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
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
|
/*
Copyright (C) 2000,2001 Franz Josef Och (RWTH Aachen - Lehrstuhl fuer Informatik VI)
This file is part of GIZA++ ( extension of GIZA ).
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307,
USA.
*/
#ifndef _d5tables_h_define
#define _d5tables_h_define
#include <math.h>
#include "D4Tables.h"
extern float d5modelsmooth_countoffset;
extern float d5modelsmooth_factor;
#define UNSEENPROB (1.0/vacancies_total)
class d5model
{
private:
typedef Vector < pair < COUNT,PROB > >Vpff;
map< m4_key,Vpff,compare1 > D1;
map< m4_key,Vpff,compareb1 > Db1;
public:
d4model&d4m;
WordClasses ewordclasses,fwordclasses;
template<class MAPPER>
void makeWordClasses(const MAPPER&m1,const MAPPER&m2,string efile,string ffile)
{
ifstream estrm(efile.c_str()),fstrm(ffile.c_str());
if( !estrm )
cerr << "ERROR: can not read classes from " << efile << endl;
else
ewordclasses.read(estrm,m1);
if( !fstrm )
cerr << "ERROR: can not read classes from " << ffile << endl;
else
fwordclasses.read(fstrm,m2);
}
d5model (d4model&_d4m)
:D1 (compare1(M5_Dependencies)), Db1 (compareb1(M5_Dependencies)),d4m(_d4m)
{}
COUNT &getCountRef_first (PositionIndex vacancies_j,
PositionIndex vacancies_jp, int F,
PositionIndex l, PositionIndex m,
PositionIndex vacancies_total)
{
massert(vacancies_j>0);
massert(vacancies_total>0);
//massert(vacancies_jp<=vacancies_total);
massert(vacancies_j <=vacancies_total);
massert(vacancies_total<=m);
m4_key key(M5_Dependencies,l,m,F,0,0,vacancies_jp,vacancies_total);
map<m4_key,Vpff,compare1 >::iterator p=D1.find(key);
if(p==D1.end())
p=D1.insert(make_pair(key,Vpff(vacancies_total+1,make_pair(0,UNSEENPROB)))).first; // !!! constrain length
massert(p!=D1.end());
return (p->second)[vacancies_j].first;
}
COUNT &getCountRef_bigger (PositionIndex vacancies_j,
PositionIndex vacancies_jp, int F,
PositionIndex l, PositionIndex m,
PositionIndex vacancies_total)
{
massert(vacancies_j>0);
massert(vacancies_total>0);
massert (vacancies_jp <= vacancies_j);
massert (vacancies_j-vacancies_jp <= vacancies_total);
m4_key key(M5_Dependencies,l,m,F,0,0,-1,vacancies_total);
map<m4_key,Vpff,compareb1 >::iterator p=Db1.find(key);
if(p==Db1.end())
p=Db1.insert(make_pair(key,Vpff(vacancies_total+1,make_pair(0,UNSEENPROB)))).first; // !!! constrain length
massert(p!=Db1.end());
return (p->second)[vacancies_j - vacancies_jp].first;
}
PROB getProb_first (PositionIndex vacancies_j, PositionIndex vacancies_jp,
int F, PositionIndex l, PositionIndex m,
PositionIndex vacancies_total) const
{
massert(vacancies_j>0);
massert(vacancies_total>0);
//massert(vacancies_jp<=vacancies_total);
massert(vacancies_j <=vacancies_total);
massert(vacancies_total<=m);
m4_key key(M5_Dependencies,l,m,F,0,0,vacancies_jp,vacancies_total);
map<m4_key,Vpff,compare1 >::const_iterator p=D1.find(key);
if( p==D1.end() )
return UNSEENPROB;
else
return max(PROB_SMOOTH,d5modelsmooth_factor/(vacancies_total)+(1-d5modelsmooth_factor)*(p->second)[vacancies_j].second);
}
PROB getProb_bigger (PositionIndex vacancies_j, PositionIndex vacancies_jp,
int F, PositionIndex l, PositionIndex m,
PositionIndex vacancies_total) const
{
massert(vacancies_j>0);
massert(vacancies_total>0);
massert (vacancies_jp <= vacancies_j);
massert (vacancies_j-vacancies_jp <= vacancies_total);
m4_key key(M5_Dependencies,l,m,F,0,0,-1,vacancies_total);
map<m4_key,Vpff,compareb1 >::const_iterator p=Db1.find(key);
if(p==Db1.end())
return UNSEENPROB;
else
return max(PROB_SMOOTH,d5modelsmooth_factor/(vacancies_total)+(1-d5modelsmooth_factor)*(p->second)[vacancies_j - vacancies_jp].second);
}
void normalizeTable ()
{
int nParams=0;
for(map<m4_key,Vpff,compare1 >::iterator i=D1.begin();i!=D1.end();++i)
{
Vpff&d1=i->second;
COUNT sum=0.0;
for(PositionIndex i=0;i<d1.size();i++)
sum+=d1[i].first+d5modelsmooth_countoffset;
for(PositionIndex i=0;i<d1.size();i++)
{
d1[i].second=sum?((d1[i].first+d5modelsmooth_countoffset)/sum):(1.0/d1.size());
nParams++;
}
}
for(map<m4_key,Vpff,compareb1 >::iterator i=Db1.begin();i!=Db1.end();++i)
{
Vpff&db1=i->second;
double sum=0.0;
for(PositionIndex i=0;i<db1.size();i++)
sum+=db1[i].first+d5modelsmooth_countoffset;
for(PositionIndex i=0;i<db1.size();i++)
{
db1[i].second=sum?((db1[i].first+d5modelsmooth_countoffset)/sum):(1.0/db1.size());
nParams++;
}
}
cout << "D5 table contains " << nParams << " parameters.\n";
}
friend ostream&operator<<(ostream&out,d5model&d5m)
{
out << "# Translation tables for Model 5 .\n";
out << "# Table for head of cept.\n";
for(map<m4_key,Vpff,compare1 >::const_iterator i=d5m.D1.begin();i!=d5m.D1.end();++i)
{
const Vpff&d1=i->second;
COUNT sum=0.0;
for(PositionIndex ii=0;ii<d1.size();ii++)sum+=d1[ii].first;
if ( sum )
{
for(unsigned ii=0;ii<d1.size();ii++)
{
print1_m5(out,i->first,d5m.ewordclasses,d5m.fwordclasses);
out << (int)(ii) << ' ' << d1[ii].second << ' ' << d1[ii].first << '\n';
}
out << endl;
}
}
out << "# Table for non-head of cept.\n";
for(map<m4_key,Vpff,compareb1 >::const_iterator i=d5m.Db1.begin();i!=d5m.Db1.end();++i)
{
const Vpff&db1=i->second;
double sum=0.0;
for(PositionIndex ii=0;ii<db1.size();++ii)sum+=db1[ii].first;
if( sum )
{
for(unsigned ii=0;ii<db1.size();ii++)
{
printb1_m5(out,i->first,d5m.fwordclasses);
out << (int)(ii) << ' ' << db1[ii].second << ' ' << db1[ii].first << '\n';
}
out << endl;
}
}
return out;
}
void readProbTable(const char*x)
{
ifstream f(x);
string l;
while(getline(f,l))
{
if(l.length()&&l[0]=='#')
continue;
istringstream is(l.c_str());
string E,F;
int v1,v2,ii;
double prob,count;
if(is>>E>>F>>v1>>v2>>ii>>prob>>count)
{
//cerr << "Read: " << E << " " << F << " " << v1 << " " << v2 << " " << prob<< endl;
if( count>0 )
if( E=="-1")
getCountRef_bigger(ii,0,fwordclasses(F),1000,1000,v2)+=count;
else
getCountRef_first(ii,v1,fwordclasses(F),1000,1000,v2)+=count;
}
}
normalizeTable();
ofstream of("M5FILE");
of << (*this);
}
void clear()
{
for(map<m4_key,Vpff,compare1 >::iterator i=D1.begin();i!=D1.end();++i)
{
Vpff&d1=i->second;
for(PositionIndex i=0;i<d1.size();i++)
d1[i].first=0.0;
}
for(map<m4_key,Vpff,compareb1 >::iterator i=Db1.begin();i!=Db1.end();++i)
{
Vpff&db1=i->second;
for(PositionIndex i=0;i<db1.size();i++)
db1[i].first=0.0;
}
}
};
#endif
|
