Richard Reichenbacher INDV 101-64 2/25/05 500 word article summary

estonianmelonΤεχνίτη Νοημοσύνη και Ρομποτική

24 Οκτ 2013 (πριν από 4 χρόνια και 16 μέρες)

112 εμφανίσεις

Richard Reichenbacher

INDV 101
-
64

2/25/05

500 word article summary

Thierry Dutoit
. "A Short Introduction to Text
-
to
-
Speech Synthesis."
A Short Introduction
to Text
-
to
-
Speech Synthesis.

December 17, 1999. TTS R
esearch Team. 2/17/05<
http://tcts.fpms.ac.be/synthesis/introtts.html >.


In Thierry Dutoit’s essay, “A Short Introduction to Text
-
to
-
Speech Synthesis,”
Du
toit poses several different questions, but mainly what is speech synthesis
, what is
TTS

used for an
d what are the fundamentals needed to go about it
. He says that
there

is a
difference between the text
-
to
-
speech machines he is talking about and other machines
such as CD players and record players.

The type that he is referring to is a machine that
can

read back written text, whether it is dictated in a computer or scanned into an Optical
Character Recognition (OCR)

system.

I believe that either of theses two TTS systems can be very useful in an economic
perspective and for medical purposes. Dutoit b
elieves the same thing as he states some of
the more obvious uses, such as telecommunications, language education, aids to
handicapped people, talking books and toys, vocal monitoring (for business memos and
notes), multimedia and highly controlled researc
h purposes. There are many more
useful
sources out there that Dutoit does not cover but these are only the more common ones.

The main problems that occur with a natural language processing (NLP)
components are that it would be almost scientifically inconc
eivable
to create it.
There are
way too many factors such as being able for the machine to interpret what the sentence it
is saying means and the actual physics of the voice box and nasal passages of a human
being. There is also the problem of the immens
ely complex algorithms that the machine
would need to use to translate the written text into spee
ch.


The basic diagram that Dutoit uses to show a typical TTS system show: the text
being inputted into the machine’s natural language processing (NLP). The N
LP
translates the text into linguistic formalisms, inference engines and logical inferences.
That is that it translates the text into logical meaning and produces the linguistics of the
sentence. The next part is that the phonemes and prosody
of the sent
ence goes into the
digital signal processing. This part of the TTS system uses mathematical models,
algorithms and computations

to form the way that the machine “speaks” the written texts.
It formulates the way that a human would say it, under certain co
nditions and depending
on the way that it is programmed.

Dutoit goes into much more detail about how each part of the system works

and
the problems associated with the current methods of creatin
g each part of the TTS system.
His conclusion states that we
are far from having a perfect TTS system yet we have come
a long way in the past years.