The difference between US and UK English for this purpose is the way that the
sounds are grouped together dependent on what is expected in the pronunciation.
For example if you say a particular phrase, the US pronunciation will be
different and the sequence of sounds represented will be different than that
expected by a UK sequence. The differences are not that large but there are
particular vowels that will be affected and also the rhotic /r/. For example,
the problem will occur in a phrase where 2 different vowels are expected in US
English but in the pronunciation of UK English the vowels will be the same.
This could mean that the wrong unit may be selected which will distort the
output.
Having to turn off the pronunciation module (due to it expecting a particular US
English sound sequence) also restricts the quality a bit as usually it would
only accept the input which is well-articulated and clear. So in this case it
may use units that are not well produced or have mistakes in them.
Basically it would improve the output a bit but I don't think that it would make
a huge jump in the quality if it was designed with a UK speaker in mind.
The webpage below gives some comparison between other speakers and their
synthesised output. There is quite a range of quality there so you can hear how
good it could be and also what people generally end up with.
http://www.modeltalker.com/comparison.html
I don't know if they are planning on designing modeltalker for UK speakers, at
the moment they are still testing and researching the whole system so it may be
a while before the next step is taken.
Cheers,
sarah
Quoting Simon Judge <[log in to unmask]>:
> Interesting, you can definitely note (I think?!) the 'americanisation' of
> the voice in the parts where it is doing the concatenation... So I assume
> this would be a _lot_ better with a UK model? Have they any plans to
> produce a UK model Sarah?
>
> Simon
>
> -----Original Message-----
> From: A discussion list for Assistive Technology professionals.
> [mailto:[log in to unmask]] On Behalf Of S Creer
> Sent: Monday, July 21, 2008 4:48 PM
> To: [log in to unmask]
> Subject: Re: Voice sampling enquiry
>
> Hello,
> I have just received my voice back from the people at Modeltalker and have
> attached a couple of example wav files. The 2 files with orig_ prefix are
> original recordings that I have done and hotel.wav and canoe.wav have been
> produced using Modeltalker. To build the voice takes ~1600 phrases to be
> recorded. They did warn me that it wouldn't be the same quality that they
> could produce for American English speakers. The tool that it uses to do the
> recordings limits what it will take as input to minimise the variation in
> terms of pitch, loudness and pronunciation, but as a British English speaker
> I had to turn off the pronunciation module.
> As I say, I have just had this back today so haven't done a full
> investigation of what it will produce but I thought you might be interested
> in hearing the output.
> All the best,
> Sarah
>
> ------------------------------------------
> Sarah Creer
> phd student
> Clinical Applications of Speech Technology Departments of Computer Science
> and Human Communication Sciences University of Sheffield
> [log in to unmask] www.dcs.shef.ac.uk/~sarahc
>
>
>
>
> Quoting "Phillips Tracy (RBF) NOC" <[log in to unmask]>:
>
> > Hi Liz,
> >
> > One of my clients has recently been asking the same questions.
> > Have you looked at the following sites to create synthesized speech?
> >
> > http://www.asel.udel.edu/speech/clinicians.html
> > http://www.modeltalker.com/
> >
> > We haven't tried it, but would be interested to hear from anyone who has.
> >
> > Best wishes,
> > Tracy
> >
> > Tracy Phillips
> > Speech and Language Therapist in AAC
> > Specialist Disability Service
> > Oxford Centre for Enablement
> > Nuffield Orthopaedic Centre
> > Windmill Road
> > Headington
> > Oxford
> > OX3 7LD
> >
> > Tel: 01865 737445
> > E-mail: [log in to unmask]
> >
> > -----Original Message-----
> > From: A discussion list for Assistive Technology professionals.
> > [mailto:[log in to unmask]] On Behalf Of Liz Scott-Tatum
> > Sent: 07 July 2008 15:44
> > To: [log in to unmask]
> > Subject: Voice sampling enquiry
> >
> > We have recently beed approached by a person with a degenerative
> > condition who wants to do some voice banking, before their speech
> > deteriorates further. We would really appreciate any advice the group
> > can offer regarding how we could go about this, any specific software
> > which can be used, what quality the sounds are, and how easy it is to
> > use the samples on different voice output communcaition aids. Finally
> > is it possible to take sufficient voice samples to create their own
> > synthesised voice - if so what would be needed to achieve this?
> >
>
>
>
> No virus found in this incoming message.
> Checked by AVG - http://www.avg.com
> Version: 8.0.138 / Virus Database: 270.5.3/1564 - Release Date: 7/21/2008
> 6:42 AM
>
>
> ***********************************************************************
> This message may contain confidential and privileged information.
> If you are not the intended recipient you should not disclose, copy
> or distribute information in this e-mail or take any action in reliance
> on its contents. To do so is strictly prohibited and may be unlawful.
> Please inform the sender that this message has gone astray before
> deleting it. Thank you.
>
> 2008 marks the 60th anniversary of the NHS. It's an opportunity to pay
> tribute to the NHS staff and volunteers who help shape the service, and
> celebrate their achievements.
>
> If you work for the NHS and would like an NHSmail email account, go
> to: www.connectingforhealth.nhs.uk/nhsmail
> ***********************************************************************
>
|