Corpora: 9 Scientist Jobs

David Horowitz dhorowitz at voxgeneration.com
Thu Apr 4 11:40:38 UTC 2002


Principal or Senior Scientist in Digital Signal Processing for
Text-to-speech Synthesis

Work on a large international project hosted at Vox Generation Ltd.  Vox
is building a scientific team of 18 individuals to focus on full
mixed-initiative conversational dialogue systems.  In this role, you
will research and innovate new techniques in Unified Hierarchical
Language Modelling for training conversational dialogue systems.  This
project is partnered with SpeechWorks International on the development
of new language modelling techniques.

We are developing a large-vocabulary TTS.  We will implement a patent
pending technique for training and unit-selection.  We will base the TTS
on an open-source standard, but productise it.  We have built an
automatic prosody markup parsing scheme, reducing the need for hand
annotated corpora for prosody.  You will extend this technique for
intelligent selection of speech segments using the ToBI technique.
Unique to the R&D effort, is to apply formant synthesis and acoustic
tube modelling techniques to interpret and join units.  We have
demonstrated human quality limited domain voices.

You should have a Ph.D. in Digital Signal Processing, Electrical
Engineering, Mathematics or Physics.  Industry experience is sought, but
you do not have had to work on TTS.  Radar signal processing, for
example, is fine.  You should be analytical and have demonstrated high
proficiency in mathematics.  A-levels in quantitative sciences should be
perfect scores or near perfect scores.

Contact David Horowitz, Chief Scientist
Vox Generation Ltd
www.voxgeneration.com
dhorowitz at voxgeneration.com


-----Original Message-----
From: David Horowitz 
Sent: 04 April 2002 12:16
To: Tadeusz Piotrowski; CORPORA at hd.uib.no
Subject: Corpora: 9 Scientist Jobs


4 Senior or Staff Scientists in Language Modelling
5 Senior or Staff Scientists in Adaptive Language Modelling

Work on a large international project hosted at Vox Generation Ltd.  Vox
is building a scientific team of 18 individuals to focus on full
mixed-initiative conversational dialogue systems.  In this role, you
will research and innovate new techniques in Unified Hierarchical
Language Modelling for training conversational dialogue systems.  This
project is partnered with SpeechWorks International on the development
of new language modelling techniques.

The Adaptive Dialogue Management will fuse multi-modal input with
conversational speech.  The ADM will model the user dependent behaviour
of preferred interaction modes, as will the language model.

We seek an individual with a Ph.D. in Computer Science, Physical
Sciences or Engineering/Mathematics.  Quantitative Computational
Linguists with experience in statistical language modelling or
statistical language processing sought.  Requirements include: excellent
programming skills in Perl.  Preferable skills in Java or C++.
Analytical background in mathematics or hard sciences preferred at
Bachelor's level.  Strong grades on A levels.

-----Original Message-----
From: Tadeusz Piotrowski [mailto:tadpiotr at plusnet.pl]
Sent: 04 April 2002 12:08
To: CORPORA at hd.uib.no
Subject: RE: Corpora: Wordsmith question


In the Windows environment there are a number of applications that allow
one to compare lists. There is a diff one for Windows, and there is the
file compare (fc) utility (far more versatile in Win NT and later).
There are also some tools inside other applications. I can recommend
Windows Commander (a sort of shell), which is very clear in showing what
the difference is, TextPad also shows you the differences. You can even
use Word, or a comparable word-editing package, to compare two texts,
including lists. 
Tadeusz Piotrowski 

> -----Original Message-----
> From: owner-corpora at lists.uib.no 
> [mailto:owner-corpora at lists.uib.no] On Behalf Of Rod de Lima-Lopes
> Sent: Wednesday, April 03, 2002 7:05 PM
> To: Van den Heuvel M Mev
> Cc: 'CORPORA at hd.uib.no'
> Subject: Re: Corpora: Wordsmith question
> 
> 
> Hi Maritza
> 
> I think you should use detailed consistencie tool (in the 
> wordlist programme), since it can tell you how quite 
> acuratelly how many lexicon is shared by two wordlists
> 
> go to wordlist > compararison > consistency (detailed
> 
> I hope it helps
> 
> []s Rod
> --------------------
> R.E. de Lima-Lopes
> http://rod.do.sapo.pt
> GNU/Linux User # 182240
> On Wed, 3 Apr 2002, Van den Heuvel M Mev wrote:
> 
> > Hi everybody,
> >
> > I'm having a spot of trouble with the Wordlist tool in the 
> Wordsmith 
> > suite that I hope someone out there can help me with. I want to 
> > compare two almost identical word lists containing the entries of a 
> > pronunciation lexicon. There are some inconsistencies between the 
> > lists, i.e. items missing in the one that should be in the 
> other and 
> > vice versa. I need to identify the missing words. I thought that I 
> > could use the "compare word lists" function in Wordlist for this 
> > purpose by setting the minimum frequency to 1 word, but it's not 
> > working. I'm obviously doing something wrong.
> >
> > If you don't have a quick answer to the Wordsmith problem, 
> but know of 
> > another tool that could help me do just this one little task with a 
> > few button clicks, I would also appreciate your response!
> >
> > TIA.
> >
> > REgards
> > Maritza van den Heuvel
> >
> >
> > Research Unit for Experimental Phonology
> > STellenbosch University
> > South Africa
> >
> >
> 
> 
> 




************************************************************@~|!+
This email and any attachments are confidential and are intended for the
addressee(s) only.
If you are not the intended recipient, please notify the sender
immediately by reply email and
delete this email and its attachments from your system. Any disclosure,
forwarding or copying 
of this email or its attachments is expressly prohibited.
Vox Generation Limited, 16 Palace Street, London SW1E 5JD, Registered in
England: 3937784.
************************************************************@~|!+



More information about the Corpora mailing list