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
|
#ifndef FF_FSA_H
#define FF_FSA_H
//TODO: actually compile this; probably full of syntax errors.
#include <stdint.h> //C99
#include <string>
#include "ff.h"
#include "sparse_vector.h"
#include "value_array.h" // used to hold state
#include "tdict.h"
#include "hg.h"
typedef ValueArray<uint8_t> Bytes;
/*
features whose score is just some PFSA over target string. TODO: could decide to give access to source span of scanned words as well if someone devises a feature that can use it
state is some fixed width byte array. could actually be a void *, WordID sequence, whatever.
*/
// it's not necessary to inherit from this.
struct FsaFeatureFunctionBase {
std::string name,usage_short,usage_verbose;
int fid; // you can have more than 1 feature of course.
void InitFid() { // call this, though, if you have a single feature
fid=FD::Convert(name);
}
std::string usage(bool param,bool verbose) {
return FeatureFunction::usage_helper(name,usage_short,usage_verbose,param,verbose);
}
FsaFeatureFunctionBase(std::string const& name,std::string const& usage_verbose="[no documentation yet]",std::string const& usage_short="[no parameters]") : name(name),usage_short(usage_short),usage_verbose(usage_verbose) { }
int state_bytes; // don't forget to set this (it may depend on params of course)
};
// example: feature val = -1 * # of target words
struct TargetPenaltyFsa : public FsaFeatureFunctionBase {
TargetPenaltyFsa(std::string const& param) : FsaFeatureFunctionBase("TargetPenalty","","-1 per target word") { InitFid(); }
const float val_per_target_word=-1;
// state for backoff
// scan
void Scan(SentenceMetadata const& smeta,WordID x,void const* prev_state,FeatureVector *features) {
features->set_value(fid,val_per_target_word);
}
// heuristic estimate of phrase
void Heuristic(WordID const* begin, WordID const* end,FeatureVector *h_features)
// return m: all strings x with the same final m+1 letters must end in this state
/* markov chain of order m: P(xn|xn-1...x1)=P(xn|xn-1...xn-m) */
int MarkovOrder() const {
return 0;
}
};
//TODO: combine 2 FsaFeatures typelist style (can recurse for more)
// the type-erased interface
struct FsaFeatureFunction {
virtual int MarkovOrder() const = 0;
virtual ~FsaFeatureFunction();
};
// conforming to above interface, type erases FsaImpl
// you might be wondering: why do this? answer: it's cool, and it means that the bottom-up ff over ff_fsa wrapper doesn't go through multiple layers of dynamic dispatch
template <class Impl>
struct FsaFeatureFunctionDynamic : public FsaFeatureFunction {
Impl& d() { return static_cast<Impl&>(*this); }
Impl const& d() { return static_cast<Impl const&>(*this); }
int MarkovOrder() const { return d().MarkovOrder(); }
};
//TODO: combine 2 (or N) FsaFeatureFunction (type erased)
/* regular bottom up scorer from Fsa feature
uses guarantee about markov order=N to score ASAP
encoding of state: if less than N-1 (ctxlen) words
either:
struct FF : public FsaImpl,FeatureFunctionFromFsa<FF> (more efficient)
or:
struct FF : public FsaFeatureFunctionDynamic,FeatureFunctionFromFsa<FF> (code sharing, but double dynamic dispatch)
*/
template <class Impl>
struct FeatureFunctionFromFsa : public FeatureFunction {
Impl& d() { return static_cast<Impl&>(*this); }
Impl const& d() { return static_cast<Impl const&>(*this); }
int M; // markov order (ctx len)
FeatureFunctionFromFsa() { }
Init() {
name=d().name;
M=d().MarkovOrder
SetStateSize(sizeof(WordID)*2*M);
} // can't do this in constructor because we come before d() in order
virtual Features Features() const { return d().Features(); }
bool rule_feature() const {
return StateSize()==0; // Fsa features don't get info about span
}
};
#endif
|