aboutsummaryrefslogtreecommitdiffstats
path: root/skate/verify.go
blob: 18b2f4eca4819b2bd985220bf4b9908402b3096e (plain)
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
// TODO: The various grouping and verification functions should probably be in
// a separate file and it should be obvious how to adjust or write a new one.

//go:generate stringer -type=Status,Reason -output verify_string.go verify.go
package skate

import (
	"bytes"
	"fmt"
	"io"
	"regexp"
	"strconv"
	"strings"
	"unicode/utf8"

	"git.archive.org/martin/cgraph/skate/set"
	json "github.com/segmentio/encoding/json"
)

// This file contains a port of fuzzycat.verify to Go.

type (
	// Status represents match strength.
	Status int
	// Reason gives more context to status result.
	Reason int
)

const (
	StatusUnknown Status = iota
	StatusExact
	StatusStrong
	StatusWeak
	StatusDifferent
	StatusAmbiguous

	ReasonUnknown Reason = iota
	ReasonAppendix
	ReasonArxiv
	ReasonArxivVersion
	ReasonBlacklisted
	ReasonBlacklistedFragment
	ReasonBookChapter
	ReasonChemFormula
	ReasonComponent
	ReasonContainer
	ReasonContainerNameBlacklist
	ReasonContribIntersectionEmpty
	ReasonCustomBSISubdoc
	ReasonCustomBSIUndated
	ReasonCustomIEEEArxiv
	ReasonCustomIOPMAPattern
	ReasonCustomPrefix1014288
	ReasonCustomPrefix105860ChoiceReview
	ReasonCustomPrefix107916
	ReasonCustomVHS
	ReasonDOI
	ReasonDataciteRelatedID
	ReasonDataciteVersion
	ReasonDatasetDOI
	ReasonFigshareVersion
	ReasonJaccardAuthors
	ReasonJstorID
	ReasonMaxClusterSizeExceeded
	ReasonNumDiff
	ReasonPMCID
	ReasonPMID
	ReasonPMIDDOIPair
	ReasonPageCount
	ReasonPreprintPublished
	ReasonPublisherBlacklist
	ReasonReleaseType
	ReasonSharedDOIPrefix
	ReasonShortTitle
	ReasonSingularCluster
	ReasonSlugTitleAuthorMatch
	ReasonSubtitle
	ReasonTitleArtifact
	ReasonTitleAuthorMatch
	ReasonTitleFilename
	ReasonTokenizedAuthors
	ReasonVersionedDOI
	ReasonWorkID
	ReasonYear
)

// Short name.
func (s Status) Short() string {
	return strings.ToLower(strings.Replace(s.String(), "Status", "", 1))
}

// Short name.
func (r Reason) Short() string {
	return strings.ToLower(strings.Replace(r.String(), "Reason", "", 1))
}

// MatchResult is the result of a verification.
type MatchResult struct {
	Status Status
	Reason Reason
}

// VerificationPair groups two identifiers and their match status and
// match reason.
type MatchPair struct {
	A      string
	B      string
	Result MatchResult
}

// AsLine returns a TSV line of the match pair.
func (m *MatchPair) AsLine() string {
	return fmt.Sprintf("%s\t%s\t%s\t%s\n", m.A, m.B, m.Result.Status, m.Result.Reason)
}

var (
	PatAppendix        = regexp.MustCompile(`appendix ?[^ ]*$`)
	PatFigshareVersion = regexp.MustCompile(`[.]v[0-9]+$`)
	PatVersionedDOI    = regexp.MustCompile(`10[.].*/v[0-9]{1,}$`)
	PatArxivVersion    = regexp.MustCompile(`(.*)v[0-9]{1,2}$`)
	PatFilenameLike    = regexp.MustCompile(`.*[.][a-z]{2,3}$`)
	PatDigits          = regexp.MustCompile(`\d+`)
	PatPages           = regexp.MustCompile(`([0-9]{1,})-([0-9]{1,})`)
)

// JsonMarshalNewline marshals a value as JSON and adds a newline.
func JsonMarshalNewline(v interface{}) ([]byte, error) {
	b, err := json.Marshal(v)
	if err != nil {
		return nil, err
	}
	b = append(b, []byte("\n")...)
	return b, nil
}

// ClusterVerifyMaxClusterSize runs verification across all pairs in the cluster. This is a
// port of https://git.io/JYgOB from fuzzycat. This is good for "self-match" verification.
func ClusterVerifyMaxClusterSize(p []byte, maxClusterSize int) ([]byte, error) {
	var (
		rc  *ReleaseCluster
		buf bytes.Buffer
		n   int
	)
	if err := json.Unmarshal(p, &rc); err != nil {
		return nil, err
	}
	if n = len(rc.Values); n > maxClusterSize {
		return nil, nil
	}
	// O(n^2) ahead, specifically, n * (n-1) / 2.
	for i := 0; i < n; i++ {
		for j := i; j < n; j++ {
			if i == j {
				continue
			}
			a := rc.Values[i]
			b := rc.Values[j]
			matchPair := &MatchPair{
				A:      a.Ident,
				B:      b.Ident,
				Result: Verify(a, b),
			}
			if _, err := io.WriteString(&buf, matchPair.AsLine()); err != nil {
				return nil, err
			}
		}
	}
	return buf.Bytes(), nil
}

// ClusterVerify runs verification process across all pairs, but skips clusters
// containing more than ten elements. If a cluster has more then 10 elements,
// it might also signal a too ambiguous title. Beside, we do not want this to
// be too slow.
func ClusterVerify(p []byte) ([]byte, error) {
	return ClusterVerifyMaxClusterSize(p, 10)
}

// RefClusterVerify deserializes a cluster document containing both converted
// references and releases and returns a tabular verification result between
// one (any) release and all references found. This depends on refs and releases
// being distinguishable, (e.g. via .extra.skate.status == "ref").
func RefClusterVerify(p []byte) ([]byte, error) {
	var (
		rc        *ReleaseCluster
		buf       bytes.Buffer
		pivot, re *Release
		err       error
	)
	if err = json.Unmarshal(p, &rc); err != nil {
		return nil, err
	}
	if pivot, err = rc.OneNonRef(); err != nil {
		return nil, err
	}
	for _, re = range rc.Values {
		if re.Extra.Skate.Status != "ref" {
			continue
		}
		matchPair := &MatchPair{
			A:      pivot.Ident,
			B:      re.Ident,
			Result: Verify(pivot, re),
		}
		if _, err := io.WriteString(&buf, matchPair.AsLine()); err != nil {
			return nil, err
		}
	}
	return buf.Bytes(), nil
}

// RefClusterToBiblioRef runs verification and creates a BiblioRef schema from
// exact and strong matches only.
func RefClusterToBiblioRef(p []byte) ([]byte, error) {
	var (
		rc        *ReleaseCluster
		br        *BiblioRef
		buf       bytes.Buffer
		pivot, re *Release
		err       error
	)
	if err = json.Unmarshal(p, &rc); err != nil {
		return nil, err
	}
	if pivot, err = rc.OneNonRef(); err != nil {
		return nil, err
	}
	for _, re = range rc.Values {
		if re.Extra.Skate.Status != "ref" {
			continue
		}
		result := Verify(pivot, re)
		switch result.Status {
		case StatusExact, StatusStrong:
			if result.Reason == ReasonDOI {
				continue // Assume we already have the DOI matches.
			}
			br = generateBiblioRef(re, pivot, result, "fuzzy")
			return JsonMarshalNewline(br)
		default:
			// XXX: may want to include non matches here.
			continue
		}
	}
	return buf.Bytes(), nil
}

// generateBiblioRef generates a bibliographic schema document. XXX: This is a bit odd.
func generateBiblioRef(source, target *Release, matchResult MatchResult, provenance string) *BiblioRef {
	var bref BiblioRef
	bref.SourceReleaseIdent = source.Ident
	bref.SourceWorkIdent = source.WorkID
	bref.SourceReleaseStage = source.ReleaseStage
	if source.ReleaseYear() > 1000 {
		bref.SourceYear = source.ReleaseYearString()
	}
	bref.RefIndex = source.Extra.Skate.Ref.Index
	bref.RefKey = source.Extra.Skate.Ref.Key
	bref.TargetReleaseIdent = target.Ident
	bref.TargetWorkIdent = target.WorkID
	bref.MatchProvenance = provenance
	bref.MatchStatus = matchResult.Status.Short()
	bref.MatchReason = matchResult.Reason.Short()
	return &bref
}

// Verify verifies two releases and will ignore short titles.
func Verify(a, b *Release) MatchResult {
	return VerifyMinTitleLength(a, b, 5)
}

// Verify follows the fuzzycat (Python) implementation of this function: it
// compares two release entities. The Go version can be used for large batch
// processing (where the Python version might take two or more days).
func VerifyMinTitleLength(a, b *Release, minTitleLength int) MatchResult {
	if a.ExtIDs.DOI != "" && a.ExtIDs.DOI == b.ExtIDs.DOI {
		return MatchResult{StatusExact, ReasonDOI}
	}
	if a.WorkID != "" && a.WorkID == b.WorkID {
		return MatchResult{StatusExact, ReasonWorkID}
	}
	aTitleLower := strings.ToLower(a.Title)
	bTitleLower := strings.ToLower(b.Title)
	if utf8.RuneCountInString(a.Title) < minTitleLength {
		return MatchResult{StatusAmbiguous, ReasonShortTitle}
	}
	if BlacklistTitle.Contains(aTitleLower) {
		return MatchResult{StatusAmbiguous, ReasonBlacklisted}
	}
	if BlacklistTitle.Contains(bTitleLower) {
		return MatchResult{StatusAmbiguous, ReasonBlacklisted}
	}
	for _, fragment := range BlacklistTitleFragments.Slice() {
		if strings.Contains(aTitleLower, fragment) {
			return MatchResult{StatusAmbiguous, ReasonBlacklistedFragment}
		}
	}
	if strings.Contains(aTitleLower, "subject index") && strings.Contains(bTitleLower, "subject index") {
		if a.ContainerID != "" && a.ContainerID != b.ContainerID {
			return MatchResult{StatusDifferent, ReasonContainer}
		}
	}
	if a.Title != "" && a.Title == b.Title &&
		a.Extra.DataCite.MetadataVersion > 0 && b.Extra.DataCite.MetadataVersion > 0 &&
		a.Extra.DataCite.MetadataVersion != b.Extra.DataCite.MetadataVersion {
		return MatchResult{StatusExact, ReasonDataciteVersion}
	}
	if strings.HasPrefix(a.ExtIDs.DOI, "10.14288/") && strings.HasPrefix(b.ExtIDs.DOI, "10.14288/") &&
		a.ExtIDs.DOI != b.ExtIDs.DOI {
		return MatchResult{StatusDifferent, ReasonCustomPrefix1014288}
	}
	if strings.HasPrefix(a.ExtIDs.DOI, "10.3403") && strings.HasPrefix(b.ExtIDs.DOI, "10.3403") {
		if a.ExtIDs.DOI+"u" == b.ExtIDs.DOI || b.ExtIDs.DOI+"u" == a.ExtIDs.DOI {
			return MatchResult{StatusStrong, ReasonCustomBSIUndated}
		}
		aSubtitle := a.Subtitle()
		bSubtitle := b.Subtitle()
		if a.Title != "" && a.Title == b.Title &&
			((len(aSubtitle) > 0 && aSubtitle[0] != "" && len(bSubtitle) == 0) ||
				(len(aSubtitle) == 0 && len(bSubtitle) > 0 && bSubtitle[0] != "")) {
			return MatchResult{StatusStrong, ReasonCustomBSISubdoc}
		}
	}
	if strings.HasPrefix(a.ExtIDs.DOI, "10.1149") && strings.HasPrefix(b.ExtIDs.DOI, "10.1149") {
		v := "10.1149/ma"
		if (strings.HasPrefix(a.ExtIDs.DOI, v) && !strings.HasPrefix(b.ExtIDs.DOI, v)) ||
			(!strings.HasPrefix(a.ExtIDs.DOI, v) && strings.HasPrefix(b.ExtIDs.DOI, v)) {
			return MatchResult{StatusDifferent, ReasonCustomIOPMAPattern}
		}
	}
	if strings.Contains(a.Title, "Zweckverband Volkshochschule") && a.Title != b.Title {
		return MatchResult{StatusDifferent, ReasonCustomVHS}
	}
	if PatAppendix.MatchString(a.Title) {
		return MatchResult{StatusAmbiguous, ReasonAppendix}
	}
	if strings.HasPrefix(a.ExtIDs.DOI, "10.6084/") && strings.HasPrefix(b.ExtIDs.DOI, "10.6084/") {
		av := PatFigshareVersion.ReplaceAllString(a.ExtIDs.DOI, "")
		bv := PatFigshareVersion.ReplaceAllString(b.ExtIDs.DOI, "")
		if av == bv {
			return MatchResult{StatusStrong, ReasonFigshareVersion}
		}
	}
	if PatVersionedDOI.MatchString(a.ExtIDs.DOI) && PatVersionedDOI.MatchString(b.ExtIDs.DOI) {
		return MatchResult{StatusStrong, ReasonVersionedDOI}
	}
	if looksLikeComponent(a.ExtIDs.DOI, b.ExtIDs.DOI) {
		return MatchResult{StatusStrong, ReasonVersionedDOI}
	}
	if len(a.Extra.DataCite.Relations) > 0 || len(b.Extra.DataCite.Relations) > 0 {
		getRelatedDOI := func(rel *Release) set.Set {
			ss := set.New()
			for _, rel := range rel.Extra.DataCite.Relations {
				if strings.ToLower(rel.RelatedIdentifierType) != "doi" {
					continue
				}
				ss.Add(rel.RelatedIdentifier())
			}
			return ss
		}
		aRelated := getRelatedDOI(a)
		bRelated := getRelatedDOI(b)
		if aRelated.Contains(b.ExtIDs.DOI) || bRelated.Contains(a.ExtIDs.DOI) {
			return MatchResult{StatusStrong, ReasonDataciteRelatedID}
		}
	}
	if a.ExtIDs.Arxiv != "" && b.ExtIDs.Arxiv != "" {
		aSub := PatArxivVersion.FindStringSubmatch(a.ExtIDs.Arxiv)
		bSub := PatArxivVersion.FindStringSubmatch(b.ExtIDs.Arxiv)
		if len(aSub) == 2 && len(bSub) == 2 && aSub[1] == bSub[1] {
			return MatchResult{StatusStrong, ReasonArxivVersion}
		}
	}
	if a.ReleaseType != b.ReleaseType {
		types := set.FromSlice([]string{a.ReleaseType, b.ReleaseType})
		ignoreTypes := set.FromSlice([]string{"article", "article-journal", "report", "paper-conference"})
		if types.Intersection(ignoreTypes).IsEmpty() {
			return MatchResult{StatusDifferent, ReasonReleaseType}
		}
		if types.Contains("dataset") && (types.Contains("article") || types.Contains("article-journal")) {
			return MatchResult{StatusDifferent, ReasonReleaseType}
		}
		if types.Contains("book") && (types.Contains("article") || types.Contains("article-journal")) {
			return MatchResult{StatusDifferent, ReasonReleaseType}
		}
	}
	if a.ReleaseType == "dataset" && b.ReleaseType == "dataset" && a.ExtIDs.DOI != b.ExtIDs.DOI {
		return MatchResult{StatusDifferent, ReasonDatasetDOI}
	}
	if a.ReleaseType == "chapter" && b.ReleaseType == "chapter" &&
		a.Extra.ContainerName != "" && a.Extra.ContainerName != b.Extra.ContainerName {
		return MatchResult{StatusDifferent, ReasonBookChapter}
	}
	if a.Extra.Crossref.Type == "component" && a.Title != b.Title {
		return MatchResult{StatusDifferent, ReasonComponent}
	}
	if a.ReleaseType == "component" && b.ReleaseType == "component" {
		if a.ExtIDs.DOI != "" && a.ExtIDs.DOI != b.ExtIDs.DOI {
			return MatchResult{StatusDifferent, ReasonComponent}
		}
	}
	aSlugTitle := strings.TrimSpace(strings.Replace(slugifyString(a.Title), "\n", " ", -1))
	bSlugTitle := strings.TrimSpace(strings.Replace(slugifyString(b.Title), "\n", " ", -1))

	if aSlugTitle == bSlugTitle {
		if a.ReleaseYear() != 0 && b.ReleaseYear() != 0 && absInt(a.ReleaseYear()-b.ReleaseYear()) > 40 {
			return MatchResult{StatusDifferent, ReasonYear}
		}
	}
	if aSlugTitle == bSlugTitle {
		ieeeArxivCheck := func(a, b *Release) (ok bool) {
			return doiPrefix(a.ExtIDs.DOI) == "10.1109" && b.ExtIDs.Arxiv != ""
		}
		if ieeeArxivCheck(a, b) || ieeeArxivCheck(b, a) {
			return MatchResult{StatusStrong, ReasonCustomIEEEArxiv}
		}
	}
	if aSlugTitle == bSlugTitle {
		if strings.HasPrefix(a.ExtIDs.DOI, "10.7916/") && strings.HasPrefix(b.ExtIDs.DOI, "10.7916/") {
			return MatchResult{StatusAmbiguous, ReasonCustomPrefix107916}
		}
	}
	if aSlugTitle == bSlugTitle {
		aSubtitle := a.Subtitle()
		bSubtitle := b.Subtitle()
		for _, aSub := range aSubtitle {
			for _, bSub := range bSubtitle {
				if slugifyString(aSub) != slugifyString(bSub) {
					return MatchResult{StatusDifferent, ReasonSubtitle}
				}
			}
		}
	}
	rawAuthors := func(rel *Release) (names []string) {
		for _, c := range rel.Contribs {
			name := strings.TrimSpace(c.RawName)
			if name == "" {
				continue
			}
			names = append(names, name)
		}
		return names
	}
	aAuthors := set.FromSlice(rawAuthors(a))
	bAuthors := set.FromSlice(rawAuthors(b))
	aSlugAuthors := set.FromSlice(mapString(slugifyString, aAuthors.Slice()))
	bSlugAuthors := set.FromSlice(mapString(slugifyString, bAuthors.Slice()))
	if aTitleLower == bTitleLower {
		if aAuthors.Len() > 0 && aAuthors.Equals(bAuthors) {
			if a.ReleaseYear() > 0 && b.ReleaseYear() > 0 && absInt(a.ReleaseYear()-b.ReleaseYear()) > 4 {
				return MatchResult{StatusDifferent, ReasonYear}
			}
			return MatchResult{StatusExact, ReasonTitleAuthorMatch}
		}
	}
	if looksLikeFilename(a.Title) || looksLikeFilename(b.Title) {
		if a.Title != b.Title {
			return MatchResult{StatusDifferent, ReasonTitleFilename}
		}
	}
	if a.Title != "" && a.Title == b.Title {
		if a.ReleaseYear() > 0 && b.ReleaseYear() > 0 && absInt(a.ReleaseYear()-b.ReleaseYear()) > 2 {
			return MatchResult{StatusDifferent, ReasonYear}
		}
	}
	// XXX: skipping chemical formula detection (to few cases; https://git.io/Jtdax)
	if len(aSlugTitle) < 10 && aSlugTitle != bSlugTitle {
		return MatchResult{StatusAmbiguous, ReasonShortTitle}
	}
	if PatDigits.MatchString(aSlugTitle) &&
		aSlugTitle != bSlugTitle &&
		unifyDigits(aSlugTitle) == unifyDigits(bSlugTitle) {
		return MatchResult{StatusDifferent, ReasonNumDiff}
	}
	if aSlugTitle != "" && bSlugTitle != "" &&
		strings.ReplaceAll(aSlugTitle, " ", "") == strings.ReplaceAll(bSlugTitle, " ", "") {
		if aSlugAuthors.Intersection(bSlugAuthors).Len() > 0 {
			if a.ReleaseYear() > 0 && b.ReleaseYear() > 0 && absInt(a.ReleaseYear()-b.ReleaseYear()) > 4 {
				return MatchResult{StatusDifferent, ReasonYear}
			}
			return MatchResult{StatusStrong, ReasonSlugTitleAuthorMatch}
		}
	}
	if a.ReleaseYear() > 0 && a.ReleaseYear() == b.ReleaseYear() && aTitleLower == bTitleLower {
		if (a.ExtIDs.PMID != "" && b.ExtIDs.DOI != "") || (b.ExtIDs.PMID != "" && a.ExtIDs.DOI != "") {
			return MatchResult{StatusStrong, ReasonPMIDDOIPair}
		}
	}
	if a.ExtIDs.Jstor != "" && b.ExtIDs.Jstor != "" && a.ExtIDs.Jstor != b.ExtIDs.Jstor {
		return MatchResult{StatusDifferent, ReasonJstorID}
	}
	if a.ContainerID != "" && a.ContainerID == b.ContainerID && a.ExtIDs.DOI != b.ExtIDs.DOI &&
		doiPrefix(a.ExtIDs.DOI) != "10.1126" &&
		doiPrefix(a.ExtIDs.DOI) == doiPrefix(b.ExtIDs.DOI) {
		return MatchResult{StatusDifferent, ReasonSharedDOIPrefix}
	}
	if aAuthors.Len() > 0 && aSlugAuthors.Intersection(bSlugAuthors).IsEmpty() {
		numAuthors := set.Min(aSlugAuthors, bSlugAuthors)
		score := averageScore(aSlugAuthors, bSlugAuthors)
		if (numAuthors < 3 && score > 0.9) || (numAuthors >= 3 && score > 0.5) {
			return MatchResult{StatusStrong, ReasonTokenizedAuthors}
		}
		aTok := set.FromSlice(strings.Fields(aSlugAuthors.Join(" ")))
		bTok := set.FromSlice(strings.Fields(bSlugAuthors.Join(" ")))
		aTok = set.Filter(aTok, func(s string) bool {
			return len(s) > 2
		})
		bTok = set.Filter(bTok, func(s string) bool {
			return len(s) > 2
		})
		if aTok.Len() > 0 && bTok.Len() > 0 {
			if aTok.Jaccard(bTok) > 0.35 {
				return MatchResult{StatusStrong, ReasonJaccardAuthors}
			}
		}
		return MatchResult{StatusDifferent, ReasonContribIntersectionEmpty}
	}
	if doiPrefix(a.ExtIDs.DOI) == "10.5860" || doiPrefix(b.ExtIDs.DOI) == "10.5860" {
		return MatchResult{StatusAmbiguous, ReasonCustomPrefix105860ChoiceReview}
	}
	// XXX: parse pages
	aParsedPages := parsePageString(a.Pages)
	bParsedPages := parsePageString(b.Pages)
	if aParsedPages.Err != nil && bParsedPages.Err != nil {
		if absInt(aParsedPages.Count()-bParsedPages.Count()) > 5 {
			return MatchResult{StatusDifferent, ReasonPageCount}
		}
	}
	if aAuthors.Equals(bAuthors) &&
		a.ContainerID == b.ContainerID &&
		a.ReleaseYear() == b.ReleaseYear() &&
		a.Title != b.Title &&
		(strings.Contains(a.Title, b.Title) || strings.Contains(b.Title, a.Title)) {
		return MatchResult{StatusStrong, ReasonTitleArtifact}
	}
	return MatchResult{
		StatusAmbiguous,
		ReasonUnknown,
	}
}

type ParsedPages struct {
	Start int
	End   int
	Err   error
}

func (pp *ParsedPages) Count() int {
	return pp.End - pp.Start + 1
}

func parsePageString(s string) *ParsedPages {
	s = strings.TrimSpace(s)
	var pp = ParsedPages{}
	if len(s) == 0 {
		pp.Err = fmt.Errorf("parse pages: empty string")
		return &pp
	}
	matches := PatPages.FindStringSubmatch(s)
	if len(matches) != 3 {
		pp.Err = fmt.Errorf("parse pages: no page pattern")
		return &pp
	}
	start, end := matches[1], matches[2]
	if len(end) == 1 && len(start) > 1 && start[len(start)-1] < end[0] {
		end = fmt.Sprintf("%s%c", start[:len(start)-1], end[0])
	}
	if pp.Start, pp.Err = strconv.Atoi(start); pp.Err != nil {
		return &pp
	}
	if pp.End, pp.Err = strconv.Atoi(end); pp.Err != nil {
		return &pp
	}
	if pp.Start > pp.End {
		pp.Err = fmt.Errorf("invalid page count: %s", s)
	}
	return &pp
}

// averageScore take a limited set of authors and calculates pairwise
// similarity scores, then returns the average of the best scores; between 0
// and 1. XXX: This should be revisited and factored out; reading:
// https://github.com/djudd/human-name.
func averageScore(a, b set.Set) float64 {
	aTrimmed := a.TopK(5)
	bTrimmed := b.TopK(5)
	maxScores := make(map[string]float64) // For each a, keep the max.
	for _, pair := range aTrimmed.Product(bTrimmed) {
		a, b := pair[0], pair[1]
		score := authorSimilarityScore(a, b)
		if v, ok := maxScores[a]; !ok || score > v {
			maxScores[a] = score
		}
	}
	var sum, avg float64
	for _, v := range maxScores {
		sum += v
	}
	avg = sum / float64(len(maxScores))
	return avg
}

// authorSimilarityScore is a hacky similarity score.
func authorSimilarityScore(s, t string) float64 {
	ss := set.FromSlice(tokenNgrams(s, 2))
	ts := set.FromSlice(tokenNgrams(t, 2))
	return ss.Jaccard(ts)
}

// tokenNgrams are groups of n char-tokens per word-token in string, e.g. for
// n=2 and string "Anne K Lam", we would get ["an", "ne", "k", "la", "m"].
func tokenNgrams(s string, n int) (result []string) {
	var buf bytes.Buffer
	for _, token := range tokenizeString(s) {
		buf.Reset()
		for i, c := range token {
			if i > 0 && i%n == 0 {
				result = append(result, buf.String())
				buf.Reset()
			}
			buf.WriteRune(c) // XXX: skipping error handling
		}
		result = append(result, buf.String())
	}
	return
}

func tokenizeString(s string) []string {
	return strings.Fields(strings.ToLower(s))
}

func doiPrefix(s string) string {
	parts := strings.Split(s, "/")
	return parts[0]
}

// unifyDigits replaces all digit groups with a hopefully rare placeholder,
// e.g. "<NUM>"; This is for discovering very similar, yet different
// publications, where e.g. titles differ only by a single char representing a
// year. Examples are yearly publications, e.g. "World Health Report 2020",
// where any plain similarity score would yield a high number, yet publications
// are obviously not the same.
func unifyDigits(s string) string {
	return PatDigits.ReplaceAllString(s, "<NUM>")
}

// looksLikeFilename returns true, if the given string could be a filename.
func looksLikeFilename(s string) bool {
	if len(strings.Fields(s)) > 1 {
		return false
	}
	return PatFilenameLike.MatchString(s)
}

// mapString applies a function on each element of a string slice.
func mapString(f func(string) string, vs []string) (result []string) {
	for _, v := range vs {
		result = append(result, f(v))
	}
	return result
}

// absInt returns the absolute value of an int.
func absInt(v int) int {
	if v < 0 {
		return -v
	}
	return v
}

// slugifyString is a basic string slugifier.
func slugifyString(s string) string {
	var buf bytes.Buffer
	for _, c := range strings.TrimSpace(strings.ToLower(s)) {
		if (c > 96 && c < 123) || (c > 47 && c < 58) || (c == 32) || (c == 9) || (c == 10) {
			fmt.Fprintf(&buf, "%c", c)
		}
	}
	return strings.Join(strings.Fields(buf.String()), " ")
}

// looksLikeComponent returns true, if either a looks like a component of b, or vice versa.
func looksLikeComponent(a, b string) bool {
	ac := strings.Split(a, ".")
	bc := strings.Split(b, ".")
	if len(ac) > 1 {
		if strings.Join(ac[0:len(ac)-1], ".") == b {
			return true
		}
	}
	if len(bc) > 1 {
		if strings.Join(bc[0:len(bc)-1], ".") == a {
			return true
		}
	}
	return false
}