blob: 593019c893cbc8fb6cbeb77c003e02a48772a450 (
plain)
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
|
#ifndef _SENTENCE_METADATA_H_
#define _SENTENCE_METADATA_H_
#include <cassert>
#include "lattice.h"
#include "scorer.h"
struct SentenceMetadata {
SentenceMetadata(int id, const Lattice& ref) :
sent_id_(id),
src_len_(-1),
has_reference_(ref.size() > 0),
trg_len_(ref.size()),
ref_(has_reference_ ? &ref : NULL) {}
// this should be called by the Translator object after
// it has parsed the source
void SetSourceLength(int sl) { src_len_ = sl; }
// this should be called if a separate model needs to
// specify how long the target sentence should be
void SetTargetLength(int tl) {
assert(!has_reference_);
trg_len_ = tl;
}
bool HasReference() const { return has_reference_; }
const Lattice& GetReference() const { return *ref_; }
int GetSourceLength() const { return src_len_; }
int GetTargetLength() const { return trg_len_; }
int GetSentenceID() const { return sent_id_; }
// this will be empty if the translator accepts non FS input!
const Lattice& GetSourceLattice() const { return src_lattice_; }
// access to document level scores for MIRA vector computation
void SetScore(Score *s){app_score=s;}
void SetDocScorer (const DocScorer *d){ds = d;}
void SetDocLen(double dl){doc_len = dl;}
const Score& GetScore() const { return *app_score; }
const DocScorer& GetDocScorer() const { return *ds; }
double GetDocLen() const {return doc_len;}
private:
const int sent_id_;
// the following should be set, if possible, by the Translator
int src_len_;
double doc_len;
const DocScorer* ds;
const Score* app_score;
public:
Lattice src_lattice_; // this will only be set if inputs are finite state!
private:
// you need to be very careful when depending on these values
// they will only be set during training / alignment contexts
const bool has_reference_;
int trg_len_;
const Lattice* const ref_;
};
#endif
|
