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
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
content / renderer / accessibility / annotations / ax_image_stopwords.cc [blame]
// Copyright 2020 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include "content/renderer/accessibility/annotations/ax_image_stopwords.h"
#include <string>
#include <string_view>
#include <vector>
#include "base/containers/contains.h"
#include "base/containers/flat_set.h"
#include "base/i18n/case_conversion.h"
#include "base/i18n/char_iterator.h"
#include "base/no_destructor.h"
#include "base/strings/string_split.h"
#include "base/strings/utf_string_conversions.h"
#include "third_party/icu/source/common/unicode/uchar.h"
namespace content {
namespace {
// List of image stopwords for all languages. See ax_image_stopwords.h
// for information about how image stopwords are defined and how they're
// used.
//
// The stopwords are encoded here as a single long string delimited by
// newlines. This is much more efficient than an array of strings, which
// in practice takes ~6x more storage in the resulting binary.
//
// Current size as of June 2020:
// 369 unique words
// 2542 bytes uncompressed
// 1127 bytes gzipped
const char kImageStopwordsUtf8[] = {
//
// Language-independent stopwords.
//
"com\n"
"edu\n"
"http\n"
"https\n"
"www\n"
//
// English.
//
// General English stopwords.
"and\n"
"are\n"
"for\n"
"from\n"
"how\n"
"online\n"
"our\n"
"the\n"
"this\n"
"with\n"
"you\n"
"your\n"
// English image-specific stopwords.
"art\n"
"avatar\n"
"background\n"
"backgrounds\n"
"black\n"
"download\n"
"drawing\n"
"drawings\n"
"free\n"
"gif\n"
"icon\n"
"icons\n"
"illustration\n"
"illustrations\n"
"image\n"
"images\n"
"jpeg\n"
"jpg\n"
"logo\n"
"logos\n"
"meme\n"
"memes\n"
"photo\n"
"photos\n"
"picture\n"
"pictures\n"
"png\n"
"stock\n"
"transparent\n"
"vector\n"
"vectors\n"
"video\n"
"videos\n"
"wallpaper\n"
"wallpapers\n"
"white\n"
// Alt text from may images starts with "Image may contain".
"may\n"
"contain\n"
// Many images on the web have the alt text "Highlights info row."
"highlights\n"
"info\n"
"row\n"
// Google Photos images are labeled as "portrait" or "landscape".
"portrait\n"
"landscape\n"
// Reddit says "post image".
"post\n"
// Months and month abbreviations. Often used as part of the date/time
// when a photograph was taken.
"january\n"
"jan\n"
"february\n"
"feb\n"
"march\n"
"mar\n"
"april\n"
"apr\n"
"may\n"
"june\n"
"jun\n"
"july\n"
"jul\n"
"august\n"
"aug\n"
"september\n"
"sep\n"
"october\n"
"oct"
"november"
"nov\n"
"december\n"
"dec\n"
// Days of the week.
"monday\n"
"mon\n"
"tuesday\n"
"tue\n"
"wednesday\n"
"wed\n"
"thursday\n"
"thu\n"
"friday\n"
"fri\n"
"saturday\n"
"sat\n"
"sunday\n"
"sun\n"
//
// French
//
// General French stopwords.
"les\n" // the
"pour\n" // for
"des\n" // of the
"sur\n" // on
"avec\n" // with
"une\n" // one
"dans\n" // in
"est\n" // is
"plus\n" // more
"par\n" // by
"vous\n" // you
"pas\n" // not
"qui\n" // who
"aux\n" // to the
"son\n" // his/her/its
"nous\n" // we
"voir\n" // see
// French Image stopwords.
"noir\n" // black
"blanc\n" // white
"dessin\n" // drawing
"font\n" // background
"peinture\n" // painting
// Months.
"janvier\n"
"janv\n"
"février\n"
"févr\n"
"mars\n"
"avril\n"
"mai\n"
"juin\n"
"juillet\n"
"juil\n"
"août\n"
"septembre\n"
"sept\n"
"octobre\n"
"oct\n"
"novembre\n"
"nov\n"
"décembre\n"
"déc\n"
// Days of the week.
"lundi\n"
"lun\n"
"mardi\n"
"mar\n"
"mercredi\n"
"mer\n"
"jeudi\n"
"jeu\n"
"vendredi\n"
"ven\n"
"samedi\n"
"sam\n"
"dimanche\n"
"dim\n"
//
// Italian
//
"con\n" // with
"per\n" // through, by
"non\n" // not
"come\n" // as
"più\n" // more
"dal\n" // da + il
"dallo\n" // da + lo
"dai\n" // da + i
"dagli\n" // da + gli
"dall\n" // da + l'
"dagl\n" // da + gll'
"dalla\n" // da + la
"dalle\n" // da + le
"del\n" // di + il
"dello\n" // di + lo
"dei\n" // di + i
"degli\n" // di + gli
"dell\n" // di + l'
"degl\n" // di + gl'
"della\n" // di + la
"delle\n" // di + le
"nel\n" // in + el
"nello\n" // in + lo
"nei\n" // in + i
"negli\n" // in + gli
"nell\n" // in + l'
"negl\n" // in + gl'
"nella\n" // in + la
"nelle\n" // in + le
"sul\n" // su + il
"sullo\n" // su + lo
"sui\n" // su + i
"sugli\n" // su + gli
"sull\n" // su + l'
"sugl\n" // su + gl'
"sulla\n" // su + la
"sulle\n" // su + le
// Images
"arte\n"
"immagini\n"
"illustrazione\n"
"fotografia\n"
"icona\n"
"bianca\n" // white
"bianco\n" // white
"nera\n" // black
"nero\n" // black
"contenere\n" // contain (image may contain...)
// Months.
"gennaio\n"
"genn\n"
"febbraio\n"
"febbr\n"
"marzo\n"
"mar\n"
"aprile\n"
"apr\n"
"maggio\n"
"magg\n"
"giugno\n"
"luglio\n"
"agosto\n"
"settembre\n"
"sett\n"
"ottobre\n"
"ott\n"
"novembre\n"
"nov\n"
"dicembre\n"
"dic\n"
// Weekdays.
"lunedì\n"
"lun\n"
"martedì\n"
"mar\n"
"mercoledì\n"
"mer\n"
"giovedì\n"
"gio\n"
"venerdì\n"
"ven\n"
"sabato\n"
"sab\n"
"domenica\n"
"dom\n"
//
// German
//
// General German stopwords.
"und\n" // and
"mit\n" // with
"für\n" // for
"der\n" // the
"die\n" // the
"von\n" // of, from
"auf\n" // on
"das\n" // the
"aus\n" // out of
"ist\n" // is
"ein\n" // one
"eine\n" // one
"sie\n" // they, she
"den\n" // the
"zum\n" // zu + dem
"zur\n" // zu + der
"bei\n" // by
"des\n" // the
"sprüche\n" // claims (to be)
"oder\n" // or
// German Image stopwords.
"bild\n" // picture
"bilder\n" // pictures
"foto\n" // photo
"schwarz\n" // black
"weiß\n" // white
// Months.
"januar\n"
"jan\n"
"jän\n"
"februar\n"
"feb\n"
"märz\n"
"april\n"
"apr\n"
"mai\n"
"juni\n"
"juli\n"
"august\n"
"aug\n"
"september\n"
"sept\n"
"oktober\n"
"okt\n"
"november\n"
"nov\n"
"dezember\n"
"dez\n"
// Weekdays.
"montag\n"
"dienstag\n"
"mittwoch\n"
"donnerstag\n"
"freitag\n"
"samstag\n"
"sonntag\n"
//
// Spanish
//
// General Spanish stopwords.
"con\n" // with
"para\n" // by
"del\n" // of the
"que\n" // that
"los\n" // the
"las\n" // the
"una\n" // one
"por\n" // for
"más\n" // more
"como\n" // how
// Spanish image stopwords.
"dibujos\n" // drawings
"imagen\n" // images
"arte\n" // art
"fondo\n" // background
"fondos\n" // backgrounds
"diseño\n" // design
"ilustración\n" // illustration
"imagenes\n" // images
"blanca\n" // white
"blanco\n" // white
"negra\n" // black
"negro\n" // black
// Months.
"enero\n"
"febrero\n"
"feb\n"
"marzo\n"
"abril\n"
"abr\n"
"mayo\n"
"junio\n"
"jun\n"
"julio\n"
"jul\n"
"agosto\n"
"septiembre\n"
"sept\n"
"set\n"
"octubre\n"
"oct\n"
"noviembre\n"
"nov\n"
"diciembre\n"
"dic\n"
// Weekdays. Weekday abbreviations in Spanish are two-letters which
// don't need to be listed here (anything less than 3 letters is
// considered a stopword already).
"lunes\n"
"martes\n"
"miércoles\n"
"jueves\n"
"viernes\n"
"sábado\n"
"domingo\n"
//
// Hindi
//
// General Hindi stopwords.
"में\n" // in
"लिए\n" // for
"नहीं\n" // no
"एक\n" // one
"साथ\n" // with
"दिया\n" // gave
"किया\n" // did
"रहे\n" // are
"सकता\n" // can
"इस\n" // this
"शामिल\n" // include
"तारीख\n" // the date
// Hindi image stopwords.
"चित्र\n" // picture
"वीडियो\n" // video
// Months
"जनवरी\n"
"फरवरी\n"
"मार्च\n"
"अप्रैल\n"
"मई\n"
"जून\n"
"जुलाई\n"
"अगस्त\n"
"सितंबर\n"
"अक्टूबर\n"
"नवंबर\n"
"दिसंबर\n"
// Weekdays.
"सोमवार\n"
"मंगलवार\n"
"बुधवार\n"
"बृहस्पतिवार\n"
"शुक्रवार\n"
"शनिवार\n"
"रविवार\n"};
} // namespace
// static
AXImageStopwords& AXImageStopwords::GetInstance() {
static base::NoDestructor<AXImageStopwords> instance;
return *instance;
}
AXImageStopwords::AXImageStopwords() {
// Parse the newline-delimited stopwords from kImageStopwordsUtf8 and store
// them as a flat_set of type string_view. This is very memory-efficient
// because it avoids ever needing to copy any of the strings; each
// string_view is just a pointer into kImageStopwordsUtf8 and flat_set
// acts like a set but basically just does a binary search.
std::vector<std::string_view> stopwords =
base::SplitStringPiece(kImageStopwordsUtf8, "\n", base::TRIM_WHITESPACE,
base::SPLIT_WANT_NONEMPTY);
// It's inefficient to add things to a flat_set one at a time. Copy them
// all over at once.
stopword_set_ = stopwords;
}
AXImageStopwords::~AXImageStopwords() = default;
bool AXImageStopwords::IsImageStopword(const char* word_utf8) const {
std::u16string word_utf16 = base::UTF8ToUTF16(word_utf8);
// It's not really meaningful, but since short words are stopwords, for
// simplicity we define the empty string to be a stopword too.
if (word_utf16.empty())
return true;
// Canonicalize case, this is like "ToLower" for many languages but
// works independently of the current locale.
word_utf16 = base::i18n::FoldCase(word_utf16);
// Count the number of distinct codepoints from a supported unicode block.
int supported_count = 0;
base::i18n::UTF16CharIterator iter(word_utf16);
for (; !iter.end(); iter.Advance()) {
int32_t codepoint = iter.get();
UBlockCode block_code = ublock_getCode(codepoint);
switch (block_code) {
case UBLOCK_BASIC_LATIN:
case UBLOCK_LATIN_1_SUPPLEMENT:
case UBLOCK_LATIN_EXTENDED_A:
case UBLOCK_DEVANAGARI:
supported_count++;
break;
default:
break;
}
}
// Treat any string of 2 or fewer characters in any of these unicode
// blocks as a stopword. Of course, there are rare exceptions, but they're
// acceptable for this heuristic. All that means is that we won't count
// those words when trying to determine if the alt text for an image
// has meaningful text or not. The odds are good that a 1 or 2 character
// word is not meaningful.
//
// Note: this assumption might not be valid for some unicode blocks,
// like Chinese. That's why the heuristic only applies to certain unicode
// blocks where we believe this to be a reasonable assumption.
//
// Note that in Devanagari (the script used for the Hindi language, among
// others) a word sometimes looks like a single character (one glyph) but it's
// actually two or more unicode codepoints (consonants and vowels that are
// joined together), which is why this heuristic still works. Anything with
// two or fewer unicode codepoints is an extremely short word.
if (supported_count == iter.char_offset() && iter.char_offset() <= 2)
return true;
return base::Contains(stopword_set_, base::UTF16ToUTF8(word_utf16));
}
} // namespace content