RE: Looking for hints on speech generation (not synthesis)

["Mark Harrison" <Mark.Harrison@xxxxxxx>]

David,

I've been thinking on this email for the last couple of weeks, and still
haven't come up with a decent solution. So here are some rambling
thoughts...

Linguists tends to assume that languages are based around axiomatic
semantics. IT people tend to think in terms of denotational semantics,
and the models continue to be virtually impossible to reconcile.

As I see things, this question is bound up with (from a people end)
cognitive psychology, and (from an IT end) with AI.

There are a number of reference works on AI, which are obviously more
about the _understanding_ of natural language.  The traditional tool is
one called Lisa, which was originally implemented in common LISP back in
the 70s/80s. (Google search: "Lisa LISP")

There's still a lot of debate in AI circles about whether
"understanding" is necessary for meaningful AI. Turing obviously
postulated one theory about intelligence (Google searach: "Turing
Test")


Roger Penrose (professor of Maths at Oxford) believes that it is, and
that furthermore, a deterministic Von Neumann architecture is
fundamentally incompatable with "intelligence." His book
"The Emperor's
New Mind" is worth a read if you've got time.

However, I disagree with his central premise. I don't believe that
non-determinism is an absolute requirement to model conciousness.
However, I do agree that any congnitive model that simulates _human_
understanding involves an inherent parallelism an order of magnitude
above what is simulatable in real-time with current processors.=20

The best book I've ever read on this (conciousness, NOT the IT modelling
of it) is "conciousness explained" by Daniel Dennett, which is
_seriously_ worth reading.


Answers to your Home Automation Problem - NO.
Interesting things to read to keep your life busy - YES.

:-)


Mark Harrison
Head of Systems, eKingfisher


-----Original Message-----
From: David Paterson [mailto:david.paterson@xxxxxxx]
Sent: 27 January 2002 23:10
To: ukha_d@xxxxxxx
Subject: [ukha_d] Looking for hints on speech generation (not synthesis)


Hi all,

The text to speech side of things seems to be well understood and fairly
easy to implement, but I'm looking for any ideas or information on
speech, or rather language, generation.  i.e. using some kind of syntax
or grammar to generate appropriate sentences based on the information
input to the system.

The idea is that various information sources like house sensors, news
and weather from web or teletext etc. would be prioritised, possibly
based on preset importance or appearance of keywords, and then the
language generation would create sentences to feed the text to speech
system to pass on the information.

I've tried various web and google searches but haven't come up with
anything which would either provide a ready made solution or even good
ideas for home brew.

With the broad experience of the people on this list I'm sure somebody
can point me at some useful resources :-)  So, any ideas folks....

David P.

--=20
David_Paterson =3D david.paterson@xxxxxxx =A6
david.paterson@xxxxxxx;

Senior R & D programmer    There are three kinds of people in the world
-
Visual Sciences Ltd.        those who can count, and those who can't...


For more information: http://www.automatedhome.co.uk=20
Post message: ukha_d@xxxxxxx=20
Subscribe:  ukha_d-subscribe@xxxxxxx=20
Unsubscribe:  ukha_d-unsubscribe@xxxxxxx=20
List owner:  ukha_d-owner@xxxxxxx=20

Your use of Yahoo! Groups is subject to
http://docs.yahoo.com/info/terms/=20



________________________________________________________________________
This email has been scanned for all viruses by the MessageLabs SkyScan
service. For more information on a proactive anti-virus service working
around the clock, around the globe, visit http://www.messagelabs.com
________________________________________________________________________