Hi
Here's Päivi's answer she wanted me to post to this list. I hope it's
satisfactory.
I'm giving up working as a middleman. You can contact Päivi directly (her
address below), if you need more information about FreeSpeech.
Best regards and Season's Greetings!
Pentti Luoma
-----Alkuperäinen viesti-----
Lähettäjä: Rissanen Päivi [mailto:[log in to unmask]]
Lähetetty: 19. joulukuuta 2006 12:21
Vastaanottaja: [log in to unmask]
Aihe: Re:VL: VL: VL: Transcription software
Hi,
I'm afraid that the program won't work the way which Roy Greenhalf has
described. I tested the program with a tv-set by putting a microphone near
the tv's loudspeaker, but I was disappointed. Sometimes it wrote something
correctly but the system demands the speaker to pronounce nearly perfectly.
Can we ask those persons whom we interview to speak exellent language? I
think not.
If Roy separates each voice and teaches the program to recognise it, perhaps
it works successfully even if everybody speaks just correctly. The
FreeSpeech program has to be "taught" by reading text to the program. So,
the owner of the voice should personally be present teaching the program. It
takes a few minutes.
Then the voice owner will be named as a user, e.g. "Roy". I think it's much
easier to use a "parrot-method"; first you listen to a part of the interview
on the earphones and then say the same words with your own voice by using
the microphone, and at the same time correct the speech of the interviewed
person.
It takes less time than type-writing.
BR,
Päivi Rissanen
Pentti Luoma (18.12.2006 15:31):
>Hei
>
>Tällaisia kysymyksiä tuli listalta. Miten vastaisit?
>
>
>Terveisin
>Pentti
>
>-----Alkuperäinen viesti-----
>Lähettäjä: qual-software [mailto:[log in to unmask]] Puolesta
>Roy Greenhalgh
>Lähetetty: 18. joulukuuta 2006 12:45
>Vastaanottaja: [log in to unmask]
>Aihe: Re: VL: VL: Transcription software
>
>Pentti
>
>Thanks for your note, which leaves me confused!
>
>I read that the software can handle 10 voices, "but it can be
>> taught to convert about 10 different speakers' voices. In that case
>> all the persons which are speaking should need a personal computer at
>> this moment, because one "FreeSpeech" is able to recognise only one
>> voice
>at a time."
>
>Does this mean that I could have 1 copy on my PC, and when I "play" the
>recording, it can a) separate out the 10 voices and b) recognise speech
>from each of them and make, say, a 90 per cent accurate "translation"?
>I assume that I need the voice characteristics file of each of the 10
>speakers on my PC, or is this no longer needed?
>
>Regards..
>
>Roy Greenhalgh
>
>
>Quoting Pentti Luoma <[log in to unmask]>:
>
>> Hi
>>
>> When I asked Päivi Rissanen, if the speech to text -option were only
>> "an urban myth", her answer was "no". She kindly promised me to
>> forward her views (see below) on the use of Philips Free Speech Viva
>> to
>the list.
>>
>> I'll send a copy of this message to Päivi, too.
>>
>> Regards
>>
>> Pentti Luoma
>>
>> -----Alkuperäinen viesti-----
>> Lähettäjä: Rissanen Päivi [mailto:[log in to unmask]]
>> Lähetetty: 18. joulukuuta 2006 9:45
>> Vastaanottaja: [log in to unmask]
>> Aihe: Re:VL: Transcription software
>>
>> Hi Pentti,
>>
>> I think that it isn't just an urban myth, while it's possibilities
>> are limited. It doesn't work directly with group discussions etc.,
>> but it can be taught to convert about 10 different speakers' voices.
>> In that case all the persons which are speaking should need a
>> personal computer at this moment, because one "FreeSpeech" is able to
>> recognise
>only one voice at a time.
>>
>> I have used it in my interwiew-research. The method I used was not
>> comfortable. Background information: I had to tell the story of the
>> person whom I had interwiewed on a video. But, the program was rather
>> good and "willing to learn to be better" recognizing my style of
>> speech during the work. It took less time to speak and correct than
>> to
>write it manually.
>>
>> The step which every researcher of human sciences has dreamed about
>> is still far away. Meaning that the program works like a quick secretary:
>> immediately, when you or any member of a group has said something,
>> even with an accent, the program writes it correctly and it also uses
>> different fonts with different speakers.
>>
>> The giant step after that would be the program, which recognises the
>> thoughts of the interwiewed persons - so that they don't even have to
>> speak during an interwiew - or is that only a dream of the future?
>> :-)
>>
>> Best Regards,
>>
>> Päivi Rissanen
>>
>
|