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
|
#ifndef _DTRAIN_KBESTGET_H_
#define _DTRAIN_KBESTGET_H_
#include <vector>
#include <string>
using namespace std;
#include "kbest.h" // cdec
#include "verbose.h"
#include "viterbi.h"
#include "ff_register.h"
#include "decoder.h"
#include "weights.h"
namespace dtrain
{
typedef double score_t; // float
struct ScoredHyp
{
vector<WordID> w;
SparseVector<double> f;
score_t model;
score_t score;
unsigned rank;
};
struct HypSampler : public DecoderObserver
{
virtual vector<ScoredHyp>* GetSamples()=0;
};
struct KBestGetter : public HypSampler
{
const unsigned k_;
const string filter_type_;
vector<ScoredHyp> s_;
KBestGetter(const unsigned k, const string filter_type) :
k_(k), filter_type_(filter_type) {}
virtual void
NotifyTranslationForest(const SentenceMetadata& smeta, Hypergraph* hg)
{
KBest(*hg);
}
vector<ScoredHyp>* GetSamples() { return &s_; }
void
KBest(const Hypergraph& forest)
{
if (filter_type_ == "unique") {
KBestUnique(forest);
} else if (filter_type_ == "no") {
KBestNoFilter(forest);
}
}
void
KBestUnique(const Hypergraph& forest)
{
s_.clear();
KBest::KBestDerivations<vector<WordID>, ESentenceTraversal,
KBest::FilterUnique, prob_t, EdgeProb> kbest(forest, k_);
for (unsigned i = 0; i < k_; ++i) {
const KBest::KBestDerivations<vector<WordID>, ESentenceTraversal, KBest::FilterUnique,
prob_t, EdgeProb>::Derivation* d =
kbest.LazyKthBest(forest.nodes_.size() - 1, i);
if (!d) break;
ScoredHyp h;
h.w = d->yield;
h.f = d->feature_values;
h.model = log(d->score);
h.rank = i;
s_.push_back(h);
}
}
void
KBestNoFilter(const Hypergraph& forest)
{
s_.clear();
KBest::KBestDerivations<vector<WordID>, ESentenceTraversal> kbest(forest, k_);
for (unsigned i = 0; i < k_; ++i) {
const KBest::KBestDerivations<vector<WordID>, ESentenceTraversal>::Derivation* d =
kbest.LazyKthBest(forest.nodes_.size() - 1, i);
if (!d) break;
ScoredHyp h;
h.w = d->yield;
h.f = d->feature_values;
h.model = log(d->score);
h.rank = i;
s_.push_back(h);
}
}
};
} // namespace
#endif
|
