Below are samples of the phonological similarity matrices we created, using the feature specifications that we provide for Spanish, English and Basque.
For access to the full matrices, please contact us.
The matrix samples can be compared to the ones in Kondrak, 2002, p. 57, which contain a set of phonemes similar to these. The scores are not exactly the same, since we adapted features and their values to represent Spanish and English phonemes more closely.
The matrix' diagonals are also different to Kondrak's, since we modified the way the similarity function scores vowel matches. The original function was intended for cognate alignment. For audio alignment, we got slightly better alignment results with a modified version. For information about this, as well as the parameters we ran the similarity function with to get these scores, see the implementation page.
IPA
æ
i:
k
n
p
ɹ
s
aj
i:
9
35
-29
-26
-37
4
-16
-46
k
-49
-29
35
0
19
-10
10
-46
n
-46
-26
0
35
4
5
5
-46
p
-57
-37
19
4
35
-6
14
-46
ɹ
-16
4
-10
5
-6
35
15
-46
s
-36
-16
10
5
14
15
35
-46
aj
10
-46
-46
-46
-46
-46
-46
35
IPA
a
i
j
k
n
p
r
s
s̪
a
35
7
2
-40
-50
-56
-30
-50
-52
i
7
35
10
-24
-26
-32
-6
-26
-28
j
2
10
35
-19
-21
-27
-1
-21
-23
k
-40
-24
-19
35
5
19
-15
5
3
n
-50
-26
-21
5
35
9
-5
-5
-7
p
-56
-32
-27
19
9
35
-11
9
11
r
-30
-6
-1
-15
-5
-11
35
-5
-7
s
-50
-26
-21
5
-5
9
-5
35
33
s̪
-52
-28
-23
3
-7
11
-7
33
35