-
Notifications
You must be signed in to change notification settings - Fork 1.5k
/
c4.py
506 lines (451 loc) · 20.9 KB
/
c4.py
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
# coding=utf-8
# Copyright 2021 The TensorFlow Datasets Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""C4 dataset based on Common Crawl."""
import collections
import functools
import json
import os
import uuid
from absl import logging
import tensorflow.compat.v2 as tf
import tensorflow_datasets.public_api as tfds
from tensorflow_datasets.text import c4_utils
_DESCRIPTION = """\
A colossal, cleaned version of Common Crawl's web crawl corpus.
Based on Common Crawl dataset: https://commoncrawl.org
To generate this dataset, please follow
[the instructions from t5](https://github.com/google-research/text-to-text-transfer-transformer#c4).
Due to the overhead of cleaning the dataset, it is recommend you prepare it with
a distributed service like Cloud Dataflow. More info at
https://www.tensorflow.org/datasets/beam_datasets.
"""
_CITATION = """
@article{2019t5,
author = {Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu},
title = {Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer},
journal = {arXiv e-prints},
year = {2019},
archivePrefix = {arXiv},
eprint = {1910.10683},
}
"""
_VERSION = tfds.core.Version("3.0.1")
# TODO(adarob): Remove supported versions. Starting with 3.0.0, all generated
# datasets are automatically forward compatible. For example,
# tfds.load('c4:3.0.0') works even if the code is at 3.0.1.
_SUPPORTED_VERSIONS = [
tfds.core.Version("2.3.1"),
tfds.core.Version("2.3.0"),
tfds.core.Version("2.2.1"),
tfds.core.Version("2.2.0"),
]
RELEASE_NOTES = {
"3.0.1": "Remove mC4 languages with less than 10k pages.",
"3.0.0": "Add multilingual version (mC4). Deterministic URL deduplication.",
"2.3.1": "Hashing change.",
"2.3.0": "Deduplicate lines within a page.",
"2.2.1": "Update dataset_info.json",
}
_DOWNLOAD_HOST = "https://commoncrawl.s3.amazonaws.com"
_WET_PATH_URL = "https://commoncrawl.s3.amazonaws.com/crawl-data/CC-MAIN-{cc_version}/wet.paths.gz"
_REALNEWS_DOMAINS_URL = "https://raw.githubusercontent.com/rowanz/grover/38f7184bd87237ae2d3bc330b99f1e2e246f6d51/realnews/domain_to_allowed_subdomains.json"
_CHECKSUMS_URL = "https://storage.googleapis.com/tfds-data/manual_checksums/c4.txt"
_OPENWEBTEXT_URLS_ZIP = "OpenWebText.zip"
_OPENWEBTEXT_URLS_URL = "https://mega.nz/#F!EZZD0YwJ!9_PlEQzdMVLaNdKv_ICNVQ"
_OPENWEBTEXT_URLS_FILE_PATTERN = "OpenWebText/Version 1/URLs/*.txt"
_EN_BADWORDS_URL = "https://raw.githubusercontent.com/LDNOOBW/List-of-Dirty-Naughty-Obscene-and-Otherwise-Bad-Words/25e679f03d96baa721cde20db9944649e8d0a844/en"
_BADWORDS_URL = "https://raw.githubusercontent.com/LDNOOBW/List-of-Dirty-Naughty-Obscene-and-Otherwise-Bad-Words/5faf2ba42d7b1c0977169ec3611df25a3c08eb13/{lang}"
_BADWORDS_LANGS = [
"ar", "cs", "da", "de", "en", "eo", "es", "fa", "fi", "fil", "fr",
"fr-CA-u-sd-caqc", "hi", "hu", "it", "ja", "kab", "ko", "nl", "no", "pl",
"pt", "ru", "sv", "th", "tlh", "tr", "zh"
]
DEFAULT_CC_VERSION = "2019-18"
OPENWEBTEXT_CC_VERSIONS = ( # August 2018 - July 2019
"2019-18", # Original default for single-crawl dataset (April 2019).
"2019-30",
"2019-26",
"2019-22",
"2019-13",
"2019-09",
"2019-04",
"2018-51",
"2018-47",
"2018-43",
"2018-39",
"2018-34")
ALL_CC_VERSIONS = ( # as of September 23, 2020
"2013-20", "2013-48", "2014-10", "2014-15", "2014-23", "2014-35", "2014-41",
"2014-42", "2014-49", "2014-52", "2015-06", "2015-11", "2015-14", "2015-18",
"2015-22", "2015-27", "2015-32", "2015-35", "2015-40", "2015-48", "2016-07",
"2016-18", "2016-22", "2016-26", "2016-30", "2016-36", "2016-40", "2016-44",
"2016-50", "2017-04", "2017-09", "2017-13", "2017-17", "2017-22", "2017-26",
"2017-30", "2017-34", "2017-39", "2017-43", "2017-47", "2017-51", "2018-05",
"2018-09", "2018-13", "2018-17", "2018-22", "2018-26", "2018-30", "2018-34",
"2018-39", "2018-43", "2018-47", "2018-51", "2019-04", "2019-09", "2019-13",
"2019-18", "2019-22", "2019-26", "2019-30", "2019-35", "2019-39", "2019-43",
"2019-47", "2019-51", "2020-05", "2020-10", "2020-16", "2020-24", "2020-29",
"2020-34")
_KNOWN_CORRUPT_WET_FILES = ( # as of September 23, 2020
# files that raise EOFError
"crawl-data/CC-MAIN-2016-50/segments/1480698543577.51/wet/CC-MAIN-20161202170903-00294-ip-10-31-129-80.ec2.internal.warc.wet.gz",
"crawl-data/CC-MAIN-2017-43/segments/1508187823309.55/wet/CC-MAIN-20171019141046-20171019161046-00789.warc.wet.gz",
"crawl-data/CC-MAIN-2017-47/segments/1510934805466.25/wet/CC-MAIN-20171119080836-20171119100836-00043.warc.wet.gz",
"crawl-data/CC-MAIN-2017-47/segments/1510934805466.25/wet/CC-MAIN-20171119080836-20171119100836-00043.warc.wet.gz",
"crawl-data/CC-MAIN-2017-47/segments/1510934805809.59/wet/CC-MAIN-20171119210640-20171119230640-00044.warc.wet.gz",
"crawl-data/CC-MAIN-2017-47/segments/1510934806543.24/wet/CC-MAIN-20171122084446-20171122104446-00120.warc.wet.gz",
"crawl-data/CC-MAIN-2017-51/segments/1512948517350.12/wet/CC-MAIN-20171212153808-20171212173808-00039.warc.wet.gz",
"crawl-data/CC-MAIN-2018-05/segments/1516084887660.30/wet/CC-MAIN-20180118230513-20180119010513-00778.warc.wet.gz",
"crawl-data/CC-MAIN-2018-09/segments/1518891815951.96/wet/CC-MAIN-20180224211727-20180224231727-00311.warc.wet.gz",
"crawl-data/CC-MAIN-2018-26/segments/1529267863518.39/wet/CC-MAIN-20180620104904-20180620124904-00402.warc.wet.gz",
"crawl-data/CC-MAIN-2018-26/segments/1529267863518.39/wet/CC-MAIN-20180620104904-20180620124904-00402.warc.wet.gz",
"crawl-data/CC-MAIN-2018-26/segments/1529267865995.86/wet/CC-MAIN-20180624005242-20180624025242-00197.warc.wet.gz",
"crawl-data/CC-MAIN-2018-30/segments/1531676591837.34/wet/CC-MAIN-20180720213434-20180720233434-00442.warc.wet.gz",
"crawl-data/CC-MAIN-2018-34/segments/1534221211167.1/wet/CC-MAIN-20180816191550-20180816211550-00078.warc.wet.gz",
"crawl-data/CC-MAIN-2018-34/segments/1534221211185.57/wet/CC-MAIN-20180816211126-20180816231126-00076.warc.wet.gz",
"crawl-data/CC-MAIN-2018-34/segments/1534221211185.57/wet/CC-MAIN-20180816211126-20180816231126-00076.warc.wet.gz",
"crawl-data/CC-MAIN-2018-34/segments/1534221219109.94/wet/CC-MAIN-20180821210655-20180821230655-00654.warc.wet.gz",
"crawl-data/CC-MAIN-2019-47/segments/1573496672170.93/wet/CC-MAIN-20191122222322-20191123011322-00000.warc.wet.gz",
"crawl-data/CC-MAIN-2020-24/segments/1590347458095.68/wet/CC-MAIN-20200604192256-20200604222256-00235.warc.wet.gz",
"crawl-data/CC-MAIN-2020-34/segments/1596439738819.78/wet/CC-MAIN-20200811180239-20200811210239-00123.warc.wet.gz",
# files that raise UnicodeDecodeError
"crawl-data/CC-MAIN-2017-13/segments/1490218203536.73/wet/CC-MAIN-20170322213003-00052-ip-10-233-31-227.ec2.internal.warc.wet.gz",
)
# Limited to languages in CLD3 that produce at least 10k pages when using the
# "multilingual" config below.
MC4_LANGUAGES = [
"af", "am", "ar", "az", "be", "bg", "bg-Latn", "bn", "ca", "ceb", "co",
"cs", "cy", "da", "de", "el", "el-Latn", "en", "eo", "es", "et", "eu", "fa",
"fi", "fil", "fr", "fy", "ga", "gd", "gl", "gu", "ha", "haw", "hi",
"hi-Latn", "hmn", "ht", "hu", "hy", "id", "ig", "is", "it", "iw", "ja",
"ja-Latn", "jv", "ka", "kk", "km", "kn", "ko", "ku", "ky", "la", "lb", "lo",
"lt", "lv", "mg", "mi", "mk", "ml", "mn", "mr", "ms", "mt", "my", "ne",
"nl", "no", "ny", "pa", "pl", "ps", "pt", "ro", "ru", "ru-Latn", "sd", "si",
"sk", "sl", "sm", "sn", "so", "sq", "sr", "st", "su", "sv", "sw", "ta",
"te", "tg", "th", "tr", "uk", "ur", "uz", "vi", "xh", "yi", "yo", "zh",
"zh-Latn", "zu"
]
class C4Config(tfds.core.BuilderConfig):
"""BuilderConfig for C4 dataset."""
def __init__(self,
name,
languages,
cc_versions=None,
clean=False,
badwords_filter=False,
paragraph_filter=False,
dedupe=True,
realnewslike=False,
webtextlike=False,
**kwargs):
"""BuilderConfig for C4.
Args:
name: string, the name for the config.
languages: list(string), the language code(s) to include.
cc_versions: tuple(string), a collection of versions of Common Crawl to
use as the raw source text. Set to None to use default.
clean: bool, whether to heuristically filter out lines and pages
considered low quality. Note: only expected to work reliably for English
pages.
badwords_filter: bool, whether to filter out pages with badwords.
paragraph_filter: bool, whether to filter out pages with too few or too
short paragraphs.
dedupe: bool, whether to deduplicate the dataset by paragraphs.
realnewslike: bool, whether to limit to news domains as compiled by
RealNews.
webtextlike: bool, whether to limit to WebText-like URLs.
**kwargs: keyword arguments forwarded to super.
"""
super(C4Config, self).__init__(
name=name,
version=_VERSION,
supported_versions=_SUPPORTED_VERSIONS,
**kwargs)
if clean and tuple(languages) != ("en",):
logging.warn(
"C4 cleaning is only expected to work reliably for English pages.")
self.languages = languages
self.cc_versions = cc_versions or (DEFAULT_CC_VERSION,)
self.clean = clean
self.badwords_filter = badwords_filter
self.paragraph_filter = paragraph_filter
self.dedupe = dedupe
self.realnewslike = realnewslike
self.webtextlike = webtextlike
class C4(tfds.core.BeamBasedBuilder):
"""C4 dataset based on Common Crawl."""
MANUAL_DOWNLOAD_INSTRUCTIONS = f"""
You are using a C4 config that requires some files to be manually downloaded.
For `c4/webtextlike`, download {_OPENWEBTEXT_URLS_ZIP} from
{_OPENWEBTEXT_URLS_URL}.
"""
BUILDER_CONFIGS = [
C4Config(
"en",
languages=["en"],
clean=True,
dedupe=True,
badwords_filter=True,
description="English C4 dataset."),
C4Config(
"en.noclean",
languages=["en"],
clean=False,
dedupe=False,
badwords_filter=False,
description="Disables all cleaning (deduplication, removal based on bad words, "
"etc.)"),
C4Config(
"realnewslike",
languages=["en"],
realnewslike=True,
clean=True,
dedupe=True,
badwords_filter=True,
description="Filters from the default config to only include content from the "
"domains used in the 'RealNews' dataset (Zellers et al., 2019)."),
C4Config(
"webtextlike",
languages=["en"],
cc_versions=OPENWEBTEXT_CC_VERSIONS,
webtextlike=True,
clean=True,
dedupe=True,
badwords_filter=True,
description="Filters from the default config to only include content from the "
"URLs in OpenWebText (https://github.com/jcpeterson/openwebtext)."),
C4Config(
"multilingual",
languages=MC4_LANGUAGES,
cc_versions=ALL_CC_VERSIONS,
clean=False,
paragraph_filter=True,
dedupe=True,
badwords_filter=True,
description="Multilingual C4 (mC4) has 101 languages and is generated from 71 "
"Common Crawl dumps."),
]
def _info(self):
features = {
"text": tfds.features.Text(),
"url": tfds.features.Text(),
"content-type": tfds.features.Text(),
"content-length": tfds.features.Text(),
"timestamp": tfds.features.Text(),
}
return tfds.core.DatasetInfo(
builder=self,
description=_DESCRIPTION,
features=tfds.features.FeaturesDict(features),
citation=_CITATION,
homepage="https://github.com/google-research/text-to-text-transfer-transformer#datasets",
)
def _split_generators(self, dl_manager, pipeline):
dl_manager.download_checksums(_CHECKSUMS_URL)
# We will automatically download the first default CC version, but others
# need to be manually downloaded.
cc_versions = set(self.builder_config.cc_versions)
files_to_download = {}
files_to_download["wet_path_urls"] = [
_WET_PATH_URL.format(cc_version=cc_version)
for cc_version in cc_versions
]
if self.builder_config.badwords_filter:
files_to_download["badwords"] = {
lang: _BADWORDS_URL.format(lang=lang)
for lang in _BADWORDS_LANGS
if lang != "en"
}
# Use older "en" file for reproducibility of the original C4.
files_to_download["badwords"]["en"] = _EN_BADWORDS_URL
if self.builder_config.realnewslike:
files_to_download["realnews_domains"] = _REALNEWS_DOMAINS_URL
file_paths = dl_manager.download_and_extract(files_to_download)
if self.builder_config.webtextlike:
owt_path = os.path.join(dl_manager.manual_dir, _OPENWEBTEXT_URLS_ZIP)
if not tf.io.gfile.exists(owt_path):
raise AssertionError(
"For the WebText-like config, you must manually download the "
"following file from {0} and place it in {1}: {2}".format(
_OPENWEBTEXT_URLS_URL, dl_manager.manual_dir,
_OPENWEBTEXT_URLS_ZIP))
file_paths["openwebtext_urls_zip"] = dl_manager.extract(owt_path)
file_paths = tf.nest.map_structure(os.fspath, file_paths)
page_content_pcollection = self._get_page_content(pipeline, file_paths,
dl_manager)
def _lang_filter(url_and_page, lang):
_, page = url_and_page
return page["language"] == lang
def _filter(url_and_page, lang, predicate_fn):
return (_lang_filter(url_and_page, lang) and
c4_utils.get_hashed_url_filter_fn(predicate_fn)(url_and_page))
train_predicate_fn = lambda x: x % 1000 != 0 # 99.9%
validation_predicate_fn = lambda x: x % 1000 == 0 # 00.1%
if len(self.builder_config.languages) == 1:
# Single-language version.
return [
tfds.core.SplitGenerator(
name=tfds.Split.TRAIN,
gen_kwargs=dict(
split="train",
page_content=page_content_pcollection,
split_filter_fn=c4_utils.get_hashed_url_filter_fn(
predicate_fn=train_predicate_fn)),
),
tfds.core.SplitGenerator(
name=tfds.Split.VALIDATION,
gen_kwargs=dict(
split="validation",
page_content=page_content_pcollection,
split_filter_fn=c4_utils.get_hashed_url_filter_fn(
predicate_fn=validation_predicate_fn)),
),
]
splits = []
for lang in self.builder_config.languages + [c4_utils.UNKNOWN_LANGUAGE]:
splits.extend([
tfds.core.SplitGenerator(
name=lang,
gen_kwargs=dict(
split=lang,
page_content=page_content_pcollection,
split_filter_fn=functools.partial(
_filter, lang=lang, predicate_fn=train_predicate_fn),
)),
tfds.core.SplitGenerator(
name=f"{lang}-validation",
gen_kwargs=dict(
split=f"{lang}-validation",
page_content=page_content_pcollection,
split_filter_fn=functools.partial(
_filter, lang=lang, predicate_fn=validation_predicate_fn),
))
])
return splits
def _get_page_content(self, pipeline, file_paths, dl_manager):
"""Build PCollection of un-split page content."""
beam = tfds.core.lazy_imports.apache_beam
def download_wet_file(path, dl_dir):
url = f"{_DOWNLOAD_HOST}/{path}"
out_path = f"{dl_dir}/{path}"
if tf.io.gfile.exists(out_path):
c4_utils.get_counter_inc_fn("download_wet_url")("exists")
return out_path
tmp_dir = f"{out_path}.incomplete{uuid.uuid4().hex}"
try:
tf.io.gfile.makedirs(tmp_dir)
downloader = tfds.download.downloader.get_downloader()
with downloader.tqdm():
dl_path = downloader.download(url, tmp_dir).get().path
tf.io.gfile.rename(os.fspath(dl_path), out_path, overwrite=True)
finally:
if tf.io.gfile.exists(tmp_dir):
tf.io.gfile.rmtree(tmp_dir)
c4_utils.get_counter_inc_fn("download_wet_url")("downloaded")
return out_path
wet_file_paths = (
pipeline
| "create_wet_path_urls" >> beam.Create(file_paths["wet_path_urls"])
| beam.io.ReadAllFromText(
compression_type=beam.io.filesystem.CompressionTypes.UNCOMPRESSED)
# Increase parallelism.
| beam.Reshuffle()
| "filter_corrupt_wet_files" >>
beam.Filter(lambda p: p not in _KNOWN_CORRUPT_WET_FILES)
| beam.Map(
download_wet_file,
dl_dir=os.path.join(dl_manager.download_dir, "c4_wet_files")))
# Parse WET files and filter by length.
# Output: url, text
page_content = (
wet_file_paths
| beam.FlatMap(c4_utils.split_wet_file)
| beam.Filter(c4_utils.is_valid_length))
# Optionally filter for RealNews domains.
# Output: url, text
if self.builder_config.realnewslike:
with tf.io.gfile.GFile(file_paths["realnews_domains"]) as f:
realnews_domains = json.load(f)
page_content = (
page_content
| beam.Filter(c4_utils.is_realnews_domain, realnews_domains))
# Normalize and deduplicate by URL.
# Output: url, text
page_content = (
page_content
| "normalize_url" >> beam.Map(c4_utils.normalize_url)
| "group_url" >> beam.GroupByKey()
| beam.Map(c4_utils.dedupe_urls))
# Optionally filter for WebText-like URLs.
# Output: url, text
if self.builder_config.webtextlike:
webtextlike_urls = (
pipeline
| "read_webtextlike_urls" >> beam.io.ReadFromText(
os.path.join(file_paths["openwebtext_urls_zip"],
_OPENWEBTEXT_URLS_FILE_PATTERN))
| "add_dummy_page" >> beam.Map(lambda x: (x, ""))
| "normal_webtext_url" >> beam.Map(c4_utils.normalize_url))
page_content = ({
"text": page_content,
"webtextlike_urls": webtextlike_urls
}
| "group_webtextlike_urls" >> beam.CoGroupByKey()
| beam.FlatMap(c4_utils.filter_by_webtextlike))
if self.builder_config.paragraph_filter:
page_content |= beam.Filter(c4_utils.paragraph_filter)
if self.builder_config.clean:
page_content = (
page_content
| "clean_pages" >> beam.FlatMap(c4_utils.get_clean_page_fn()))
if self.builder_config.dedupe:
page_content = (
# Also removes documents with too few sentences after deduplication.
c4_utils.remove_duplicate_text(page_content) # pylint:disable=g-long-ternary
if self.builder_config.clean else
# If we are not cleaning, do not remove too-few-sentence documents.
c4_utils.remove_duplicate_text(page_content, min_num_sentences=0))
# Add detected language.
if self.builder_config.languages == ["en"]:
# Use langdetect for reproducibility of the original C4.
page_content |= beam.FlatMap(c4_utils.detect_english)
else:
page_content = c4_utils.detect_languages(
page_content, valid_languages=self.builder_config.languages)
if self.builder_config.badwords_filter:
# Create dictionary of badwords regex for each available language.
badwords = collections.defaultdict(set)
for lang, path in file_paths["badwords"].items():
lang = lang.split("-")[0] # remove suffix if present
with tf.io.gfile.GFile(path) as f:
badwords[lang].update(l.strip() for l in f)
page_content |= beam.Filter(c4_utils.get_badwords_filter_fn(badwords))
return page_content
def _build_pcollection(self, unused_pipeline, split, page_content,
split_filter_fn):
beam = tfds.core.lazy_imports.apache_beam
def _emit_examples(el):
c4_utils.get_counter_inc_fn(split)("examples")
_, features = el
return features["url"], {
"url": features["url"],
"text": features["text"],
"content-type": features["content-type"],
"content-length": features["content-length"],
"timestamp": features["timestamp"]
}
return (page_content
| beam.Filter(split_filter_fn)
| beam.Map(_emit_examples))