from_token2crf.c
6.41 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
/* From a TOKEN file with POS file to a CRF++ training file */
/* FRED 0309 */
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <strings.h>
/* ................................................................ */
#define TailleLigne 8000
#define True 1
#define False 0
void
ERREUR(char *ch1, char *ch2)
{
fprintf(stderr, "ERREUR : %s %s\n", ch1, ch2);
exit(0);
}
#define SI_MINUSCULE(a) (((a)>='a')&&((a)<='z'))
/* ................................................................ */
/*
* Format:
*
* file .tk
*
* <sentence id="s0001"> <text>merci � tous d' �couter France_Inter euh vous
* �tes en ce moment un million deux cent mille � �tre branch�s sur
* l' antenne de France_Inter , et_puis ce qui nous fait plaisir aussi , c'
* est que le petit matin marche tr�s_bien . </text> <tokens count="93">
* <token id="s0001_t0001" type="sgmltag"><Sync time="0"/></token> <token
* type="space" id="s0001_t0002"/> <token type="wtoken"
* id="s0001_t0003">merci</token> <token type="space" id="s0001_t0004"/>
* <token type="wtoken" id="s0001_t0005">�</token> <token type="space"
* id="s0001_t0006"/> <token type="wtoken" id="s0001_t0007">tous</token>
*
* file .pos
*
* <word id="s0001_w0001" token="s0001_t0003" pos="NMS"> merci </word> <word
* id="s0001_w0002" token="s0001_t0005" pos="PREPADE"> � </word> <word
* id="s0001_w0003" token="s0001_t0007" pos="AINDMP"> tous </word> <word
* id="s0001_w0004" token="s0001_t0009" pos="PREPADE"> d' </word> <word
* id="s0001_w0005" token="s0001_t0011" pos="VINF"> �couter </word> <word
* id="s0001_w0006" token="s0001_t0014" pos="XSOC"> France_Inter </word>
* <word id="s0001_w0007" token="s0001_t0017" pos="ADV"> euh </word> <word
* id="s0001_w0008" token="s0001_t0019" pos="PPER2P"> vous </word> <word
* id="s0001_w0009" token="s0001_t0021" pos="VE2P"> �tes </word> <word
* id="s0001_w0010" token="s0001_t0023" pos="PREP"> en </word> <word
* id="s0001_w0011" token="s0001_t0025" pos="DETMS"> ce </word>
*
*
* CRF++ training:
*
* Patricia XPREF B-pers Martin XFAMIL I-pers que COSUB O voici PREP O que COSUB
* O voilà PREP O
*
*/
void
analyze_pos(char *ch, char **chid, char **chpos, char **chword)
{
int i;
for (i = 0; (ch[i]) && (strncmp(ch + i, "token=", 6)); i++);
if (!ch[i])
ERREUR("bad POS string:", ch);
(*chid) = ch + i + 7;
for (i += 7; (ch[i]) && (ch[i] != '"'); i++);
if (!ch[i])
ERREUR("bad POS string:", ch);
ch[i] = '\0';
for (++i; (ch[i]) && (strncmp(ch + i, "pos=", 4)); i++);
if (!ch[i])
ERREUR("bad POS string:", ch);
(*chpos) = ch + i + 5;
for (i += 5; (ch[i]) && (ch[i] != '"'); i++);
if (!ch[i])
ERREUR("bad POS string:", ch);
ch[i] = '\0';
for (++i; (ch[i]) && (ch[i] != '>'); i++);
if (!ch[i])
ERREUR("bad POS string:", ch);
for (++i; (ch[i]) && (ch[i] == ' '); i++);
(*chword) = ch + i;
for (++i; (ch[i]) && (ch[i] != ' ') && (ch[i] != '<'); i++);
if (!ch[i])
ERREUR("bad POS string:", ch);
ch[i] = '\0';
}
/*
* analyse NE <token id="s0001_t0003" type="sgmltag"><Event desc="time.hour"
* type="entities" extent="begin"/></token>
*/
void
analyze_tag(char *ch, char **desc, char **type, char **extent)
{
int i;
*desc = *type = *extent = NULL;
for (i = 0; ch[i]; i++)
if (!strncmp(ch + i, "desc=", 5)) {
*desc = ch + i + 6;
for (i += 6; (ch[i]) && (ch[i] != '"'); i++);
if (!ch[i])
ERREUR("bad format:", ch);
ch[i] = '\0';
} else if (!strncmp(ch + i, "type=", 5)) {
*type = ch + i + 6;
for (i += 6; (ch[i]) && (ch[i] != '"'); i++);
if (!ch[i])
ERREUR("bad format:", ch);
ch[i] = '\0';
} else if (!strncmp(ch + i, "extent=", 7)) {
*extent = ch + i + 8;
for (i += 8; (ch[i]) && (ch[i] != '"'); i++);
if (!ch[i])
ERREUR("bad format:", ch);
ch[i] = '\0';
}
}
int
main(int argc, char **argv)
{
char ch[TailleLigne], ch2[TailleLigne], *chid, *chpos,
*chword, *chdesc, *chtype, *chextent, chNE[100];
int nb, insent, i, j, beginNE;
FILE *file_tk, *file_pos;
file_tk = file_pos = NULL;
if (argc > 1)
for (nb = 1; nb < argc; nb++)
if (!strcmp(argv[nb], "-tk")) {
if (nb + 1 == argc)
ERREUR("an option must follow option:", argv[nb]);
if (!(file_tk = fopen(argv[++nb], "rt")))
ERREUR("can't open:", argv[nb]);
} else if (!strcmp(argv[nb], "-pos")) {
if (nb + 1 == argc)
ERREUR("an option must follow option:", argv[nb]);
if (!(file_pos = fopen(argv[++nb], "rt")))
ERREUR("can't open:", argv[nb]);
} else if (!strcmp(argv[nb], "-h")) {
fprintf(stderr, "Syntax: %s [-h] -tk <file> -pos <file>\n", argv[0]);
exit(0);
} else
ERREUR("unknown option:", argv[nb]);
if ((!file_tk) || (!file_pos))
ERREUR("bad syntax, check '-h'", "");
while ((fgets(ch2, TailleLigne, file_pos)) && (!strstr(ch2, "<word")));
if (!feof(file_pos))
analyze_pos(ch2, &chid, &chpos, &chword);
else
ERREUR("empty POS file ??", "");
for (chNE[0] = '\0', beginNE = insent = False, nb = 0; fgets(ch, TailleLigne, file_tk); nb++)
if (strstr(ch, "<tokens"))
insent = True;
else if (strstr(ch, "</tokens")) {
insent = False;
printf("\n");
} else if (insent) {
if (strstr(ch, "<token "))
if (!strstr(ch, "sgmltag")) {
if (chid == NULL)
ERREUR("files TOKEN and POS are not synchronized !!", "");
for (i = 0; (ch[i]) && (strncmp(ch + i, "id=", 3)); i++);
if (!ch[i])
ERREUR("bad token string:", ch);
for (j = i + 4; (ch[j]) && (ch[j] != '"'); j++);
if (!ch[j])
ERREUR("bad token string:", ch);
ch[j] = '\0';
/*
* printf("[%s] ==
* [%s]\n",ch+i+4,chid) ;
*/
if (!strcmp(ch + i + 4, chid)) {
printf("%s\t%s\t", chword, chpos);
if (chNE[0])
printf("%c-%s\n", beginNE ? 'B' : 'I', chNE);
else
printf("O\n");
while ((fgets(ch2, TailleLigne, file_pos)) && (!strstr(ch2, "<word")));
if (!feof(file_pos))
analyze_pos(ch2, &chid, &chpos, &chword);
else
chid = chpos = chword = NULL;
}
beginNE = False;
} else {
analyze_tag(ch, &chdesc, &chtype, &chextent);
if (!strcmp(chtype, "entities")) {
if (!strcmp(chextent, "begin")) {
for (i = 0; (chdesc[i]) && (chdesc[i] != '.'); i++)
chNE[i] = SI_MINUSCULE(chdesc[i]) ? chdesc[i] - ('a' - 'A') : chdesc[i];
chNE[i] = '\0';
beginNE = True;
} else {
beginNE = False;
chNE[0] = '\0';
}
}
}
}
exit(0);
}