1
|
#!/usr/bin/python
|
2
|
# -*- coding: utf-8 -*-
|
3
|
|
4
|
import argparse
|
5
|
from asflight import AsfLight
|
6
|
|
7
|
# --------------------
|
8
|
|
9
|
|
10
|
def main():
|
11
|
|
12
|
# all consonants
|
13
|
phn_cons = [ 'b','c','C','d','D','w','W','f','g','h','x','G','j','k','l','L','m','M','H','n','J','N','p','r','P','R','Q','s','S','t','T','v','z','Z','!' ]
|
14
|
|
15
|
# consonants without r, l, P, L
|
16
|
phn_cons_m_left = [ 'b','c','C','d','D','w','W','f','g','h','x','G','j','k','m','M','H','n','J','N','p','R','Q','s','S','t','T','v','z','Z','!' ]
|
17
|
|
18
|
# consonants without r, l, P, L, J
|
19
|
phn_cons_m_right = [ 'b','c','C','d','D','w','W','f','g','h','x','G','j','k','m','M','H','n','N','p','R','Q','s','S','t','T','v','z','Z','!' ]
|
20
|
|
21
|
parser = argparse.ArgumentParser( description = 'Convert ASF file to MLF.' )
|
22
|
|
23
|
parser.add_argument( type=str, dest="asf_in", help="input ASF file", metavar="ASF_IN" )
|
24
|
|
25
|
parser.add_argument( type=str, dest="asf_out", help="input ASF file", metavar="ASF_OUT" )
|
26
|
|
27
|
parser.add_argument( "-c", "--code-page", type=str, dest="code_page",
|
28
|
help="encoding of all files, default value: %(default)s", default="utf-8" )
|
29
|
|
30
|
# parse arguments
|
31
|
args = parser.parse_args()
|
32
|
|
33
|
asf = AsfLight( args.asf_in )
|
34
|
|
35
|
for utt_name, asf_labels in asf.utts.iteritems():
|
36
|
|
37
|
n_labels = len( asf_labels )
|
38
|
|
39
|
for lab_idx in range( len( asf_labels ) ):
|
40
|
|
41
|
curr_lab = asf_labels[ lab_idx ]
|
42
|
prev_lab = asf_labels[ lab_idx-1 ] if ( lab_idx > 0 ) else None
|
43
|
next_lab = asf_labels[ lab_idx+1 ] if ( lab_idx+1 < n_labels ) else None
|
44
|
|
45
|
word_beg = ( curr_lab['word'] != '.' )
|
46
|
word_end = ( next_lab is None ) or ( next_lab['word'] != '.' )
|
47
|
|
48
|
if curr_lab['phone'] == 'n':
|
49
|
if ( not word_end ) and ( next_lab['phone'] in [ 'g', 'k' ] ):
|
50
|
curr_lab['phone'] = 'N'
|
51
|
|
52
|
elif curr_lab['phone'] == 'm':
|
53
|
if ( not word_beg ) and ( prev_lab['phone'] in phn_cons_m_left ) and ( word_end or ( next_lab['phone'] in phn_cons_m_right ) ):
|
54
|
curr_lab['phone'] = 'H'
|
55
|
|
56
|
elif ( not word_end ) and ( next_lab['phone'] in [ 'f', 'v' ] ):
|
57
|
curr_lab['phone'] = 'M'
|
58
|
|
59
|
elif curr_lab['phone'] == 'x':
|
60
|
if next_lab and ( next_lab['phone'] in [ 'b','d','D','w','W','g','h','G','z','Z' ] ): # voiced paired consonants without 'v'
|
61
|
curr_lab['phone'] = 'G'
|
62
|
|
63
|
elif curr_lab['phone'] == 'l':
|
64
|
if ( not word_beg ) and ( prev_lab['phone'] in phn_cons ) and ( word_end or ( next_lab['phone'] in phn_cons ) ):
|
65
|
curr_lab['phone'] = 'L'
|
66
|
|
67
|
elif curr_lab['phone'] == 'r':
|
68
|
if ( not word_beg ) and ( prev_lab['phone'] in phn_cons ) and ( word_end or ( next_lab['phone'] in phn_cons ) ):
|
69
|
curr_lab['phone'] = 'P'
|
70
|
|
71
|
asf.write( args.asf_out )
|
72
|
|
73
|
|
74
|
#
|
75
|
# ----------------------------------
|
76
|
#
|
77
|
|
78
|
# run the main program
|
79
|
if ( __name__ == "__main__" ):
|
80
|
main()
|