Database Reference
In-Depth Information
represented as agent-internal digital code, e.g., seven-bit ASCII. This conver-
sion is instantiated in four basic variants, namely the speak mode and the hear
mode each in the modalities of vision (optics, writing) and audition (acoustics,
speech).
In the hear mode, today's systems of
speech recognition
convert external
acoustic surfaces into modality-free digital code, mainly by the statistical
method of Hidden Markov Models (HMMs). In the optical modality, today's
systems turn images of letters into modality-free digital code based on the
software of
optical character recognition
(OCR).
In the speak mode, today's systems of
speech synthesis
convert digitally rep-
resented text into artificially realized speech, usually by concatenating pieces
of recorded speech. In the modality of vision, there is the conversion from dig-
ital code to the familiar letter images on our computer screens, which may be
called
optical character synthesis
(OCS).
The conversion from a modality-dependent to a modality-free representa-
tion (hear mode) abstracts away from properties of the external surfaces such
as speed, pitch, intonation, etc. in spoken language, and font, size, color, etc.
in written language, which from a certain point of view may be regarded
as accidental. The result of this token-type mapping is a
surface template
(2.3.1, 2.6.1). Conversely, the conversion from a modality-free to a modality-
dependent representation (speak mode) must settle on which of these proper-
ties, for example, pitch, speed, dialect, etc., should be selected for the external
surface (type-token mapping).
If speak mode and hear mode utilize the
same
modality and are realized by
different
agents, we have inter-agent communication. Examples are agent A
writing a letter (speak mode, visual modality) and agent B reading the letter
(hear mode, visual modality), and accordingly in the auditory modality:
2.3.1 I
NTER
-
AGENT COMMUNICATION USING SPEECH
agent B in hear mode
unanalyzed
external surface
agent A in speak mode
surface
template
surface
template
auditory
modality
type
type
token
modality free
internal coding
modality free
internal coding
external world
Search WWH ::
Custom Search