Datasets:

Modalities:
Text
Libraries:
Datasets
License:
File size: 28,688 Bytes
41a7e13
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bbcca53
41a7e13
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
# coding=utf-8
# Copyright 2022 The HuggingFace Datasets Authors and the current dataset script contributor.
#
# 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.

"""
A dataset loader for the MuchMore Springer Bilingual Corpus

homepage

* https://muchmore.dfki.de/resources1.htm

description of annotation format

* https://muchmore.dfki.de/pubs/D4.1.pdf

Four files are distributed

* springer_english_train_plain.tar.gz (english plain text of abstracts)
* springer_german_train_plain.tar.gz (german plain text of abstracts)
* springer_english_train_V4.2.tar.gz (annotated xml in english)
* springer_german_train_V4.2.tar.gz (annotated xml in german)

Each tar file has one member file per abstract.
There are keys to join the english and german files
but there is not a 1-1 mapping between them (i.e. some
english files have no german counterpart and some german
files have no english counterpart). However, there is a 1-1
mapping between plain text and annotations for a given language
(i.e. an abstract in springer_english_train_plain.tar.gz will
also be found in springer_english_train_V4.2.tar.gz)

Counts,

* 15,631 total abstracts
* 7,823 english abstracts
* 7,808 german abstracts
* 6,374 matched (en/de) abstracts
* 1,449 english abstracts with no german
* 1,434 german abstracts with no english

Notes

* Arthroskopie.00130237.eng.abstr.chunkmorph.annotated.xml seems to be empty


* entity spans can overlap. an example from the first sample:

{'id': 'Arthroskopie.00130003.eng.abstr-s1-t1',
  'type': 'umlsterm',
  'text': ['posterior'],
  'offsets': [[4, 13]],
  'normalized': [{'db_name': 'UMLS', 'db_id': 'C0032009'}]},
{'id': 'Arthroskopie.00130003.eng.abstr-s1-t8',
  'type': 'umlsterm',
  'text': ['posterior cruciate ligament'],
  'offsets': [[4, 31]],
  'normalized': [{'db_name': 'UMLS', 'db_id': 'C0080039'}]},
{'id': 'Arthroskopie.00130003.eng.abstr-s1-t2',
  'type': 'umlsterm',
  'text': ['ligament'],
  'offsets': [[23, 31]],
  'normalized': [{'db_name': 'UMLS', 'db_id': 'C0023685'},
   {'db_name': 'UMLS', 'db_id': 'C0023686'}]},


* semantic relations are defined beween concepts but entities can
  have multiple concpets associated with them. in the bigbio
  schema we skip relations between multiple concept of the
  same entity. an example of a relation that is kept from the
  source schema is below,

In [35]: dsd['train'][0]['sentences'][0]['tokens']
Out[35]:
[{'id': 'w1', 'pos': 'DT', 'lemma': 'the', 'text': 'The'},
 {'id': 'w2', 'pos': 'JJ', 'lemma': 'posterior', 'text': 'posterior'},
 {'id': 'w3', 'pos': 'JJ', 'lemma': 'cruciate', 'text': 'cruciate'},
 {'id': 'w4', 'pos': 'NN', 'lemma': 'ligament', 'text': 'ligament'},
 {'id': 'w5', 'pos': 'PUNCT', 'lemma': None, 'text': '('},
 {'id': 'w6', 'pos': 'NN', 'lemma': None, 'text': 'PCL'},
 {'id': 'w7', 'pos': 'PUNCT', 'lemma': None, 'text': ')'},
 {'id': 'w8', 'pos': 'VBZ', 'lemma': 'be', 'text': 'is'},
 {'id': 'w9', 'pos': 'DT', 'lemma': 'the', 'text': 'the'},
 {'id': 'w10', 'pos': 'JJS', 'lemma': 'strong', 'text': 'strongest'},
 {'id': 'w11', 'pos': 'NN', 'lemma': 'ligament', 'text': 'ligament'},
 {'id': 'w12', 'pos': 'IN', 'lemma': 'of', 'text': 'of'},
 {'id': 'w13', 'pos': 'DT', 'lemma': 'the', 'text': 'the'},
 {'id': 'w14', 'pos': 'JJ', 'lemma': 'human', 'text': 'human'},
 {'id': 'w15', 'pos': 'NN', 'lemma': 'knee', 'text': 'knee'},
 {'id': 'w16', 'pos': 'JJ', 'lemma': 'joint', 'text': 'joint'},
 {'id': 'w17', 'pos': 'PUNCT', 'lemma': None, 'text': '.'}]


In [36]: dsd['train'][0]['sentences'][0]['semrels'][0]
Out[36]: {'id': 'r1', 'term1': 't3.1', 'term2': 't6.1', 'reltype': 'surrounds'}

In [37]: dsd['train'][0]['sentences'][0]['umlsterms'][2]
Out[37]:
{'id': 't3',
 'from': 'w11',
 'to': 'w11',
 'concepts': [{'id': 't3.1',
   'cui': 'C0023685',
   'preferred': 'Ligaments',
   'tui': 'T024',
   'mshs': [{'code': 'A2.513'}]},
  {'id': 't3.2',
   'cui': 'C0023686',
   'preferred': 'Articular ligaments',
   'tui': 'T023',
   'mshs': [{'code': 'A2.513.514'}, {'code': 'A2.835.583.512'}]}]}

In [38]: dsd['train'][0]['sentences'][0]['umlsterms'][5]
Out[38]:
{'id': 't6',
 'from': 'w16',
 'to': 'w16',
 'concepts': [{'id': 't6.1',
   'cui': 'C0022417',
   'preferred': 'Joints',
   'tui': 'T030',
   'mshs': [{'code': 'A2.835.583'}]}]}

"""

import itertools
import os
import re
import tarfile
import xml.etree.ElementTree as ET
from collections import defaultdict
from typing import Dict, List
from xml.etree.ElementTree import Element

import datasets
from datasets import Features, Value

# TODO: home page has a list of publications but its not clear which to choose
# https://muchmore.dfki.de/papers1.htm
# to start, chose the one below.
# Buitelaar, Paul / Declerck, Thierry / Sacaleanu, Bogdan / Vintar, Spela / Raileanu, Diana / Crispi, Claudia: A Multi-Layered, XML-Based Approach to the Integration of Linguistic and Semantic Annotations. In: Proceedings of EACL 2003 Workshop on Language Technology and the Semantic Web (NLPXML’03), Budapest, Hungary, April 2003.
from .bigbiohub import kb_features
from .bigbiohub import text2text_features
from .bigbiohub import BigBioConfig
from .bigbiohub import Tasks

_LANGUAGES = ['English', 'German']
_PUBMED = True
_LOCAL = False
_CITATION = """\
@inproceedings{buitelaar2003multi,
  title={A multi-layered, xml-based approach to the integration of linguistic and semantic annotations},
  author={Buitelaar, Paul and Declerck, Thierry and Sacaleanu, Bogdan and Vintar, {\v{S}}pela and Raileanu, Diana and Crispi, Claudia},
  booktitle={Proceedings of EACL 2003 Workshop on Language Technology and the Semantic Web (NLPXML'03), Budapest, Hungary},
  year={2003}
}
"""

_DESCRIPTION = """\
The corpus used in the MuchMore project is a parallel corpus of English-German scientific
medical abstracts obtained from the Springer Link web site. The corpus consists
approximately of 1 million tokens for each language. Abstracts are from 41 medical
journals, each of which constitutes a relatively homogeneous medical sub-domain (e.g.
Neurology, Radiology, etc.). The corpus of downloaded HTML documents is normalized in
various ways, in order to produce a clean, plain text version, consisting of a title, abstract
and keywords. Additionally, the corpus was aligned on the sentence level.

Automatic (!) annotation includes: Part-of-Speech; Morphology (inflection and
decomposition); Chunks; Semantic Classes (UMLS: Unified Medical Language System,
MeSH: Medical Subject Headings, EuroWordNet); Semantic Relations from UMLS.
"""

_DATASETNAME = "muchmore"
_DISPLAYNAME = "MuchMore"

_HOMEPAGE = "https://muchmore.dfki.de/resources1.htm"

# TODO: website says the following, but don't see a specific license
# TODO: add to FAQs about what to do in this situation.

# "The cross-lingual information access prototype system for the medical domain
# will be made publicly accessible through the internet. It provides access to
# multilingual information on the basis of a domain ontology and classification.
# For the main task of multilingual domain modelling, the project will focus
# on German and English. "
_LICENSE = 'License information unavailable'
_URLs = {
    "muchmore_source": [
        "https://muchmore.dfki.de/pubs/springer_english_train_plain.tar.gz",
        "https://muchmore.dfki.de/pubs/springer_english_train_V4.2.tar.gz",
        "https://muchmore.dfki.de/pubs/springer_german_train_plain.tar.gz",
        "https://muchmore.dfki.de/pubs/springer_german_train_V4.2.tar.gz",
    ],
    "muchmore_bigbio_kb": [
        "https://muchmore.dfki.de/pubs/springer_english_train_V4.2.tar.gz",
        "https://muchmore.dfki.de/pubs/springer_german_train_V4.2.tar.gz",
    ],
    "muchmore_en_bigbio_kb": "https://muchmore.dfki.de/pubs/springer_english_train_V4.2.tar.gz",
    "muchmore_de_bigbio_kb": "https://muchmore.dfki.de/pubs/springer_german_train_V4.2.tar.gz",
    "plain": [
        "https://muchmore.dfki.de/pubs/springer_english_train_plain.tar.gz",
        "https://muchmore.dfki.de/pubs/springer_german_train_plain.tar.gz",
    ],
    "plain_en": "https://muchmore.dfki.de/pubs/springer_english_train_plain.tar.gz",
    "plain_de": "https://muchmore.dfki.de/pubs/springer_german_train_plain.tar.gz",
    "muchmore_bigbio_t2t": [
        "https://muchmore.dfki.de/pubs/springer_english_train_plain.tar.gz",
        "https://muchmore.dfki.de/pubs/springer_german_train_plain.tar.gz",
    ],
}

# took version from annotated file names
_SOURCE_VERSION = "4.2.0"
_BIGBIO_VERSION = "1.0.0"
_SUPPORTED_TASKS = [
    Tasks.TRANSLATION,
    Tasks.NAMED_ENTITY_RECOGNITION,
    Tasks.NAMED_ENTITY_DISAMBIGUATION,
    Tasks.RELATION_EXTRACTION,
]

NATIVE_ENCODING = "ISO-8859-1"
FILE_NAME_PATTERN = r"^(.+?)\.(eng|ger)\.abstr(\.chunkmorph\.annotated\.xml)?$"
LANG_MAP = {"eng": "en", "ger": "de"}


class MuchMoreDataset(datasets.GeneratorBasedBuilder):
    """MuchMore Springer Bilingual Corpus"""

    DEFAULT_CONFIG_NAME = "muchmore_source"
    SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
    BIGBIO_VERSION = datasets.Version(_BIGBIO_VERSION)

    BUILDER_CONFIGS = [
        BigBioConfig(
            name="muchmore_source",
            version=SOURCE_VERSION,
            description="MuchMore source schema",
            schema="source",
            subset_id="muchmore",
        ),
        BigBioConfig(
            name="muchmore_bigbio_kb",
            version=BIGBIO_VERSION,
            description="MuchMore simplified BigBio kb schema",
            schema="bigbio_kb",
            subset_id="muchmore",
        ),
        BigBioConfig(
            name="muchmore_en_bigbio_kb",
            version=BIGBIO_VERSION,
            description="MuchMore simplified BigBio kb schema",
            schema="bigbio_kb",
            subset_id="muchmore_en",
        ),
        BigBioConfig(
            name="muchmore_de_bigbio_kb",
            version=BIGBIO_VERSION,
            description="MuchMore simplified BigBio kb schema",
            schema="bigbio_kb",
            subset_id="muchmore_de",
        ),
        BigBioConfig(
            name="muchmore_bigbio_t2t",
            version=BIGBIO_VERSION,
            description="MuchMore simplified BigBio translation schema",
            schema="bigbio_t2t",
            subset_id="muchmore",
        ),
    ]

    # default config produces english annotations at the moment
    def _info(self):

        if self.config.schema == "source":
            features = Features(
                {
                    "sample_id": Value("string"),
                    "corresp": Value("string"),
                    "language": Value("string"),
                    "abstract": Value("string"),
                    "sentences": [
                        {
                            "id": Value("string"),
                            "corresp": Value("string"),
                            "umlsterms": [
                                {
                                    "id": Value("string"),
                                    "from": Value("string"),
                                    "to": Value("string"),
                                    "concepts": [
                                        {
                                            "id": Value("string"),
                                            "cui": Value("string"),
                                            "preferred": Value("string"),
                                            "tui": Value("string"),
                                            "mshs": [
                                                {
                                                    "code": Value("string"),
                                                }
                                            ],
                                        }
                                    ],
                                }
                            ],
                            "ewnterms": [
                                {
                                    "id": Value("string"),
                                    "to": Value("string"),
                                    "from": Value("string"),
                                    "senses": [
                                        {
                                            "offset": Value("string"),
                                        }
                                    ],
                                }
                            ],
                            "semrels": [
                                {
                                    "id": Value("string"),
                                    "term1": Value("string"),
                                    "term2": Value("string"),
                                    "reltype": Value("string"),
                                }
                            ],
                            "chunks": [
                                {
                                    "id": Value("string"),
                                    "to": Value("string"),
                                    "from": Value("string"),
                                    "type": Value("string"),
                                }
                            ],
                            "tokens": [
                                {
                                    "id": Value("string"),
                                    "pos": Value("string"),
                                    "lemma": Value("string"),
                                    "text": Value("string"),
                                }
                            ],
                        }
                    ],
                }
            )

        elif self.config.schema == "bigbio_kb":
            features = kb_features

        elif self.config.name in ("plain", "plain_en", "plain_de"):
            features = Features(
                {
                    "sample_id": Value("string"),
                    "sample_id_prefix": Value("string"),
                    "language": Value("string"),
                    "abstract": Value("string"),
                }
            )

        elif self.config.schema == "bigbio_t2t":
            features = text2text_features

        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            supervised_keys=None,
            homepage=_HOMEPAGE,
            license=str(_LICENSE),
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""
        my_urls = _URLs[self.config.name]
        data_dirs = dl_manager.download(my_urls)
        # ensure that data_dirs is always a list of string paths
        if isinstance(data_dirs, str):
            data_dirs = [data_dirs]

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "file_names_and_pointers": itertools.chain(
                        *[dl_manager.iter_archive(data_dir) for data_dir in data_dirs]
                    ),
                    "split": "train",
                },
            ),
        ]

    @staticmethod
    def _get_umlsterms_from_xsent(xsent: Element) -> List:
        xumlsterms = xsent.find("./umlsterms")

        umlsterms = []
        for xumlsterm in xumlsterms.findall("./umlsterm"):

            concepts = []
            for xconcept in xumlsterm.findall("./concept"):

                mshs = [
                    {"code": xmsh.get("code")} for xmsh in xconcept.findall("./msh")
                ]

                concept = {
                    "id": xconcept.get("id"),
                    "cui": xconcept.get("cui"),
                    "preferred": xconcept.get("preferred"),
                    "tui": xconcept.get("tui"),
                    "mshs": mshs,
                }
                concepts.append(concept)

            umlsterm = {
                "id": xumlsterm.get("id"),
                "from": xumlsterm.get("from"),
                "to": xumlsterm.get("to"),
                "concepts": concepts,
            }
            umlsterms.append(umlsterm)

        return umlsterms

    @staticmethod
    def _get_ewnterms_from_xsent(xsent: Element) -> List:
        xewnterms = xsent.find("./ewnterms")

        ewnterms = []
        for xewnterm in xewnterms.findall("./ewnterm"):

            senses = [
                {"offset": xsense.get("offset")}
                for xsense in xewnterm.findall("./sense")
            ]

            ewnterm = {
                "id": xewnterm.get("id"),
                "from": xewnterm.get("from"),
                "to": xewnterm.get("to"),
                "senses": senses,
            }
            ewnterms.append(ewnterm)

        return ewnterms

    @staticmethod
    def _get_semrels_from_xsent(xsent: Element) -> List[Dict[str, str]]:
        xsemrels = xsent.find("./semrels")
        return [
            {
                "id": xsemrel.get("id"),
                "term1": xsemrel.get("term1"),
                "term2": xsemrel.get("term2"),
                "reltype": xsemrel.get("reltype"),
            }
            for xsemrel in xsemrels.findall("./semrel")
        ]

    @staticmethod
    def _get_chunks_from_xsent(xsent: Element) -> List[Dict[str, str]]:
        xchunks = xsent.find("./chunks")
        return [
            {
                "id": xchunk.get("id"),
                "to": xchunk.get("to"),
                "from": xchunk.get("from"),
                "type": xchunk.get("type"),
            }
            for xchunk in xchunks.findall("./chunk")
        ]

    @staticmethod
    def _get_tokens_from_xsent(xsent: Element) -> List[Dict[str, str]]:
        xtext = xsent.find("./text")
        return [
            {
                "id": xtoken.get("id"),
                "pos": xtoken.get("pos"),
                "lemma": xtoken.get("lemma"),
                "text": xtoken.text,
            }
            for xtoken in xtext.findall("./token")
        ]

    def _generate_original_examples(self, file_names_and_pointers):
        """Generate something close to the original dataset.

        This will yield one sample per abstract with the plaintext
        and the annotations combined into one object. If an abstract
        is available in both english and german each language version
        will be a distinct example.
        """
        abstracts = {}
        samples = {}
        for file_name, fp in file_names_and_pointers:

            if file_name.endswith(".abstr"):
                sample_id = file_name
                abstracts[sample_id] = fp.read().decode(NATIVE_ENCODING)

            elif file_name.endswith(".abstr.chunkmorph.annotated.xml"):
                content_bytes = fp.read()
                content_str = content_bytes.decode(NATIVE_ENCODING)
                if content_str == "":
                    continue

                xroot = ET.fromstring(content_str)

                sentences = []
                for xsent in xroot.findall("./"):
                    sentence = {
                        "id": xsent.get("id"),
                        "corresp": xsent.get("corresp"),
                        "umlsterms": self._get_umlsterms_from_xsent(xsent),
                        "ewnterms": self._get_ewnterms_from_xsent(xsent),
                        "semrels": self._get_semrels_from_xsent(xsent),
                        "chunks": self._get_chunks_from_xsent(xsent),
                        "tokens": self._get_tokens_from_xsent(xsent),
                    }
                    sentences.append(sentence)

                sample_id = xroot.get("id")
                samples[sample_id] = {
                    "sample_id": sample_id,
                    "corresp": xroot.get("corresp"),
                    "language": xroot.get("lang"),
                    "sentences": sentences,
                }

        for _id, (sample_id, sample) in enumerate(samples.items()):
            sample["abstract"] = abstracts[sample_id]
            yield _id, sample

    def _generate_bigbio_kb_examples(self, file_names_and_pointers):
        """Generate big science biomedical kb examples."""

        def snippets_tokens_from_sents(sentences):
            snippets = []
            for sentence in sentences:
                snippet = [el["text"] for el in sentence["tokens"]]
                snippets.append(snippet)
            return snippets

        def sid_to_text_off(sid, snip_txts_lens):
            ii_sid = int(sid[1:])
            start = sum(snip_txts_lens[: ii_sid - 1]) + (ii_sid - 1)
            end = start + snip_txts_lens[ii_sid - 1]
            return start, end

        def sid_wid_to_text_off(sid, wid, snip_txts_lens, snip_toks_lens):
            s_start, s_end = sid_to_text_off(sid, snip_txts_lens)
            ii_sid = int(sid[1:])
            ii_wid = int(wid[1:])
            w_start = sum(snip_toks_lens[ii_sid - 1][: ii_wid - 1]) + (ii_wid - 1)
            start = s_start + w_start
            end = start + snip_toks_lens[ii_sid - 1][ii_wid - 1]
            return start, end

        for _id, (file_name, fp) in enumerate(file_names_and_pointers):

            content_bytes = fp.read()
            content_str = content_bytes.decode(NATIVE_ENCODING)
            if content_str == "":
                continue

            xroot = ET.fromstring(content_str)

            sentences = []
            for xsent in xroot.findall("./"):
                sentence = {
                    "id": xsent.get("id"),
                    "corresp": xsent.get("corresp"),
                    "umlsterms": self._get_umlsterms_from_xsent(xsent),
                    "ewnterms": self._get_ewnterms_from_xsent(xsent),
                    "semrels": self._get_semrels_from_xsent(xsent),
                    "chunks": self._get_chunks_from_xsent(xsent),
                    "tokens": self._get_tokens_from_xsent(xsent),
                }
                sentences.append(sentence)

            snip_toks = snippets_tokens_from_sents(sentences)
            snip_txts = [" ".join(snip_tok) for snip_tok in snip_toks]
            snip_txts_lens = [len(el) for el in snip_txts]
            snip_toks_lens = [[len(tok) for tok in snip] for snip in snip_toks]
            text = " ".join(snip_txts)
            passages = [
                {
                    "id": "{}-passage-0".format(xroot.get("id")),
                    "type": "abstract",
                    "text": [text],
                    "offsets": [(0, len(text))],
                }
            ]

            entities = []
            rel_map = {}
            for sentence in sentences:
                sid = sentence["id"]
                ii_sid = int(sid[1:])

                for umlsterm in sentence["umlsterms"]:
                    umlsterm_id = umlsterm["id"]
                    entity_id = f"{sid}-{umlsterm_id}"
                    wid_from = umlsterm["from"]
                    wid_to = umlsterm["to"]
                    ii_wid_from = int(wid_from[1:])
                    ii_wid_to = int(wid_to[1:])

                    tok_text = " ".join(
                        snip_toks[ii_sid - 1][ii_wid_from - 1 : ii_wid_to]
                    )
                    w_from_start, w_from_end = sid_wid_to_text_off(
                        sid, wid_from, snip_txts_lens, snip_toks_lens
                    )
                    w_to_start, w_to_end = sid_wid_to_text_off(
                        sid, wid_to, snip_txts_lens, snip_toks_lens
                    )

                    offsets = [(w_from_start, w_to_end)]
                    main_text = text[w_from_start:w_to_end]
                    umls_cuis = [el["cui"] for el in umlsterm["concepts"]]
                    for concept in umlsterm["concepts"]:
                        rel_map[concept["id"]] = entity_id

                    entity = {
                        "id": "{}-{}".format(xroot.get("id"), entity_id),
                        "offsets": offsets,
                        "text": [tok_text],
                        "type": "umlsterm",
                        "normalized": [
                            {"db_name": "UMLS", "db_id": cui} for cui in umls_cuis
                        ],
                    }
                    entities.append(entity)

            relations = []
            for sentence in sentences:
                sid = sentence["id"]
                for semrel in sentence["semrels"]:
                    semrel_id = semrel["id"]
                    rel_id = "{}-{}-{}-{}".format(
                        sid, semrel_id, semrel["term1"], semrel["term2"],
                    )
                    arg1_id = "{}-{}".format(xroot.get("id"), rel_map[semrel["term1"]])
                    arg2_id = "{}-{}".format(xroot.get("id"), rel_map[semrel["term2"]])
                    # some semrels are between multiple normalizations of
                    # a single entity. we skip these. see docstring at top
                    # of module for more complete description
                    if arg1_id == arg2_id:
                        continue
                    relation = {
                        "id": "{}-{}".format(xroot.get("id"), rel_id),
                        "type": semrel["reltype"],
                        "arg1_id": arg1_id,
                        "arg2_id": arg2_id,
                        "normalized": []
                    }
                    relations.append(relation)

            yield _id, {
                "id": xroot.get("id"),
                "document_id": xroot.get("id"),
                "passages": passages,
                "entities": entities,
                "coreferences": [],
                "events": [],
                "relations": relations,
            }

    def _generate_plain_examples(self, file_names_and_pointers):
        """Generate plain text abstract examples."""
        for _id, (file_name, fp) in enumerate(file_names_and_pointers):
            match = re.match(FILE_NAME_PATTERN, file_name)
            yield _id, {
                "sample_id_prefix": match.group(1),
                "sample_id": file_name,
                "language": LANG_MAP[match.group(2)],
                "abstract": fp.read().decode(NATIVE_ENCODING),
            }

    def _generate_translation_examples(self, file_names_and_pointers):
        sample_map = defaultdict(list)
        for file_name, fp in file_names_and_pointers:
            if file_name.endswith("eng.abstr"):
                language = "en"
            elif file_name.endswith("ger.abstr"):
                language = "de"
            else:
                raise ValueError()
            sample_id_prefix = re.sub(".(eng|ger).abstr$", "", file_name)
            sample_id = file_name
            abstract = fp.read().decode(NATIVE_ENCODING)
            sample_map[sample_id_prefix].append(
                {"language": language, "sample_id": sample_id, "abstract": abstract}
            )

        _id = 0
        for sample_id_prefix, sample_pair in sample_map.items():
            if len(sample_pair) != 2:
                continue
            en_idx = 0 if sample_pair[0]["language"] == "en" else 1
            de_idx = 0 if en_idx == 1 else 1
            yield _id, {
                "id": sample_id_prefix,
                "document_id": sample_id_prefix,
                "text_1": sample_pair[en_idx]["abstract"],
                "text_2": sample_pair[de_idx]["abstract"],
                "text_1_name": "en",
                "text_2_name": "de",
            }
            _id += 1

    def _generate_examples(self, file_names_and_pointers, split):

        if self.config.schema == "source":
            genny = self._generate_original_examples(file_names_and_pointers)

        elif self.config.schema == "bigbio_kb":
            genny = self._generate_bigbio_kb_examples(file_names_and_pointers)

        elif self.config.name in ("plain", "plain_en", "plain_de"):
            genny = self._generate_plain_examples(file_names_and_pointers)

        elif self.config.schema == "bigbio_t2t":
            genny = self._generate_translation_examples(file_names_and_pointers)

        for _id, sample in genny:
            yield _id, sample