| 1 |
#include <iostream> |
| 2 |
#include <string> |
| 3 |
#include <vector> |
| 4 |
#include <fstream> |
| 5 |
#include <strstream> |
| 6 |
#include <algorithm> |
| 7 |
using namespace std; |
| 8 |
|
| 9 |
#include "setutils.h" |
| 10 |
|
| 11 |
#define DEBUG_SUBSETS 0 |
| 12 |
#define DEBUG_SMALLMERGE 0 |
| 13 |
|
| 14 |
|
| 15 |
// SMALLMERGE_RATIO: the golden ratio used for smallmerge algorithm. |
| 16 |
const double SMALLMERGE_RATIO = .5; |
| 17 |
|
| 18 |
// PITHY_FILTER: remove all meanings with less than this many synonyms listed. |
| 19 |
const unsigned int PITHY_FILTER = 10; |
| 20 |
|
| 21 |
|
| 22 |
template<class T> |
| 23 |
void printvec(const vector<T>& v, const char* label) |
| 24 |
{ |
| 25 |
cout << label << " = { "; |
| 26 |
copy(v.begin(), v.end(), ostream_iterator<T>(cout, ", ")); |
| 27 |
cout << "}\n"; |
| 28 |
} |
| 29 |
|
| 30 |
|
| 31 |
void cleanup(vector< vector<string> >& classes, vector< vector<string> >& output) |
| 32 |
{ |
| 33 |
output.reserve(50); |
| 34 |
|
| 35 |
// First, destroy any classes that only have two words. There |
| 36 |
// tend to be extremely obscure. |
| 37 |
for(int i = 0;i < classes.size();++i) |
| 38 |
{ |
| 39 |
if (classes[i].size() == 2) |
| 40 |
classes[i].clear(); |
| 41 |
|
| 42 |
sort(classes[i].begin(), classes[i].end()); |
| 43 |
} |
| 44 |
|
| 45 |
for(int i = 0;i < classes.size();++i) |
| 46 |
{ |
| 47 |
if (classes[i].empty()) |
| 48 |
continue; |
| 49 |
|
| 50 |
for(int j = i+1;j < classes.size();++j) |
| 51 |
{ |
| 52 |
if (classes[j].empty()) |
| 53 |
continue; |
| 54 |
|
| 55 |
vector<string> left, right, common; |
| 56 |
set_comparison(left, right, common, classes[i], classes[j]); |
| 57 |
|
| 58 |
|
| 59 |
// Now we will delete proper subsets. There shouldn't |
| 60 |
// be many of these. |
| 61 |
if (left.empty()) |
| 62 |
{ |
| 63 |
#if DEBUG_SUBSETS |
| 64 |
cout << i << " is a subset of " << j << endl; |
| 65 |
printvec(classes[i], " lhs"); |
| 66 |
printvec(classes[j], " rhs"); |
| 67 |
#endif // DEBUG_SUBSETS |
| 68 |
classes[i].clear(); |
| 69 |
break; |
| 70 |
} |
| 71 |
else if (right.empty()) |
| 72 |
{ |
| 73 |
#if DEBUG_SUBSETS |
| 74 |
cout << j << " is a subset of " << i << endl; |
| 75 |
printvec(classes[i], " lhs"); |
| 76 |
printvec(classes[j], " rhs"); |
| 77 |
#endif |
| 78 |
classes[j].clear(); |
| 79 |
} |
| 80 |
|
| 81 |
|
| 82 |
// Now we will attempt to merge small sets into large sets. |
| 83 |
// There should (hopefully) be many of these merges. |
| 84 |
// |
| 85 |
// We will make the convention that merges will be placed |
| 86 |
// into classes[i]. In other words, we want classes[i] to |
| 87 |
// become our large set, and clear out classes[j]. |
| 88 |
double lratio = double(common.size()) / double(common.size() + left.size()); |
| 89 |
if (lratio >= SMALLMERGE_RATIO) |
| 90 |
{ |
| 91 |
#if DEBUG_SMALLMERGE |
| 92 |
cout << "smallmerge: " << i << " into " << j << " (ratio=" << lratio << ")" << endl; |
| 93 |
printvec(classes[i], " lhs"); |
| 94 |
printvec(classes[j], " rhs"); |
| 95 |
#endif |
| 96 |
|
| 97 |
for(int k = 0;k < classes[j].size();++k) |
| 98 |
{ |
| 99 |
const string& word = (classes[j])[k]; |
| 100 |
if (!binary_search(classes[i].begin(), classes[i].end(), word)) |
| 101 |
{ |
| 102 |
classes[i].push_back(word); |
| 103 |
sort(classes[i].begin(), classes[i].end()); |
| 104 |
} |
| 105 |
} |
| 106 |
|
| 107 |
classes[j].clear(); |
| 108 |
|
| 109 |
--i; |
| 110 |
break; |
| 111 |
|
| 112 |
#if DEBUG_SMALLMERGE |
| 113 |
cout << "Post merge setup: " << endl; |
| 114 |
printvec(classes[i], " lhs"); |
| 115 |
printvec(classes[j], " rhs"); |
| 116 |
#endif |
| 117 |
} |
| 118 |
|
| 119 |
double rratio = double(common.size()) / double(common.size() + right.size()); |
| 120 |
if (rratio >= SMALLMERGE_RATIO) |
| 121 |
{ |
| 122 |
#if DEBUG_SMALLMERGE |
| 123 |
cout << "smallmerge: " << j << " into " << i << " (ratio=" << rratio << ")" << endl; |
| 124 |
printvec(classes[i], " lhs"); |
| 125 |
printvec(classes[j], " rhs"); |
| 126 |
#endif |
| 127 |
|
| 128 |
for(int k = 0;k < classes[j].size();++k) |
| 129 |
{ |
| 130 |
const string& word = (classes[j])[k]; |
| 131 |
if (!binary_search(classes[i].begin(), classes[i].end(), word)) |
| 132 |
{ |
| 133 |
classes[i].push_back(word); |
| 134 |
sort(classes[i].begin(), classes[i].end()); |
| 135 |
} |
| 136 |
} |
| 137 |
|
| 138 |
classes[j].clear(); |
| 139 |
|
| 140 |
--i; |
| 141 |
break; |
| 142 |
|
| 143 |
#if DEBUG_SMALLMERGE |
| 144 |
cout << "Post merge setup: " << endl; |
| 145 |
printvec(classes[i], " lhs"); |
| 146 |
printvec(classes[j], " rhs"); |
| 147 |
#endif |
| 148 |
} |
| 149 |
} |
| 150 |
} |
| 151 |
|
| 152 |
|
| 153 |
for(int i = 0;i < classes.size();++i) |
| 154 |
{ |
| 155 |
if (classes[i].size() > PITHY_FILTER) |
| 156 |
{ |
| 157 |
// copy(classes[i].begin(), classes[i].end(), |
| 158 |
// ostream_iterator<string>(cout, " ")); |
| 159 |
// cout << endl; |
| 160 |
output.push_back(classes[i]); |
| 161 |
} |
| 162 |
} |
| 163 |
} |
| 164 |
|