[b-hebrew] Phonemes versus phones
jc.bhebrew at googlemail.com
Fri Jan 22 07:03:44 EST 2010
a few notes on my avoidance of the terms phonemes and phones. Below is a
A The Hebrews invented a phonetic alphabet
B No they didn't! They invented a phonemic one
This is a classic example of getting hung up over terminology. Phonemes are
groups of phones (allophones) and so some linguists may feel it is more
accurate to say 'The Hebrews invented a phonemic alphabet'. Whatever! The
general point is that they thought they were inventing a phonetic alphabet
but ended up inventing a phonemic one. That is to say they thought they were
inventing a system with one to one mapping between units of a speech signal
and the symbols of their alphabet. This evidently is an impossible feat and
when it comes down to it anybody with any real experience with processing
speech signals should know that any definition of phones is a useless one
also limited by powers of perception. Speech is a continuous signal. An
unsegmented waveform. Any attempt to segment the unsegmentable is an
exercise in futility. And yet the human perception is able to do it
elegantly. This is due to mechanisms of generalisation.
And so, all in all, it really is sad that some who consider themselves
higher educated linguists can get so hung up over terminology. Did the
inventors of the Hebrew languages know what a phoneme was? Did they know
what a phone was? Did they know that the distinction itself is subjective
when we start analysing real speech signals? Of course, they didn't! They,
in the simplest of terms, just attempted to map their perceptions of unique
'sounds' to unique symbols.
If anybody thinks that my attempts to make this explanation accessible by
avoiding ambiguous terms like 'phoneme' and 'phone' makes me a simpleton
(I'm talking to you Randall) then all I can say is this. Start working with
real speech signals and try to invent an exhaustive list of phones (not
phonemes) that works without exception in speech recognition tasks and you
will soon see that what you attempt to define as phones are what you have
defined as phonemes. Another useful exercise for you could be the converse.
Take your system of phones and try to generate a speech signal by
concatenating your phones together and see if a native speaker of the
language is willing to believe if a real person uttered the signal rather
than its being artificially generated. You will soon see that while we find
it difficult to make our perception explicit we can quite easily tell the
difference between such synthesized speech and a real speech signal.
All of this enters into the psycholinguistic realms of perception.
Perception is full of illusions. e.g. Did you know that when you listen to
somebody speak in a room that your ears receive the speech signal many times
as the signal bounces of the walls and arrives at your ears later than the
original speech stream? Probably not! Why! Because our hearing mechanisms
filter this noise out and our perception presents us only with the original
stream. But there is a limitation to this filtering mechanism. When the same
signal arrives too late for the filter to associate it with noise then we
perceive the bouncing retarded signals as an 'echo'. And so there are echoes
and there are perceived echoes.
In conclusion, all I can say to those theoretical linguists who think they
know better and get hung up of terminology which just doesn't work is this.
Turn your theoretical linguistics into practical linguistics. Take some real
speech signals and start working with them and see how useful your
definitions are to implementing a speech system that is capable of
a) recognising speech or
b) synthesising speech of high enough quality to fool humans
Then we might be able to start seeing eye to eye. Until then, I will
continue to make my arguments more accessible to the layman whose heart is
more open to understanding the simplicity of it all.
More information about the b-hebrew