciation, but differences
> like
>>> the
>>>>> difference between "six" and "sux" are subtle enough acoustically
> that
>>>>> they often go undetected.
>>>>>
>>>>> Sometimes a work around is to substitute an alternative pronunciation
>>> for
>>>>> the word. For instance, transcribe sucks as SSAXKKSS instead of
>>>>> SSAHKKSS
>>>>> (the AH is a more stressed variant of AX). Unfortunately, in Jim's
>>>>> voice
>>>>> that too sounds like "six".
>>>>>
>>>>> Possibly the best fix is to add one or two instances of words that
> have
>>>>> the syllables suck and sucks in them to the inventory as user defined
>>>>> utterances. At the same time, we may want to change the way we
>>> transcribe
>>>>> words like "successful" in this inventory since that talker
>>>>> consistently
>>>>> used a more front vowel than our transcriptions expect.
>>>>>
>>>>> We haven't the staffing to do a lot of this, but if people identify
>>>>> specific problems, rerecord utterances or add new ones, and upload
> the
>>>>> changes, we can run off new versions of voices. But please, save up
>>>>> your
>>>>> changes until you have identified the most serious problems so we can
>>> fix
>>>>> them all at once.
>>>>>
>>>>> Thanks,
>>>>>
>>>>> t
>>>>>
>>>>> snowman wrote:
>>>>>> Oh no!!! This won't do at all! My voice mispronounces a very
>>> important
>>>>>> word. When it means "sucks", it says "six". And, of course, I have
>>>>>> my
>>>>>> thoughts about that. but, I don't really have any good way to tell
> it
>>>>>> what I think, because, when I say what I really think, it thinks I
>>> mean
>>>>>> six. And of course I don't. I guess I could always try saying six,
>>> and
>>>>>> see what it makes of that. But, I dint think of that until just
> now.
>>> So
>>>>>> , I'll go try that now. But I don't think it will help.
>> ------------------------------------------------------------------------
>>>>>> _______________________________________________
>>>>>> mt-discuss mailing list
>>>>>> [EMAIL PROTECTED]
>>>>>> https://wizard.asel.udel.edu/mailman/listinfo/mt-discuss
>>>>> --
>>>>> Head, Speech Research Lab | Research Associate Professor
>>>>> duPont Hospital for Children | Computer and Info Sciences/
> Linguistics
>>>>> Voice: 1+302+651-6835 | University of Delaware
>>>>> Fax: 1+302+651-6895 | URL: www.asel.udel.edu/speech/
>>>>>
>>>>
>>>> _______________________________________________
>>>> mt-discuss mailing list
>>>> [EMAIL PROTECTED]
>>>> https://wizard.asel.udel.edu/mailman/listinfo/mt-discuss
>>>>
>>>
>>
>
>
> _______________________________________________
> mt-discuss mailing list
> [EMAIL PROTECTED]
> https://wizard.asel.udel.edu/mailman/listinfo/mt-discuss

--
Head, Speech Research Lab | Research Associate Professor
duPont Hospital for Children | Computer and Info Sciences/ Linguistics
Voice: 1+302+651-6835 | University of Delaware
Fax: 1+302+651-6895 | URL: www.asel.udel.edu/speech/

_______________________________________________
mt-discuss mailing list
[EMAIL PROTECTED]
https://wizard.asel.udel.edu/mailman/listinfo/mt-discuss



  Peace......................................................how do we get it????


Get your email and more, right on the new Yahoo.com


_______________________________________________
mt-discuss mailing list
[EMAIL PROTECTED]
https://wizard.asel.udel.edu/mailman/listinfo/mt-discuss

Reply via email to