-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathkl.c
479 lines (441 loc) · 15 KB
/
kl.c
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
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
/* Weight-setting and scoring for Kuback-Leiber classification */
/* Copyright (C) 1997, 1998, 1999 Andrew McCallum
Written by: Andrew Kachites McCallum <[email protected]>
This file is part of the Bag-Of-Words Library, `libbow'.
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Library General Public License
as published by the Free Software Foundation, version 2.
This library 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
Library General Public License for more details.
You should have received a copy of the GNU Library General Public
License along with this library; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111, USA */
#include <bow/libbow.h>
#include <math.h>
/* For changing weight of unseen words.
I really should implement `deleted interpolation' */
/* M_EST_P summed over all words in the vocabulary must sum to 1.0! */
#if 1
/* This is the special case of the M-estimate that is `Laplace smoothing' */
#define M_EST_M (barrel->wi2dvf->num_words)
#define M_EST_P (1.0 / barrel->wi2dvf->num_words)
#define WORD_PRIOR_COUNT 1.0
#else
#define M_EST_M (cdoc->word_count \
? (((float)barrel->wi2dvf->num_words) / cdoc->word_count) \
: 1.0)
#define M_EST_P (1.0 / barrel->wi2dvf->num_words)
#endif
/* Function to assign `Naive Bayes'-style weights to each element of
each document vector. */
void
bow_kl_set_weights (bow_barrel *barrel)
{
int ci;
bow_cdoc *cdoc;
int wi; /* a "word index" into WI2DVF */
int max_wi; /* the highest "word index" in WI2DVF. */
bow_dv *dv; /* the "document vector" at index WI */
int dvi; /* an index into the DV */
int weight_setting_num_words = 0;
int total_num_words = 0;
/* We assume that we have already called BOW_BARREL_NEW_VPC() on
BARREL, so BARREL already has one-document-per-class. */
assert (!strcmp (barrel->method->name, "kl"));
max_wi = MIN (barrel->wi2dvf->size, bow_num_words());
/* Get the total number of terms in each class; store this in
CDOC->WORD_COUNT. */
/* Get the total number of unique terms in each class; store this in
CDOC->NORMALIZER. */
/* Calculate the total number of occurrences of each word; store this
in DV->IDF. */
for (ci = 0; ci < barrel->cdocs->length; ci++)
{
cdoc = bow_array_entry_at_index (barrel->cdocs, ci);
cdoc->word_count = 0;
cdoc->normalizer = 0;
}
for (wi = 0; wi < max_wi; wi++)
{
dv = bow_wi2dvf_dv (barrel->wi2dvf, wi);
if (dv == NULL)
continue;
dv->idf = 0;
for (dvi = 0; dvi < dv->length; dvi++)
{
cdoc = bow_array_entry_at_index (barrel->cdocs,
dv->entry[dvi].di);
cdoc->word_count += dv->entry[dvi].count;
total_num_words += dv->entry[dvi].count;
dv->idf += dv->entry[dvi].count;
cdoc->normalizer++;
}
}
for (wi = 0; wi < max_wi; wi++)
{
dv = bow_wi2dvf_dv (barrel->wi2dvf, wi);
if (dv == NULL)
continue;
for (dvi = 0; dvi < dv->length; dvi++)
{
float pr_w_c;
cdoc = bow_array_entry_at_index (barrel->cdocs,
dv->entry[dvi].di);
pr_w_c = (float)dv->entry[dvi].count / cdoc->word_count;
}
}
/* Set the weights in the BARREL's WI2DVF so that they are
equal to the log likelihood-ratio, Pr(w|C)/Pr(w|~C). */
for (wi = 0; wi < max_wi; wi++)
{
double pr_w = 0.0;
dv = bow_wi2dvf_dv (barrel->wi2dvf, wi);
/* If the model doesn't know about this word, skip it. */
if (dv == NULL)
continue;
pr_w = ((double)dv->idf) / total_num_words;
/* Now loop through all the elements, setting their weights */
for (dvi = 0; dvi < dv->length; dvi++)
{
double pr_w_c;
double pr_w_not_c;
double log_likelihood_ratio;
cdoc = bow_array_entry_at_index (barrel->cdocs,
dv->entry[dvi].di);
/* Here CDOC->WORD_COUNT is the total number of words in the class */
/* We use Laplace Estimation. */
pr_w_c = ((double)dv->entry[dvi].count
/ (cdoc->word_count + cdoc->normalizer));
pr_w_c = (((double)dv->entry[dvi].count + 1)
/ (cdoc->word_count + barrel->wi2dvf->num_words));
pr_w_not_c = ((dv->idf - dv->entry[dvi].count
+ barrel->cdocs->length - 1)
/
(total_num_words - cdoc->word_count
+ (barrel->wi2dvf->num_words
* (barrel->cdocs->length - 1))));
log_likelihood_ratio = log (pr_w_c / pr_w_not_c);
dv->entry[dvi].weight = log_likelihood_ratio;
dv->entry[dvi].weight = pr_w_c * log_likelihood_ratio;
}
weight_setting_num_words++;
/* Set the IDF. Kl doesn't use it; make it have no effect */
dv->idf = 1.0;
}
#if 0
fprintf (stderr, "wi2dvf num_words %d, weight-setting num_words %d\n",
barrel->wi2dvf->num_words, weight_setting_num_words);
#endif
}
int
bow_kl_score (bow_barrel *barrel, bow_wv *query_wv,
bow_score *bscores, int bscores_len,
int loo_class)
{
double *scores; /* will become prob(class), indexed over CI */
int ci; /* a "class index" (document index) */
int wvi; /* an index into the entries of QUERY_WV. */
int dvi; /* an index into a "document vector" */
float pr_w_c; /* P(w|C), prob a word is in a class */
int num_scores; /* number of entries placed in SCORES */
int query_word_count;
double score_increment = 0;
double pr_w_d;
#define KL_AGAINST_UNCOND 0
#if KL_AGAINST_UNCOND
int total_num_words = 0; /* number of word occurrences in all classes */
int total_num_w = 0; /* number of WI occurrences in all classes */
double pr_w; /* unconditional probability of WI. */
#endif
int count_w_c;
int count_c;
int num_smoothes = 0;
double entropy_d = 0;
/* Allocate space to store scores for *all* classes (documents) */
scores = alloca (barrel->cdocs->length * sizeof (double));
/* Calculate the total number of words in QUERY_WV. Should we start
at one to prevent getting a zero here? Also, would this be
statistically more correct, too? */
query_word_count = 0;
for (wvi = 0; wvi < query_wv->num_entries; wvi++)
{
/* Only count those words that are in the model's vocabulary. */
if (bow_wi2dvf_dv (barrel->wi2dvf, query_wv->entry[wvi].wi))
query_word_count += query_wv->entry[wvi].count;
}
if (query_word_count == 0)
query_word_count = 1;
#if KL_AGAINST_UNCOND
for (ci = 0; ci < barrel->cdocs->length; ci++)
{
bow_cdoc *cdoc = bow_array_entry_at_index (barrel->cdocs, ci);
assert (cdoc->type == model);
total_num_words += cdoc->word_count;
}
#endif
/* Initialize the SCORES to the class prior probabilities. */
if (bow_print_word_scores)
printf ("%s\n",
"(CLASS PRIOR PROBABILIES)");
for (ci = 0; ci < barrel->cdocs->length; ci++)
{
bow_cdoc *cdoc = bow_array_entry_at_index (barrel->cdocs, ci);
if (bow_uniform_class_priors)
{
scores[ci] = 0.0;
}
else
{
/* assert (cdoc->prior > 0 && cdoc->prior <= 1.0); */
assert (cdoc->prior >= 0 && cdoc->prior <= 1.0);
if (cdoc->word_count == 0)
/* There was no training data for this class. Give it a
special, impossible value. Positive scores are treated
as a special case that can never win. */
scores[ci] = 999.0;
else
scores[ci] = log (cdoc->prior) / query_word_count;
}
}
/* Loop over each word in the word vector QUERY_WV, putting its
contribution into SCORES. */
for (wvi = 0; wvi < query_wv->num_entries; wvi++)
{
int wi; /* the word index for the word at WVI */
bow_dv *dv; /* the "document vector" for the word WI */
/* Get information about this word. */
wi = query_wv->entry[wvi].wi;
dv = bow_wi2dvf_dv (barrel->wi2dvf, wi);
/* If the model doesn't know about this word, skip it. OOV */
if (!dv)
continue;
if (bow_print_word_scores)
printf ("%-30s (queryweight=%.8f)\n",
bow_int2word (wi),
query_wv->entry[wvi].weight * query_wv->normalizer);
pr_w_d = (float)query_wv->entry[wvi].count / query_word_count;
/* pr_w_d = 1.0 / query_wv->num_entries; xxx !!! */
entropy_d += pr_w_d * log (pr_w_d);
#if KL_AGAINST_UNCOND
/* Calculate the unconditional probability of word WI. */
for (dvi = 0; dvi < dv->length; dvi++)
total_num_w += dv->entry[dvi].count;
pr_w =
(((float)total_num_w + M_EST_M * M_EST_P * barrel->cdocs->length)
/ (total_num_words + M_EST_M * barrel->cdocs->length));
#endif
/* Loop over all classes, putting this word's (WI's)
contribution into SCORES. */
for (ci = 0, dvi = 0; ci < barrel->cdocs->length; ci++)
{
bow_cdoc *cdoc;
cdoc = bow_array_entry_at_index (barrel->cdocs, ci);
assert (cdoc->type == bow_doc_train);
if (cdoc->word_count == 0)
{
assert (scores[ci] > 0);
/* xxx Why isn't this true? assert (cdoc->prior == 0); */
continue;
}
/* Assign PR_W_C to P(w|C), either using a DV entry, or, if
there is no DV entry for this class, using M-estimate
smoothing */
if (dv)
while (dvi < dv->length && dv->entry[dvi].di < ci)
dvi++;
if (dv && dvi < dv->length && dv->entry[dvi].di == ci)
{
/* The count for this word in this class is non-zero. */
if (loo_class == ci)
{
/* xxx This is not exactly right, because
BARREL->WI2DVF->NUM_WORDS might have changed with the
removal of QUERY_WV's document. */
pr_w_c = ((float)
((M_EST_M * M_EST_P) + dv->entry[dvi].count
- query_wv->entry[wvi].count)
/ (M_EST_M + cdoc->word_count
- query_word_count));
if (pr_w_c <= 0)
bow_error ("A negative word probability was calculated.\n"
"This can happen if you are using "
"--test-files-loo and the test files are\n"
"not being lexed in the same way as they "
"were when the model was built.\n"
"Value is %f\n"
"Word is `%s'\n", pr_w_c, bow_int2word (wi));
count_w_c = dv->entry[dvi].count-query_wv->entry[wvi].count;
count_c = cdoc->word_count - query_wv->entry[wvi].count;
}
else
{
pr_w_c = ((float)
((M_EST_M * M_EST_P) + dv->entry[dvi].count)
/ (M_EST_M + cdoc->word_count));
if (pr_w_c <= 0)
bow_error ("A negative word probability was calculated. "
"This can happen if you are using\n"
"--test-files-loo and the test files are "
"not being lexed in the same way as they\n"
"were when the model was built.");
assert (pr_w_c > 0 && pr_w_c <= 1);
count_w_c = dv->entry[dvi].count;
count_c = cdoc->word_count;
if (bow_smoothing_method == bow_smoothing_wittenbell)
{
pr_w_c = ((float)dv->entry[dvi].count
/ (cdoc->word_count + cdoc->normalizer));
}
}
}
else
{
/* The count for this word in this class is zero. */
num_smoothes++;
if (loo_class == ci)
{
/* xxx This is not exactly right, because
BARREL->WI2DVF->NUM_WORDS might have changed with the
removal of QUERY_WV's document. */
pr_w_c = ((M_EST_M * M_EST_P)
/ (M_EST_M + cdoc->word_count
- query_word_count));
assert (pr_w_c > 0 && pr_w_c <= 1);
count_w_c = 0;
count_c = cdoc->word_count - query_wv->entry[wvi].count;
}
else
{
pr_w_c = ((M_EST_M * M_EST_P)
/ (M_EST_M + cdoc->word_count));
assert (pr_w_c > 0 && pr_w_c <= 1);
count_w_c = 0;
count_c = cdoc->word_count;
if (bow_smoothing_method == bow_smoothing_wittenbell)
{
if (cdoc->word_count)
/* There is training data for this class */
pr_w_c =
(cdoc->normalizer
/ ((cdoc->word_count + cdoc->normalizer)
*(barrel->wi2dvf->num_words-cdoc->normalizer)));
else
/* There is no training data for this class. */
pr_w_c = 1.0 / barrel->wi2dvf->num_words;
}
}
}
assert (pr_w_c > 0 && pr_w_c <= 1);
#if KL_AGAINST_UNCOND
/* score_increment = pr_w_d * log (pr_w_c / (pr_w)); */
score_increment = pr_w_d * log (pr_w_c / (pr_w * pr_w_d));
#else
/* score_increment = pr_w_d * log (pr_w_c); */
score_increment = pr_w_d * log (pr_w_c / pr_w_d);
#endif
assert (score_increment == score_increment);
scores[ci] += score_increment;
assert (scores[ci] == scores[ci]);
if (bow_print_word_scores)
printf (" %5d/%-6d %8.2e %7.4f %-25s %8.5f\n",
count_w_c,
count_c,
pr_w_c,
score_increment,
(strrchr (cdoc->filename, '/') ? : cdoc->filename),
scores[ci]);
}
}
/* Now SCORES[] contains a KL divergence for each class. */
#if 1
/* Normalize the SCORES so they all sum to minus one. */
{
double scores_sum = 0;
for (ci = 0; ci < barrel->cdocs->length; ci++)
{
if (scores[ci] <= 0)
scores_sum += scores[ci];
}
if (scores_sum)
{
for (ci = 0; ci < barrel->cdocs->length; ci++)
{
if (scores[ci] > 0)
scores[ci] = -FLT_MAX;
else
scores[ci] /= -scores_sum;
assert (scores[ci] == scores[ci]);
/* assert (scores[ci] > 0); */
}
}
else
{
for (ci = 0; ci < barrel->cdocs->length; ci++)
scores[ci] = -1.0 / barrel->cdocs->length;
}
}
#endif
/* Return the SCORES by putting them (and the `class indices') into
SCORES in sorted order. */
{
num_scores = 0;
for (ci = 0; ci < barrel->cdocs->length; ci++)
{
if (num_scores < bscores_len
|| bscores[num_scores-1].weight < scores[ci])
{
/* We are going to put this score and CI into SCORES
because either: (1) there is empty space in SCORES, or
(2) SCORES[CI] is larger than the smallest score there
currently. */
int dsi; /* an index into SCORES */
if (num_scores < bscores_len)
num_scores++;
dsi = num_scores - 1;
/* Shift down all the entries that are smaller than SCORES[CI] */
for (; dsi > 0 && bscores[dsi-1].weight < scores[ci]; dsi--)
bscores[dsi] = bscores[dsi-1];
/* Insert the new score */
bscores[dsi].weight = scores[ci];
bscores[dsi].di = ci;
}
}
}
#if 0
printf ("kl %8.6f %8.6f %d %d %8.6f %8.6f ",
-bscores[0].weight * scores_sum, /* unnormalized high score */
bscores[0].weight, /* normalized high score */
query_word_count, /* document length */
query_wv->num_entries, /* num unique words in query */
(float)num_smoothes/query_word_count,
entropy_d);
#endif
return num_scores;
}
rainbow_method bow_method_kl =
{
"kl",
bow_kl_set_weights,
0, /* no weight scaling function */
NULL, /* bow_barrel_normalize_weights_by_summing, */
bow_barrel_new_vpc_merge_then_weight,
bow_barrel_set_vpc_priors_by_counting,
bow_kl_score,
bow_wv_set_weights_to_count,
NULL, /* no need for extra weight normalization */
bow_barrel_free,
0
};
void _register_method_kl () __attribute__ ((constructor));
void _register_method_kl ()
{
static int done = 0;
if (done)
return;
bow_method_register_with_name ((bow_method*)&bow_method_kl, "kl",
sizeof (rainbow_method),
NULL);
done = 1;
}