Special cases in data.
(
features: Features,
processor: spm.SentencePieceProcessor)
| 361 | |
| 362 | |
| 363 | def _handle_exceptional_examples( |
| 364 | features: Features, |
| 365 | processor: spm.SentencePieceProcessor) -> List[AnswerSpan]: |
| 366 | """Special cases in data.""" |
| 367 | if features.id == 'qw_6687--Viola.txt': |
| 368 | pattern = 'three strings in common—G, D, and A'.encode('utf-8') |
| 369 | i = features.context.find(pattern) |
| 370 | if i != -1: |
| 371 | span = AnswerSpan(i + len(pattern) - 1, i + len(pattern), 'A') |
| 372 | span = realign_answer_span(features, None, processor, span) |
| 373 | assert span is not None, 'Span should exist.' |
| 374 | return [span] |
| 375 | if features.id == 'sfq_26183--Vitamin_A.txt': |
| 376 | pattern = ('Vitamin A is a group of unsaturated nutritional organic ' |
| 377 | 'compounds that includes retinol').encode('utf-8') |
| 378 | i = features.context.find(pattern) |
| 379 | if i != -1: |
| 380 | span = AnswerSpan(i + pattern.find(b'A'), i + pattern.find(b'A') + 1, 'A') |
| 381 | span = realign_answer_span(features, None, processor, span) |
| 382 | assert span is not None, 'Span should exist.' |
| 383 | spans = [span] |
| 384 | span = AnswerSpan(i, i + pattern.find(b'A') + 1, 'Vitamin A') |
| 385 | span = realign_answer_span(features, None, processor, span) |
| 386 | return spans + [span] |
| 387 | if features.id == 'odql_292--Colombia.txt': |
| 388 | pattern = b'Colombia is the third-most populous country in Latin America' |
| 389 | i = features.context.find(pattern) |
| 390 | if i != -1: |
| 391 | span = AnswerSpan(i, i + len(b'Colombia'), 'Colombia') |
| 392 | span = realign_answer_span(features, None, processor, span) |
| 393 | assert span is not None, 'Span should exist.' |
| 394 | return [span] |
| 395 | if features.id == 'tc_1648--Vietnam.txt': |
| 396 | pattern = 'Bảo Đại'.encode('utf-8') |
| 397 | i = features.context.find(pattern) |
| 398 | if i != -1: |
| 399 | span = AnswerSpan(i, i + len(pattern), 'Bảo Đại') |
| 400 | span = realign_answer_span(features, None, processor, span) |
| 401 | assert span is not None, 'Span should exist.' |
| 402 | return [span] |
| 403 | if features.id == 'sfq_22225--Irish_mythology.txt': |
| 404 | pattern = 'Tír na nÓg'.encode('utf-8') |
| 405 | spans = [] |
| 406 | i = 0 |
| 407 | while features.context.find(pattern, i) != -1: |
| 408 | i = features.context.find(pattern) |
| 409 | span = AnswerSpan(i, i + len(pattern), 'Tír na nÓg') |
| 410 | span = realign_answer_span(features, None, processor, span) |
| 411 | assert span is not None, 'Span should exist.' |
| 412 | spans.append(span) |
| 413 | i += len(pattern) |
| 414 | return spans |
| 415 | return [] |
| 416 | |
| 417 | |
| 418 | class FindAnswerSpans(beam.DoFn): |
no test coverage detected