|
SPRAAK
|
The distance matrix can, if desired, be specified using features which allows for a decomposition of the problem.
A simple distance matrix has the following format
[distance] / / it1 it2 it3 ... / C00 C01 C02 C03 ... it1 C10 C11 C12 C13 ... it2 C20 C21 C22 C23 ... it3 C30 C31 C32 C33 ... ... ... ... ... ... ...
The elements in the example above have the following function:
For a feature based matrix, one first has to list the features and the mapping between items and features:
[features] % some comment feat1 feat2 ... [feature_map] it1 featX it2 featY ... [distance] ...
The elements in the example above have the following function:
Multiple (sub) distance matrices may be defined. The cost assigned to an operation (sub/ins/del) is the sum of the costs defined in the various (sub) distance matrices. This allows for a simple feature based decomposition of the costs.
A extensive example is given below; all costs were assigned add-hoc:
% phoneme properties
% = can be considered to consist out of the given two parts
% X specials: silence / speaker noises
% C consonant
% p plosief (occlusive)
% f fricatief
% n nasaal
% l lateraal+trill
% g glide (approximant)
% bl bilabiaal
% ld labiodentaal
% al alveolair
% rf retroflex
% pt palataal
% vl velair
% gt glottaal
% v0 unvoiced
% v1 voiced
% V vowel
% o1 close
% o2 close-mid
% o3 open-mid
% o4 open
% p1 front
% p2 front-central
% p3 central
% p4 back
% s0 long
% s1 short
% ! stressed
%
% Phoneme table
% # X (f-v0)
% </s> = #
% p C-p-bl-v0
% b C-p-bl-v1
% t C-p-al-v0
% d C-p-al-v1
% T C-p-rf-v0
% D C-p-rf-v1
% k C-p-vl-v0
% g C-p-vl-v1
% f C-f-ld-v0
% v C-f-ld-v1
% s C-f-al-v0
% z C-f-al-v1
% S C-f-pt-v0 % S,Z were made palatal instead of postalveolar (closer to the /j/ + postalveolar was not in use)
% Z C-f-pt-v1
% h C-f-gt-v0
% m C-n-bl-v1
% n C-n-al-v1
% N C-n-vl-v1
% l C-l-al-v1-t0 % t0/t1 == trill
% r C-g-al-v1-t1 % classification based on the Chinese pronunciation of the /r/ :-)
% j C-g-pt-v1
% w C-g-bl-v1
% t+S = t S
% d+Z = d Z
% Note: last column indicates the linking phoneme the vowel induces (j/w/?)
% i V-o1-p1-s0-pt % worry-j-above
% E V-o3-p1-s1-pt %
% { V-o4-p1-s1-pt %
% I V-o1-p2-s1-pt %
% I! V-o1-p2-s1-pt! %
% @ V-o2-p3-s1-gt % extra-?-above
% @! V-o2-p3-s1-gt! %
% u V-o1-p4-s0-bl % flew-w-above
% U V-o1-p4-s1-bl %
% @+U V-o2-p4-s0-bl % yellow-w-above
% O V-o3-p4-s1-bl % outlaw-w-above
% A V-o4-p4-s1-gt % panama-?-above
% } = @ r
% }! = @! r
% e+I = E j
% a+U = A w
% a+I = a j
% O+I = O j
[features]
% major distinction between all 'phonemes'
#
V
Cp
CP
Cf
Cn
Cl
Cg
[feature_map]
# #
</s> #
p Cp
b Cp
t Cp
d Cp
T CP
D CP
k Cp
g Cp
f Cf
v Cf
s Cf
z Cf
S Cf
Z Cf
h Cf
m Cn
n Cn
N Cn
l Cl
r Cl
j Cg
w Cg
t+S CP
d+Z CP
i V
E V
{ V
I V
I! V
@ V
@! V
u V
U V
@+U V
O V
A V
} Cl
}! Cl
e+I V
a+U V
a+I V
O+I V
[distance_matrix]
/ # V Cp CP Cf Cn Cl Cg /
# 0.0 1.0 0.5 0.5 0.5 1.0 1.0 1.0 0.5
V 1.0 0.0 1.0 1.0 1.0 0.5 0.7 0.5 1.0
Cp 0.5 1.0 0.0 0.1 0.3 1.0 1.0 1.0 0.5
CP 0.5 1.0 0.1 0.0 0.2 1.0 1.0 1.0 0.5
Cf 0.5 1.0 0.3 0.2 0.0 1.0 1.0 1.0 0.5
Cn 1.0 0.5 1.0 1.0 1.0 0.0 0.7 0.5 1.0
Cl 1.0 0.7 1.0 1.0 1.0 0.7 0.0 0.7 1.0
Cg 1.0 0.5 1.0 1.0 1.0 0.5 0.7 0.0 0.7
/ 0.5 1.0 0.5 0.5 0.5 1.0 1.0 0.7 0.0
[features]
% further distinction between the consonants w.r.t. manner
bl
ld
al
rf
pt
vl
gt
[feature_map]
p bl
b bl
t al
d al
T rf
D rf
k vl
g vl
f ld
v ld
s al
z al
S pt
Z pt
h gt
m bl
n al
N vl
l al
r al
j pt
w bl
t+S al
d+Z al
} al
}! al
[distance_matrix]
/ bl ld al rf pt vl gt
bl 0.0 0.2 0.4 0.4 0.4 0.4 0.4
ld 0.2 0.0 0.2 0.4 0.4 0.4 0.4
al 0.4 0.2 0.0 0.2 0.2 0.4 0.4
rf 0.4 0.4 0.2 0.0 0.2 0.2 0.4
pt 0.4 0.4 0.2 0.2 0.0 0.2 0.4
vl 0.4 0.4 0.4 0.2 0.2 0.0 0.2
gt 0.4 0.4 0.4 0.4 0.4 0.2 0.0
[features]
% further distinction between the plosives and fricatives w.r.t. voiced/unvoiced
v0
v1
[feature_map]
p v0
b v1
t v0
d v1
T v0
D v1
k v0
g v1
f v0
v v1
s v0
z v1
S v0
Z v1
h v0
t+S v0
d+Z v1
} v1
}! v1
[distance_matrix]
/ v0 v1
v0 0.0 0.1
v1 0.1 0.0
[features]
% further distinction between trill and others
t0
t1
tx
[feature_map]
* tx
l t0
r t1
} t1
}! t1
[distance_matrix]
/ t0 t1 tx
t0 0.0 0.3 0.0
t1 0.3 0.0 0.1
tx 0.0 0.1 0.0
[features]
% further distinction between the vowels w.r.t. place
o1
o2
o3
o4
[feature_map]
i o1
E o3
{ o4
I o1
I! o1
@ o2
@! o2
u o1
U o1
@+U o2
O o3
A o4
} o2
}! o2
e+I o3
a+U o4
a+I o4
O+I o3
[distance_matrix]
/ o1 o2 o3 o4
o1 0.0 0.1 0.2 0.2
o2 0.1 0.0 0.1 0.2
o3 0.2 0.1 0.0 0.1
o4 0.2 0.2 0.1 0.0
[features]
% further distinction between the vowels w.r.t. openness
p1
p2
p3
p4
[feature_map]
i p1
E p1
{ p1
I p2
I! p2
@ p3
@! p3
u p4
U p4
@+U p4
O p4
A p4
} p3
}! p3
e+I p1
a+U p4
a+I p1
O+I p4
[distance_matrix]
/ p1 p2 p3 p4
p1 0.0 0.1 0.2 0.2
p2 0.1 0.0 0.1 0.2
p3 0.2 0.1 0.0 0.1
p4 0.2 0.2 0.1 0.0
[features]
% further distinction between the vowels w.r.t. length
s0
s1
[feature_map]
i s0
E s1
{ s1
I s1
I! s1
@ s1
@! s1
u s0
U s1
@+U s0
O s1
A s1
} s1
}! s1
e+I s0
a+U s0
a+I s0
O+I s0
[distance_matrix]
/ s0 s1
s0 0.0 0.1
s1 0.1 0.0
[features]
% further distinction between the vowels w.r.t. stress
s0
s1
[feature_map]
i s1
E s1
{ s1
I s0
I! s1
@ s0
@! s1
u s1
U s1
@+U s1
O s1
A s1
} s0
}! s1
e+I s1
a+U s1
a+I s1
O+I s1
[distance_matrix]
/ s0 s1
s0 0.0 0.1
s1 0.1 0.0
[features]
% further distinction between the vowels w.r.t. manner
pt
gt
bl
[feature_map]
i pt
E pt
{ pt
I pt
I! pt
@ gt
@! gt
u bl
U bl
@+U bl
O bl
A gt
} gt
}! gt
e+I pt
a+U bl
a+I pt
O+I pt
[distance_matrix]
/ pt gt bl
pt 0.0 0.2 0.2
gt 0.2 0.0 0.2
bl 0.2 0.2 0.0
[features]
% difference between diphtongs/monophtongs/sylabics
1
^
,
[feature_map]
i 1
E 1
{ 1
I 1
I! 1
@ 1
@! 1
u 1
U 1
@+U 1
O 1
A 1
} ,
}! ,
e+I ^
a+U ^
a+I ^
O+I ^
[distance_matrix]
/ 1 ^ ,
1 0.0 0.1 0.1
^ 0.1 0.0 0.3
, 0.1 0.3 0.0
[features]
% correction for the diphtongs/sylabics
2j
2w
2r
n
j
w
lr
@
[feature_map]
S 2j
Z 2j
t+S 2j
d+Z 2j
j j
w w
l lr
r lr
i 2j
u 2w
U 2w
@+U 2w
e+I 2j
a+U 2w
a+I 2j
O+I 2j
} 2r
}! 2r
@ @
@! @
I @
I! @
E @
{ @
[distance_matrix]
/ @ j w lr 2j 2w 2r
@ 0.0 0.0 0.0 0.0 0.0 0.0 -.6
j 0.0 0.0 0.0 0.0 -.1 0.0 0.0
w 0.0 0.0 0.0 0.0 0.0 -.1 0.0
lr 0.0 0.0 0.0 0.0 0.0 0.0 0.1
2j 0.0 -.1 0.0 0.0 0.0 0.0 0.0
2w 0.0 0.0 -.1 0.0 0.0 0.0 0.0
2r -.6 0.0 0.0 0.1 0.0 0.0 0.0
1.8.6