~d
of
Computational
Linguistics
A
C A S E
H I S T O R Y
G D M P U T E R
E X P L O R A T I O N
F A S T
S P E E C H R U L E S
Douglas
B.
M o r a n
"S;.
of
C o m p u t e r
&Communication
Sciences
University
of
Michigan
Ann
A r b o r ,
Mlchigan
C o p y r i g h t
1 9 7 5
A b s t r a c t
-
-In c o n v e r s a t i o n a l
s p e e c h ,
words
r u n
t o q 6 ther
a n d
i n t e r a c t
c a u s i n g
t h e i
p h o n o l o g i c a l
f ~ r m s
t o d i f f e rfrom
t h e i rc i t a t i o n
forms.
F a s t s p e e c h
r u l e s
a t t e m p t t3 q e s c r i b et h e s e
c h a n g e s
ass p e e c h
becomes
f a s t e r
a n dmore
c a s u a l ,
In
d e v e l o p i n g
3n y
s e tof
p h o n o l o g i c a l
rules,
c o m p u t e r i z e d
grammar
t e s t e r s
a r e
au s e f u l
andi m p o r t a n t
a i d . T h e yn e c e s s i t a t e
(3.
p r e c i s e ,
a o n s i s t e n t E o r n u l a t i o n
of the
r u l e s a n dallow
t h eg e n e r a t i o n
of
s - p l e
d . e r i v a t i o n s .
I n
a p p l y i n g
these
~ u l e s
to
ad i v e r s e
setof utterances,
w e can
f i r s t
confirm
t h a t
t h e
r u l e s
r e a l l y
d c
a p p l ywhere
wee x p e c t
them to,
a n d
t h e n
experiment
v i t h
v a r i o u s
r u l e
o r d e r i n g s to o b s e r ~ e t h e i r e f f e c t s .T h e
P h o n o l o g i c a l
G r a m m a r
Tester
of F r i e d m a n
a n d
Morin
was
use-d
t o t e s t
t w o s e t s Of*f a s t
s p e ' e c h
rules from
t h e
A R Q A S p s e c hU n d e r s t a n d i n g
R e s e a r c h
c c m m u n i t y .
Working
w i t h
Chese
rules
l a d
t o
c e r t a i n o b s e r v a t i o n s
a b o u t t h ei n t e r a c t i o n s and
n a t u r e
of
f a s t
s p e e c h
r u l e s
in
g e n e r a l .
3x1 a d d i t i o nt o
t e s t i n gt h e s e
t w osets
of
f a s t s p e e c hr u l e s ,
wewere also
i n t e r e s t e d
in
t h ep r o b l e m s
o f
t e S t i n g
such
agrammar
v i t h
t h i sprogram.
A p p e h d i c e s i n c l t l d e
an
overview
of
t h egrammar
tester
a n d
Table
of
C o n t e n t s
---.lC
...
T e s t i n g
t h e
F a s t
S p e e c h
R u l e s
...
5
...
Sutgqested Wodif
i e a t i o n s
to
t ' h eP G T
10Observations
a b o u t
Fast
S p e e c hRules
...
1 2References
...
1 5A p p e n d i x A
.
Using
the
P h o n o l o g i c a l
Grammar
Tester
...m...m...
18.
A p p e n d i x B S a m p l e O u t p u t
.m...
2 1.
A p p e n d i x
C F i r s tSet of
R u l e s
...
2 4...
4
fi Case
History
i n
Computer E x p l o r a t i o n
o f
F a s t S p e e c h
R u l e s
C o u g l a s
B.
Moran
The
U n i v e r s i t y
of
M i c h i g a n
I n t r o d u c t i o n
- - - I - - - - * -
In
d e v e l o p i n g
a n y
sebt
o f
p h o n o l o g i c a l
r u l e s , computerized
grammar
t e s t e r s
are
a
useful
a n d
i m p o r t a n t
a i d .
T h e y
n e c e s s i t a t e
a
precise,
consistent
formulation
of
t h e
r u l e s
a n d
aL1ow t h eg e n e r a t i c n
of
sample
d e r i v a t i o n s .
I n a p p l y i n g
t h ~ s e
r u l e s
to
ad i v e r s e
set
of
utterances,
we can
f i r s t
confirm
t h a t
t h e
rules
r e a l l y
d o
3 p p l y
where
ue expectt h e m
to,
a n d
t h e nexperiment
w i t h
v a r i o u s
rule
orderings
t oobserve
t h e i r
e f f e c t s ,
T r a d i t
iona
1
p h o n o l o g i c a l
r u l e s
d e s c r i b e
t r a n s i t i o n s
from
the
u n d e r l y i n g
form
t ot h e
surface
form o f
aword.
However,
in
casual
(fast) speech,
words
are
n o t
d i s t i n c t , i n d e p e n d e n t
u n i t s ,
but
r u n
together
a n d
i n t e r a c t .
F a s ts p e e c h
rules
a t t e ~ p t *-od e s c r i b e
t h e s e
c h a n g e s
t h a t
occur
as speech
becomes
f a s t e r
a n d
more
c a s u a l
( 4 ,8,
9 ) .T h i s
is
a.
case# h i s t o r y
of
t h e
examination of
t w op a r t i c u l a r
sets of
r u l e s
on
the
P h o n o l o g i c a l
GrammarTester
(PGT) of
Friedman
a n d
Borin
(3)
.
Anoverview
is
g i v e n
in A p p e n d i x
Aa n d
as a m ~ l e
of t h e
output
i n
A p p e n d i x
8.In
a d d i t i o n
t ot e s t i n g t h e s e
two
s e t sof
f a s t
s p e e c h
rules,
we
were
also
i n t e r e s t e di n
the
Testing
the
Fast Seeech
Rules
-
-
- - I -- - - C m I I I L I . I .F o r
our
f
jc s t t e s t ,
w
chose
t h e
fast
s p e e c h
r u l e s
of
Neu
(6) a s
the
b e s t
a v a i l a b l e
s e tof
r u l e s
g i v e n
i n
a
r e a s o n a b l y .
c o n s i s t e n t
n o t a t i o n .
The
r u l e s
Mere
g i v e n
in
t w oformats:
one
u s i d g
phonemes
a n d
t h eother
u s i n g
d i s t i n c t i v e
f e a t u r e s .
T h e
f i r s t
t a s k wasto
a d j u s tt h e
n o t a t i o n
i n t o
amachine
u s a b l e
form.
For
t h e
purposes
of this
t e s t , we usedt h e
feature
s y s t e hof
Chomsky
a n d
H a l l e
('I),w i t h
three
t y p e s
of
b o u n d a r i e s : s y l l a b l e
( +
,
morpheme
( 3 )
a n d word
( # I )
.
For example,
t h e
f
eatur'es
a n t e r i o r
---a
and
---
coronal
h a d
to
be
s u b s t i t u t e d
f o r
&ace'
_ofa r t i ~ u 1 a t i . o ~ .
A l s o ,t h e
r u l e s
h a d
t o
b e
m o d i f i e d
sot h a t
each
f e a t u r - e had
t h e same
t y p e
of
v a l u e ( i n t e g e r ,
b i n a r y ,
g r
marked-
unmarked)
t h ~ o n g b o u t t h e
grammar.
For
e x a m p l e ,
i n
the
PG?,
l s t r e s s
is
not
e q u a l
to
+ssess
and
is
less t h a n
=stress,
w h i c h
---
fc o u l d
causem a n f
unexpected
p r o b l e m s
i n
the
grammar.
Another
minor
problem
wasaetermining
w h e t h e r
aC
Ii n
t h e
rules meant
not
----.-- a vo we1,
I+
consl,
or
up--
v o c a l l .S a m p l e
d e r i v a t i o n s
were
g i v e n
w i t h
t h erules,
and t h e s e .
were
.used as a
f i r s t
t e s t .We
q u i c k l y
f o u n d
anumber
of
small
p r o b l e m s ,
most
being
t h a t t h e
rule
had
b e e n
incorrectly
or
incompletely
s t a t e d .
For example,
Geminate
R e d u c t i o n
a n d
A l v e o l a r
F l a p p i n g
would
not
a p p l y
across
t h e
word boundary
i n
"want
to",
w h i l e t h e
E
i n
" c a p t a i n 1 v
b l o c k e d t h e
a p p l i c a t i o n
of
lott tali eat ion.
erro-1s
were
f a u n d ,
such
a aNasal
Consonant Deletion
b l e e d i n g
T r a n s i t i o n a l
S t o p
I n s e r t i o n
and
thereh
y
i n d i ' c a t i n g
t h a t
it
s h o u l d
be
ordered
a f t e r
Transtop.
W e
s t i l l
had
SOB-e
u n r e s b l
vedproblems,
such
a s"Ts
/f
e da
1/
aproper
f a s t
s p e e c h form
of
' f . e d e r a l ' ? l g ,
when
t h e
r e v i s i o n s a n d
a d d i t i o n s
(7)
a r r i v e d ,
Manyof
thes
er e v i s i o n s corresponded
to
o u r
s o l u t i o n s
of
problems
encoun\telCed
d u r i n g
t e s t f
ng o f
these
r u l e s .
The
f
i a a l form
of
t h i s f i r s t
sotis
g i v e n
i n
l p p e n d i x
C.
S i n c e
there
wasl i t t l e
o b v i o u sorder
in
the r u l e s ,
g r a p h i n g
the
~ a r t i a l
o r d e r i n g
g i v e n
wasvery
h e l p f u l
in
e x p l o r i n g
the
r u l e
o r d e r i n g .
T h e
p a r t i a l o r d e r i n g g i v e n
w i t h
the r u l a s ,
Ficrure
1,was
q u i c k l y r e f i n e d
t o that
i n
F i g u r e 2
by
t e s t i n g
t h e
s a m p l e
d e r
i
va'ti
cns
.
In
these
n e w
'rules,
p a r t i c u l a r
a t t e n t i o n
was
p a i d
t o
t h e
presence
a n d
p o s i t
i o nof
s y l l a b l e , a o r p h e n e
a n d
word b o u n d a r i e s
i n
t h e
e n v i r o n m e n t s
for
w h i c h
t h er u l e s
s h o u l d a p p l y .
O t h e r
f a s t
speech
rules
from
members.
of
t h e
A R F AS p e e c h
U n d e r s t a n d i p g
Research
( S U R )community
(11, 12, 1 3 )vere
used3s
references,
e s p e c i a l l y
in
respect
t bb o u n d a r y
p o s i t i o n ,
i n
i m p l e d e n t i n g f h
asenew
rules.
In
t e s t i n g
the
r u l e s ,
most
o ft h e
p r o b l e m s
i n v o l v ' e d
t h e
p o s i t i o n i n g
and poesence
of
b o u n d a r i - e s .
T h e
second
D e n t a l D e l e t i o n
r u l e
wasdropped
from
t h e
grammar
because
it
a p p e a r e d
t cd u p l i c a t e
the effect
o f
a s s i m i l a t i n g
a n
a l v e o l a r
f l a p p e d
g
t othe
p r e c e d i n g
5.
G R A P H O F P A R T I A L O R D E R I N G
-
-L I N E A R I Z E D
-....---
SCHWA D E L E T I O N < I N G > R E D U C T I O N
R U H - R E D U C T I O N R U H - L E S S N E S S
N A S A L I Z E D VOWELS N A S A L C O N S O N A N T
EEL
ET I0
N
D E N T A L
DELETION
A L V E O L A R F L A P P I N G
G L O T T A L S T O P FoRHqTIoN
P A L A T A L I Z A T I O N
A
SSI
31 LATZO
N
TFiAN S I T I O N A L
STOP
INS E R T I O N
G E K I N A T E REDUCTION
D A R K
[ L J
[ RJ-FLAPPING
F I G U R " 1 I N I T I A L R U L E O R D E R I N G
-L I N E A R I 7 ; E D
---
VOWEL R E D U C T I O N
SCHWA O E L F T I O N
< I k G >
R E D U C T I O NR U H -
R E D U C T X O NN A S A L I Z E D V O W Y L S
T R A N S I T I O N A L STOP
I N S E R T I O N
P A L A T A L I Z A T I O Y
A L V E O L A P
F L A P P X
:
!
N
GLOTTAL S T O P
F O R M A T I O N
D E N T A L D E L E T I O N
G E H X
N A T E RS'D U C T I O NN A S A L C O N S O N A N T
REGBCT I O N
D A R K
[ L ]G R t A P H O F P 4 R T I : A L O R D E R I N G
--..-.----..---
---
F I G U R E 2 P I N A L RULE
ORD~~~l4
testing
t h e s e
rules
a
major problem
wasi n
d e t e r m i n i n g
wha-t
d i d
and d i d
n o tc o n s t i t u t e proper
fast speech
forms.
I n
slow
speech,
it
is
f a i r l y
e a s yto
break
up
the
utterance i n t o
asequence
ofp h o n e m e s ,
b u ti n
f a s t
s p e e c h ,
the
utterance
is
more
c o n t i q u o u s
a n d
it
is
harder
t o
d e c i d e
what
is
a
segment
a n d
wh.ere t h e
segments
b e g i n
a n d
end.
For
example,
t h e
p a l a t a l i z a t i o n
rules
produce t h e f o l l o w i n g
d e r i v a t i o n :
( a )
.
/ d i d # C y u /
b )
.
/ d i # #
yyu/( c ) .
/ d i t t y u /
We
a r e
n o ti n c l i n e d
t ob e l i e v e
t h a t
a n
i n t e r m e d i a t e
form
/ d i d # # Y y u /
exists
b e t w e e n
s t e p s
(a)a n d
( b )in
t h i s
d e r i v a t i o n .
I
t h i n k
t h a t
(b)
i s
a
t r u e
i n t e r m e d i a t e
form,
existing
i n
slower,
casual
speech,
but not
i n f a s t
speech.
O u r
b e s t
e v i d e n c e
i n d i c a t e d t h a t
t h e
R U H
i st h e
same a st h e
s y l l a b i c
"rU
It h u s
r a i s i n g
the
p r o b l e m
of
ordering
t h e
S y l l a b i c i z i n g
a n d Ilu
h - r s d u c t i o n
rules.
Arethe
proper
f o r m s
/g
u
v+-
n
r/
ahd
/ fe
d
+
r
l/
or,
more likely,
/gu
v
+
n
r/
0
a n d
/f
e
d
+
r I/?
Q c ,c o u l d
it
p o s s i b l y
be
/gu
v+
n
r/
a n d
/ f e
d + r I./,w i t h
t h e
S y l l a b i c i z i n g
r u l e n o t
deleting
the
schua, b u t
n a k i n g
it v e r y
short
a n d
p a s s i b l y
devoicing
it?
Another d i s t u r b i n g
r e s u l t
of
t h e
testing
v a sthe a p p l i c a t i o n
of
rules
across deleted
s e g m e n t s .
For
exaaple, after
qqvernqg
has
b e e n
r e d u c e d
to
/ gu v
+
n
r/,
Proqressive
A s s i m i l a t i o n
applies,
producing
/g
u v
+
m
r/.
Similarly,
----
s a n i t q
---
clause
c l o s i n g
i n
on
a.fast
speech
form
of
s a n t a
C l a u s .
T h e final
form
of
t h e
second
s e tof
r u l e s
a n d the o u t p u t
are
given
i n
A p p e n d i x
D
6Suggested
& d i f i c _ g t i o n s
to
the
PGT
-
---..-
---I-T h e
PGT wasw r i t f e n f o r t h e
t e s t i n g
o ft r a d i t i o n a l
p h ~ n o l o g i c a t
grammars
w h i c h
t a k e
a n u n d e r l y i n g
form
t o
aBurface
form,
a n d
t h u s
fast
speech
r u l e s
p r e s e n t
s e v e r a l
u n f o r e s e e n
problems.
F a s t
speech
r'ules
do
not
c o n s t i t . u t ea
c o n p l e k e
grammar,
b u t
a r ei o s t e a d
i n t e r s p e r s e d ,
among
t h e
t r a d i t i o n a l
r u l e s ,
Ln
p r o d u c i n g
an
i n t e r m e d i a t e
or
s u r f h c e
for^f o r
i n p u t
to
t h e
grammar,
t h e
use of
d i a c r i t i c s
would
p r a d u c e
2simpler
a n d
moreeconcmical
represent
ation
st
t h e
i n p u t
F o r
e x a m p l e ,
i n
the
p r e s e n t
system,
it
is necessary
to
give
as e p a r a t e
d e f i n i t i o n
f o r
each
stress
a s s i g n m e n t
for
each
v o w e l ,i n s t e a d
o f
d e f i n i n g
each
towel
cnce
a n d
a s s i g n i n g
t h e s t r e s s through t h e
useof
d i a c r i t i c s .
S y l l a b i f i c a t i o n a D p e a r s
to
o c c u r before
someof
t h e
r u l e s ,
an
dafter
o t h e r s .
TO
a v o i d
h a v i n g
to
i n s e r t ,
r u l e s
for
s y l l a b i f i c a t i o ~ ~
or
h a v i n g
the
r u l e s
p r e c e d i n g s y l l a b i f i c a t i o n
ignore
s y l l a b l e b o u n d a r i e s
i n t h e
i n p u t ,
the
PGTs h o u l d
h a v e
al@human
i n t e r v e n t i o n "
r u l e w h i c h w o u l d a l l o w
the
user
to
c h m g e
WmissingfI
rules.
"Human
interventionn
rules
c o u l d b e u s e d
t o
postFone Q r i t i n g
rules, either
ro
al a t e r s e s s i o n
or i n d e f i n i t e l y
if
the
rules
are
t o oh a t d
to
write
or
if t h e y
are
& a t s i . d e t h e
scope
of
t h e
problem.
Another
problem
a r i s e s
from
most,
if
not
a l l ,
f a s ts p z e c h
r u l e s
b e i n g
o p t i o n a l .
T h e
present
systemgenerates
as i n g l e
d e r i v a t i o n
u s i n g
aando om
number
g e n e r a t o r
to
d e c i d e
w h e t h e r
or
not- t o
a p p l y o p t i o n a l
rules.
Asa
result,
t o
obtain
a good
b e s tof the
e f f e c t
o f a n optional rule
i n
t h o
grammar,
we
h a v e t odo
two
tests,
O h e w i t ht h e
rule
o b l i g a t o r y
and
one
w i t h o u t
t h e
rule.
With.
alarge number
of
optional r u l e s
o p e r a t i n g
on
ar r o n - t r i v i a l
set
of
test data,
h a v i n g
to
test
e a c h
possible
rule combination
i n d i v i d u a l l y
would
b e
a-staggering
t a s k .
Amore
satisfactory
method
w q u l d
be
for
t h e
PGTto p r o d u c e
aset
o f
d e r i v a t i o n s s u c h
t h a t
a n y
t i n ea n
o p t i o n a l
rule
can
a p p l y ,
t h e
PGT
produces
ad e r i v a t i o n for
t h e
casewhere
the.
rule
wasa p p l i e d
a n d
one
f o r
the case
where
it
wasn't.
Cohen
ana
Mercer
(2)
have
i m p l e m e n t e d
s u c h
af e a t u r e
i n t h e i r
rule
tester
b y
storing
the
r e s u l t
of the
a p p l i c a t i o n
of
ar u l e
a s adirected gra.ph.
S u b s e q u e n t
r u l e s
are
a p p l i e d
t oa l l
p a t h s t h r o u g h
t h eg r a p h ,
p r o d u c i n g
a
neu graph.
In
c o n s i d e r i n g
the problem of
when
a n
o p t i o n a l
r u l e
s h o u l d
be
a p p l i e d ,
weare
c o n s i d e r i n g p g f o u a n c e
(10).
To t h i s
end,
we
m i g h t
be b e t t e r
served by
h a v i n g
t h e
v a r i a b l e
r u l e s
of
Iabov
(5)
ia a d d i t i o n
t o t r a d i t i o n a l p h o n o l o g i c a 1
r u l e s .
f
n s t e a d
of
the
a l l -
or-
n o t h i n g
m a t c h i n g
of
normal r u l a s ,
v a r i a b l e
rules employ
a p p l i e d b e i n g
af u n c t i o n
of
t h edegree
o f
match,
t h e
r a t e
o f
speech,
and
the
n a t u r e
of
t h ec o n v e r s a t i o n
( e , g .Is
i tformal
or
c a s u a l ? )
.
For
example,
for:
t h eFroqrelssive A s s i m i l a t i o n
mile
for
p o i n t
of
articuigtion
(PROARTICJ
,
/ sO
f
n/
((soften)
becomes
/S (3
f m/
only
i n
v e r y
f a s t speech,
h u t
/ p
r
+
f yu
+
n/
v N
( p r o f u s i o n )
becomes
/p c +f
yu
+ zn/
i n
sost
f a s t
s p e e c h .
I n
a d d i t i a n ,
f o r
eachv a r i a b l e
rule
a p p l i e d ,
t h e
PGTs h o u l d
usethe
a s s o c i a t e d p r o b a b i l i
t v
t cproduce
aE i n a l
p r o b a b i l i t y
f o r
each,
d e r i v a t i o n .
U n f o r t u n a t e l y r
t h e
i l b p l e a e n t a t i o n
of
v a r i a b l e
r u l e s
would
n e c e s s i t a t e
major
c h a n g e s
to t h e
PGT.C o n v e n i e n t ,
b u t
not
n e c e s s a r y ,
w o u l d
bet h e
ability
t o
make
a r u l e
" b l i n d wt o certain
t y p e s
of
n o d e s ,
For
example,
Chomskp
a n d
Halle
(1)a l l o w
sosleb o u n d a r y
symbols
t o
be
invisible
t o
t h e i r
phonological
rules.
T n
some
cases,t h i s
a b i l i t y
m i g h t
be
a b l e
t of u l f i l l
t h e
f u n c t i o n
of
a
" h u m a n i n t e r v e n t i o n v r u l e ,
O b s e r v a t
i o n s a b o u t
F a s t
Skeech
R
ules
---.----
---
T e s t i n g
t h e
fast
s p e e c h
r u l e s
led
ta
certain
i n s i g h t s
into
t h e
interactions
of
t h e
r u l e s .
The
most
v e x i n g problems verbe
with
t h e r e p r e s e n t a t i o n
of
b o u n d a r i e s .
The use
of
2 ,
&,
a n d
g t
seems
a r t i f i c i a l
a n d
cumhersome,
Amoce n a t u r a l
a p p r o a c h
seemst o b e
to
a s s i g n
t h e
b o u n d a r i e s
v a r y i n g
degrees
of
$treggf_h,
a n d
t h e n
allow
t h e
p h o n o l c g i c a l
rules
t oo p e r a t e
on
t h e
s t r e n g t h
assignments.
Consider
t h e problem
of
HJohngs
going
t o
uorkll
where q o i n q
i s
a n
auxiliary
b u t not i n t h e
caseuhere
it
is
t h e
v e r b .
Rather
t h a n h a v i n g the
p h o n o l o g i c a l
rule s e n s i t i v e
to
t h e
s y n t a x ,
-it
seemsmuch
c l e a n e r
to
h a v e
it
s e n s i t i v e
t pt h e
b o u n d a r y
strength
between
o
n
a n d
to.
To
t h i s
end,
we must
allow t h e
s y n t a x
t oi n f l u e n c e
t h e
i n i t i a l
a s s i g n a e n t s
of boundacy
s t r e n g t h .
The
use ofboundary
strengths also
seems
to
s o l v e
t h e
problem,
of
how
boundaries
a r e
r e d u c e d .
For
example,
in
r e d u c i n g
want
to
t o
wanna
s h o u l d
t h eword
b o u n d a r y
b ereduced
t oa
me--- ---I
morpheme
or
a
s y l l a b l e
bounda'ry?
A n o t h e r
aseof
b o u n d a r y
s t r e n g t h s
m i g h t
be t o
mark
t h e
p o s i t i o n
of
segments
d e l e t e d
fcom aword
b y
s t r e n g t h e n i n g
t h e
n e i g h b o r i n g
boundaty.
For
e x a m p l e ,
when
governor
---
i sreduced
to
s v
Iner,
the
p
is
l e n g t h e n e d ,
i n d i c a t i n g t h e p o s s i b l e
presence
o f
a
stronger
then
normal syllable
b o u n d a r y .
This stronger boundary.
w o u l d
t h e n p r e v e n t
the
from
c h a n g i h g
t h e
g
t o a n
-
m
u n d e r
Progressive
A s s i m i l a t i o n .
An
a l t e r n a t i v e
is
t o
h a r e
a" s t r o n g
s y l l a b l e w boundary
s y m b o l w h i c h
w o u l d
b e "strongerwq
t h a n
a"weak
wordm bounaary
symbol,
b u t
t h i s
woulil
be
mdssyand
u n n a t u r a l .
It
voola
p r o b a b l y
b e
more
productive
t o
represent
f a s t
s p e e c h
forms
w i t h
phonemes
u t i l i z i n g m u l t i - v a l u e d
(1
t h r o u g h
7i n s t e a d
of
+
a n d
)f e a t u r e s .
G i v i n g
the segments
l e n g t h
a t t r i b u t e s
would
a l l o w s e g m e n t s
t o
be
d e l e t e d
b y
acombination
of
r u l e s
which reduce
segment l e n g t h s .
Also,
i n c o n p a r i n g
forms,
av e r y
short
seqment
could be
considered
t obe
d e l e t e d .
w h i c h t h e y
can
a p p l y .
Z w i c k y
(14)
g i v e s e x a m p l e s
a n dc o u n t e r -
e x a m p l e s
for
v a r i o u s
f a s t
speech
r u l e s ;
However,
it
seems
t h a t
t h e
counter-examples
a r e
words
w h i c h
do
n o tf t e q u e n t l y
occ,ur
i n
&aqua1
speech,
w h i l e
the
e x a m p l e s
were
common.
Thiso o i n t
wasi l l u s t r a t e d when
,
i n a n e a r l y
run,
t h a t
SchvaD e l e t i o n
p r o d u c e d
/k
a e
n
+
g
r
u/
from
'tkangaroow,
A Tfirst, t h i s
East speech
form
was u n a c c e p t a b l e ,
b u t
w i t h
u s e ,it
b e c d m e
more
a n d
more
acceptable.
Interestingly
e n o u g h ,
a
s m a l l
r a n d o m
p o l l i n g
s e ; n e d
t o i n d i c a t e
t h a t
this
form is
a c c e p t a b l e
"if
y o u tre
a n
AUS
t r a l i a n t r
.
T h a t
fast
speechr u l e s a p p l y
t~
what 1 scommon
i n
the
speaker's
v o c a b u l a r y
is
also
e v i d e n c e d
b y
the
h e a v y
a p p l i c a t i o n
of
f a s t
speech
r u l e s
t o
l o c a l
names.
T h e
similarity
af
the
p a l a t i z a t i o n
r u l e s i n
t h i s f a s ts p a e c f i
grammar
a n d
i n
n o r m a l F n g l i s h ( 1 ) r a i s e st h a
q u e s t i o n
o f
W h a ta r e
' f a s t s p e e c h
rules?".
If
normal r u l e s are r e s t r i c t e d
v e r s i o n s
of
t h e
moreg e n e r a l
p r o c e s s e s
w h i c h
a l s o
a c c o u n t
for
t h e
P a s t
s p e e c h
forms
(14),
t h e n
we
must
be
c a r e f u l
i n using
slow s p e e c h
s u r f a c e
forms
to
t e s tt h e fast
s p e e c h
r u l e s .
F o r example,
p a r t
of
the
d a t a
covered
b y
t h e
Schwa Deletion
r u l e
m i g h t
a c t u s l l y
result
from
schwas being *ser&gq
b e t w e e n
c o n s o n a n t s
i n
slow
y o r k was s u p p o r t e d u n d e c
N a t i o n a l S c i e n c e
F o u \ n a a a i i o n
1 3 0 9 , w i t h t h e supervision a n d q u i d a n c ~of
J o y c eReferences
---
sky, Noam G
Y o r r ~ s
H a l l e , The S2gm&-Patternof
E2gL~l.&,rper a n d
Row,
Yew Y o r k , Yew Y o r k , 1 9 6 8Re P . S . &
R.
L.
Mercer,
Y " e P h o n o l o g i c a l Component o f PAutomat
ic S ~ e e c h - R e c o g n i t i o n
S y s t e m f f , C o n t r i b u t e dears.
-
I E F E
Sypl~osium
on
SpeechRecognition,
Lee E r m a nB , 1 9 7 4
mlan,
J o y c ea
nd Yve s C h .n o r i n ,
ghonejhqiial
c;rggmggDescription,
B
Per:
---
Rat u r a l
L a n g u a g e S t u d i e sNo.
9 ,~ n o t i c s L a b o r a t o r y , The University
of
M i c h i g a n , 1 9 7 1Ilk
an,
J-oyce, vaCcmputer
Explorationof
F a s t S p e e c hIesm, I E E E
T r a n s a c t i o n s
on
A c o u s t i c s , S p e e c ha n d
S i g n a l $Zessig.q, A S S P - 2 3 , 1,F e b r u a r y
1 9 7 56.
Neu,
Helene,
"SomeOptional P h o n o l o g i c a l
Rules Occurring
i n
C a s u a l
Speech",
Speecn
t o m a u n i d a t i o n
Research
Laboratory,
S a n t a
Barbara,
C a l i f o r r i a ,
S U RNote
9 5 ,
N I C 1 7 6 7 7 ,J u l y
11, 19737 .
Neu,
Helene
anll
Beatrice
O s h i k a ,
* O p t i o h a l
Phonological
Rules
-
Revisions an?
Additions",
Speech Comrnunica
L i o n s
Research
Laboratory,
S a n t a
Barbara,
C a l i f o r n i a ,
SUR
Note
1 2 2 ,
N I C 21717,February
5.
1974
8.
OIMalley,
Michael H.
a n d
Alan
C o l e ,' V e s t i n g P h o n o l o g i c a l
R u l e s m , C o n t r i b u t e d Papers
-
I E E ESymposium
on
S p e a c h
R e c o g n i t i o n ,
Lee
Erman
(ed)
,
19749 .
Oshika,
Beatrice
f . ,
Victor
W.
Zue,
Rollin
V.
Weeks a n d
Helene
Neu,
" T h eRole
of
P h o n o - l o g i c a l
Rules
i n
S p e a c h .
U n d e r s t a n d i n g
Researchd,
IEEE Transactions
on
Acoustics,
Speech
and
Signal P r o c e s s i n g ,
A S S P d 2 3 , 1 ,F e b r u a r y
197510.
Robinson,
J a n e
J,
,
wPerformance
GramatarsM,
Technical
Note
TN
97,
Stanford
Research
I n s t i t u t e ,
197U,
t o 'appear
i n
S-eeech
p r e s e n t e d
at
t h e ,
1 9 7 4I E E E
R e c o s ~ i ~ ~ I ~ ~ f i e d P a p e r s - -
---
---
---
Syemosi
urn
Acadenic
Press
-
7-I11.
Shockey,
Linda,
u n p u b l i s h e d
collection
of
r u l e s ,
Dept.
o f
Computer
S c i , e n c e ,
C a r n e g i e - M e l l o n
U n i v e r s i t y ,
Pi
t t s b g r g h
,
12,
W e e k s ,
R o l l i n
V.,
'9sDCP h o n o l o g i c a l
R u l e s w ,
SystemsD ~ ~ e l o p m e n t
C b r p . ,SUR
Note
t 1 9 , N I C 2 1 4 9 0 ,J a n u a r y
3 1 ,1974
13. Zue,
Victor,
W.,
" O p t i o n a l P h o n o l o g i c a l
R u l e s n ,
M I TL i n c o l n
Laboratory,
L e x i n g t o n ,
Mass.,SUR
Note
1 2 4 ,NIC 21952,
F e b ~ u a r y 19, 1974
1 4 . Z w i c k y ,
A r n o l d
N.,"Note
on
a
~ h o n o l o g i c a l
Hierarchy
i nXngl,ishit
,
i n
R.
P.
S t o c k w e l l
a n d
R.
K .S.
~ a c a u l a y
(eds .)
,
L & h q u i s t
&
C h a n 4 e
and
Gen9ratij.e
Theory,
Indiana
&mend
j x
A-
U s i n %
t
h eP
hono&oqjcaJlG
rdmgqg-Tes
ter
-
- -
-
--
----*In
t h e
PG?,
t h e
phonological
r u l e s
are t r e a t e d
asa
t r a n s f ormat
ional
grammar
w h i c h
o p e r a t e s
on
t r e e s whose
terminal
n o d e s
are
u s u a l l y phonemes
a n d
b o u n d a r y syabolsb..
Y h i . 1 ~most
p h o n o l o g i c a l
rules
uill
use
o n l y
these
t e r m i n a l
nodes,
i t m a y be
d e s i t a b l e ,
or
e v e n
necessary, t o h a v e
t h erule s e n s i t i v e
to
t h e
s y n t a x ,
w h i c h
c o u l d
indicated
b y
t h e
n o n - t e r m i n a l
n o a e s .
Trees
can be
l i n e a r i z e d
by
U s i n g
I t < " and l g > v t t oenclose
aleft
to
r i g h t
l i s t i n g
of sibling
nodes,
such t h a t
ad e s c r i p t i o n
of
each
noh-term5nal n o d e ' s d e s c e n d a n t s
is
a p p e n d e d
to
its
name.
thus:
becomes
S<A B < D 5 )C>.
In
d e f i n i n g t h e
grammar,
r u l e s
a r e
stated
i n
one
of
two
foras.
Themore g e n e r a l
t r a n s f o r m a t i o n a l
rule
formath a s
twoparts:
t h e
s t r u c t u r a l
d e s c r i p t i o n ,
w h i c h must
m a t c h
t h e
tree
e x a c t l y
for
t h e
r u l e
t oa p p l y ,
and
t h e
s t r u c t u r a l c h a n g e , v h i c h d e s c r i b e s
the e f f e c t
of
t h e
rule
upcn
the tree.
S i n c e
these p h o n o l c g i c a l
rules
use
only
the
immediate
c o n t e x t ,
t h e
S D ' sb e g i n
a n d
end
w i t h
the
unbounded
skip
symbol
"Xn.Bounded
s k i p s
a r el i n e a r i z e d
b y
p r e f i x i n g t h e s y m b o l
w i t h
its
s e p a r a t i n g
them
w i t h
commas
a n d
e n c l o s i n g
t h e m
i n
p a r e n t h e s e s .
\ n
o p t i o n a l
part
o f
tho
structural
d e s c r i p t i o n
is
e n c l o s e d
in
p a r e n t h e s e s
IN o d e s
usedin
t h e structural
chang-e
are
referenced
by
an
i n t e g e r
p r e f i x .
If
o n l y
one
node
i sto be
e r a s e d ,
changed
or
i n s e r t e d ,
t h e
second
formof
t h e r u l e , w h i c h
resembles
t h a t
t r a d i t i o n a l l y
used
in
p h o n o l o g y ,
c a n
be
used.
Example:
VOWTL
N A S A L I Z A T I O NTRANS
NASVOW.sn
'V
( ( # ) )*
(+CONS
-VOCAL* + N A S A L 1%.
SC.
( + N A S A L (NERGEP
I ,
The,rules
can
e i t h e r
b e
o b l i g a t a r y
or
o p t i o n a l a n d
c a n
h a v e
one
of
three
modes
of
a p p l i c a t i o n :
s i n g l e ,
s i m u l t a n e o u s ,
or
r e p e t i t i v e .
The
PGTalso
c o n t a i n s
p r o v i s i o n
for
ac o n t r o l
program
which
will
d i r e c t
the order
in
w h i c h t h e
rules are
t r i e d ,
w i t h the
d e f a u l t
being
that
t h e
r u l e s are
tried
once
i n
t h e
order
g i v e n .
w i t h
the P G T
has
b e e n
d e v e l o p e d ,
n o r
H a s
aprocedure
for
a c q u i r i n g
an
a p p r o p r i a t e
i n p u t
set.
Our
p r o c e d u r e
has
b e e nt o
a d j u s t
t h e
g r a m m a r
w h i l e
t r y i n g
to
o b t a i n
the
samplz derivations
Note: e x p l a n a t i o n s
are
in
lower
case.' ~ 3 : N T E i O D U C Y I O N : l m S<'!lB N < ' f l B
'XI2
'N 'SB
IT
'R
$03
'SB
'D
' U H 1
'K
'SR
' S H'EH2
'N
'MB>
' M E > *echoes the
i n p u t
T R E E
R E A D B Y
P T R I N1
S
2 X3
N
# #
11/21
N
+
T
R
0131
+
D
U f l / 1 /
K
+
S HEH/2/ N
# 4o u t p u t
t h e
form
i n
tree
a n d
linear
form
rltrlr**rk TRANSFORBATIONS
I****
SCANC A L L E D
AT1
I
start scanning
on
top
llSmnode
ANTEST CALLEC FOR
1
" R
E D V OW
(AACC),SD=
2.
RES=
5 .TOP=
1:S
attempt
t oa p p l y
rule
redvow t oform
AN
TEST
R E T U R N S**
I * *
r u l e redvov h a s matched once
absenco
of this message i n d i c a t e s
that the
rule
d i d
not
a p p l y
CHBWGE, H l V E CSERCH FOR HERGEF
IN
10merge
f e a t u r e s
i n t o
node
10(making
it
aschwa)
ANTEST CALLED
F O B
211SCFIHDA (AACC),
SD=
3.
RES=7.
TOP= 1 :SANTEST CALLED
F O R
3"
SCH
W D B
(AACC),SD=
4.
RES=
9 .TOP=
1 :SANTEST CALLEC FOR
U ING
(AACC),
SD=
5,
RES=O 0
TOP.= 1:sANTEST CALLED
F O E
SV1SYZLAB
tt (AACC),SD=
6.
RES-
11.TOP=
1 :SANTEST RETURNS
**
I * *
CHANGE. CALL ELENOP