Passes de relecture

syntax_book v2 load: 5053 ayahs vs Shamela 86

export /tmp/v2b86_stage run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 rollback of 203 named ayahs: 203 ayahs

book 86 Gemini-challenge fixes: Gemini doubt, Opus wrong, second Opus agrees, validators, blind re-score

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 440 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 rollback of 4 named ayahs: 4 ayahs

fix4: 4:61 T38, 31:7 T49, 3:73 T33, 89:16 RSLT-ف

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 540 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 rollback of 6 named ayahs: 6 ayahs

cert edge direction in not_stated (6 ayahs)

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 4818 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 rollback of 318 named ayahs: 318 ayahs

cert edge direction: V -> CERT per tagset (318 ayahs); before reload

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 45 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 34 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 60 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 227 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 99 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 52 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 123 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 286 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 62 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 60 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 75 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 83 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 43 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 88 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 77 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 78 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 112 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 128 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 50 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 129 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 165 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 200 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 74 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 73 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 135 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 206 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 176 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 285 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 22 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 35 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 37 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 82 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 42 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 33 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 59 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 49 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 43 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_book v2 load: 6195 ayahs vs Shamela 86

export quran/review/treebank/v2-book86 run v2-book86-full

Cette passe ne stocke aucun constat ; elle a chargé des données.

syntax_* cells whose latest finding is agrees (not read by the Phase 4 UM class check): every cited book entry scanned for a rival alternative, flagged cells hand-read under the UM rule (downgrade-only)

Phase 4 rival sweep Step 5: agrees->uncertain downgrades from batches out001-074 (838 cells), verifier-confirmed (270 cells checked, 36 refutations, 1 revision round; 22 reverted to agrees, 14 newly confirmed). Rehearsed on a cp -c clone: dbdiff CHANGE SET MATCHES, gates 119/2, drift clean, sync 11/0, coverage 100%.

838 constats Concorde 0 Incertain 838

syntax_* claim cells without a finding after the syntax Phase 4 apply (not_stated sweep: hand readings, the TAQ-fixed Phase 3 comparer, consolidation decisions, and each book's stated silence)

Phase 4 not_stated sweep over the 17166 treebank cells with no finding after runs 16-25: {'1 hand reading': 16026, '4 not_stated (cross-reference only)': 1135, '3 TAQ-fixed comparer': 5}. Hand readings: every cell no comparer tied a statement to was read by hand in 22916, 86 and 150578 (opus readers over 98 batches + the 649 unaligned / untied / named-head cells of rounds 1-3), strict extent and elided-POS rules (a phrase end agrees only on a stated span), a targeted re-check of the batches read before those rules, a class-ruling consolidation pass, and a fresh adversarial verifier; TAQ-fixed comparer: compare_phase3.py with «وجوبا تقديره» / «تقديره: فهم» read (phase4/taq, 17 changed cells hand-checked); not_stated notes give each book's reason and cite the passage read; cross-reference-only cells quote the cross-reference (the morphology run 12 precedent). Disjoint from every earlier finding.

17,166 constats Concorde 8,677 Diffère 11 Incertain 2,734 Non mentionné 5,744

syntax_* and ayah_word_segments cells corrected by the syntax Phase 4 apply (post-fix state)

one finding per claim cell the Phase 4 apply edited or inserted (39 decisions, keyed by node, edge or word; data_corrections reason "Syntax Phase 4 book review"), citing the page its decision rests on: 22916 first in surahs 1-9, 86 first in 59-114, then 22928 / 150578; no overrides; 22916/86 pages checked against the page cache (vol/page printed on the page id, a quote piece verbatim). All agrees. Includes the morphology cells of 2:186:11 دعان (segment 122397 aspect / person / number)

80 constats Concorde 80

syntax_* cells corrected by the syntax Phase 3 apply (post-fix state)

one finding per treebank cell the Phase 3 apply edited or inserted (88 decisions, keyed by node or edge; data_corrections reason "Syntax Phase 3 book review"), citing the page its decision rests on: 22916 first in surahs 1-9, 86 first in 59-114, then 22928 / 150578; no overrides; 22916/86 pages checked against the page cache (vol/page printed on the page id, a quote piece verbatim). agrees, except 1 cell set by the treebank convention (decision edge 6602's VS phrase_last_node_id, the clause-extent practice): not_stated with the reason

175 constats Concorde 174 Non mentionné 1

syntax_nodes ELIDED part_of_speech_id/elided_text and PHRASE phrase_type/phrase_first_node_id/phrase_last_node_id vs Shamela 86 «إعراب القرآن الكريم» (Dār al-Ṣaḥāba) word analyses and clause statements (syntax Phase 3)

comparer syntax/phase3/compare_phase3.py (imports phase1/compare_relation.py alignment + phase2/compare_head.py named heads) + conventions_phase3.v1.tsv: ELIDED POS only from «تقديره X», «بفعل محذوف», an explicit كائن/استقرّ word, the named hidden pronoun, «حرف نداء محذوف» (S2 rule 6; «متعلّق بمحذوف خبر/نعت/صلة» = existence only, not_stated); elided_text = the pronoun the book names; PP from the preposition's object read «جار ومجرور» / «حرف جرّ» + «مجرور» / «(PP) متعلّق»; clause and SC phrases from the statements whose quote starts at the phrase's first written word (tightest phrase), type from the statement's qualifier (DIR) or derived from the clause head's category (S2 rule 7), last only when the quote reaches the last written word; S2 rules 1/2 (a DIR mismatch differs only under a row measured ok, else uncertain; the book's alternative = uncertain), 8, 9 (C-NASIKH, C-TASWIYA, C-SHAN, REPCONV uncertain, never fixes); S6 R1 (150578 giving the treebank value makes a differs uncertain); class CLASS-MUBTADA-TAQDIR held uncertain by the comparer (D15) until the consolidation decided it whole (the decisions file: 8 members differs, fixed by the Phase 3 apply; 42004/20944 uncertain; noun تقديرs agree); quotes verbatim at their page offset; 50 random agrees hand-checked 50/50 (seed 20260924c); differs / both-books-state-another-value / class / carried cells read by hand (1 decision file(s), 85 cells folded in); not_stated/not_found not recorded (D12; Phase 4 sweep)

29,244 constats Concorde 28,969 Diffère 15 Incertain 260

syntax_nodes ELIDED part_of_speech_id/elided_text and PHRASE phrase_type/phrase_first_node_id/phrase_last_node_id vs Shamela 22916 «الجدول في إعراب القرآن» «الإعراب» sections and clause statements (syntax Phase 3)

comparer syntax/phase3/compare_phase3.py (imports phase1/compare_relation.py alignment + phase2/compare_head.py named heads) + conventions_phase3.v1.tsv: ELIDED POS only from «تقديره X», «بفعل محذوف», an explicit كائن/استقرّ word, the named hidden pronoun, «حرف نداء محذوف» (S2 rule 6; «متعلّق بمحذوف خبر/نعت/صلة» = existence only, not_stated); elided_text = the pronoun the book names; PP from the preposition's object read «جار ومجرور» / «حرف جرّ» + «مجرور» / «(PP) متعلّق»; clause and SC phrases from the statements whose quote starts at the phrase's first written word (tightest phrase), type from the statement's qualifier (DIR) or derived from the clause head's category (S2 rule 7), last only when the quote reaches the last written word; S2 rules 1/2 (a DIR mismatch differs only under a row measured ok, else uncertain; the book's alternative = uncertain), 8, 9 (C-NASIKH, C-TASWIYA, C-SHAN, REPCONV uncertain, never fixes); S6 R1 (150578 giving the treebank value makes a differs uncertain); class CLASS-MUBTADA-TAQDIR held uncertain by the comparer (D15) until the consolidation decided it whole (the decisions file: 8 members differs, fixed by the Phase 3 apply; 42004/20944 uncertain; noun تقديرs agree); quotes verbatim at their page offset; 50 random agrees hand-checked 50/50 (seed 20260924c); differs / both-books-state-another-value / class / carried cells read by hand (1 decision file(s), 85 cells folded in); not_stated/not_found not recorded (D12; Phase 4 sweep)

29,077 constats Concorde 28,771 Diffère 13 Incertain 293

syntax_* cells corrected by the syntax Phase 2 apply (post-fix state)

one finding per treebank cell the Phase 2 apply edited or inserted (24 decisions, data_corrections reason "Syntax Phase 2 book review"), citing the page its decision rests on: 22916 first in surahs 1-9, 86 first in 59-114, then 22928 / 150578; no overrides; 22916/86 pages checked against the page cache (vol/page printed on the page id, a quote piece verbatim). all agrees

25 constats Concorde 25

syntax_edges.head_node_id vs Shamela 86 «إعراب القرآن الكريم» (Dār al-Ṣaḥāba) word analyses and clause statements (syntax Phase 2)

comparer syntax/phase2/compare_head.py (imports phase1/compare_relation.py alignment) + conventions_head.v1.tsv: NAMED heads (the word the book names after its role word, resolved in the ayah; a PHRASE head = the phrase containing it; an elided head by the named function) and POSITIONAL heads (S2 rule 3: only when this book's Phase 1 relation finding agrees; the unique slot grammar fixes); S2 rules 1/2 (a NAMED mismatch differs only under a row measured ok, else uncertain; the book's alternative = uncertain), 4 (conj chain), 5 (link never fixed on two books), 8 (short quotes tied to the entry); S6 R1 (150578 naming the treebank head makes a differs uncertain); representation rows H1/H3/H4/H5 uncertain, never fixes; quotes verbatim at their page offset; 50 random agrees hand-checked 50/50 (seed 20260924b); differs / both-books-name-another-head / carried cells read by hand (1 decision file(s), 1317 cells folded in); not_stated/not_found not recorded (D12; Phase 4 sweep)

35,734 constats Concorde 34,371 Diffère 24 Incertain 1,339

syntax_edges.head_node_id vs Shamela 22916 «الجدول في إعراب القرآن» «الإعراب» sections and clause statements (syntax Phase 2)

comparer syntax/phase2/compare_head.py (imports phase1/compare_relation.py alignment) + conventions_head.v1.tsv: NAMED heads (the word the book names after its role word, resolved in the ayah; a PHRASE head = the phrase containing it; an elided head by the named function) and POSITIONAL heads (S2 rule 3: only when this book's Phase 1 relation finding agrees; the unique slot grammar fixes); S2 rules 1/2 (a NAMED mismatch differs only under a row measured ok, else uncertain; the book's alternative = uncertain), 4 (conj chain), 5 (link never fixed on two books), 8 (short quotes tied to the entry); S6 R1 (150578 naming the treebank head makes a differs uncertain); representation rows H1/H3/H4/H5 uncertain, never fixes; quotes verbatim at their page offset; 50 random agrees hand-checked 50/50 (seed 20260924b); differs / both-books-name-another-head / carried cells read by hand (1 decision file(s), 1317 cells folded in); not_stated/not_found not recorded (D12; Phase 4 sweep)

31,775 constats Concorde 30,194 Diffère 24 Incertain 1,557

syntax_* cells corrected by the syntax Phase 1 apply (post-fix state)

one finding per treebank cell the Phase 1 apply edited or inserted (145 decisions, data_corrections reason "Syntax Phase 1 book review"), citing the page its decision rests on: 22916 first in surahs 1-9, 86 first in 59-114, then 22928 / 150578; overrides 43028->22928, 33173->86, 35511->86, 15497->150578; 22916/86 pages checked against the page cache (vol/page printed on the page id, a quote piece verbatim). agrees, except elided POS cells set by the treebank convention where the books ground existence only (S2 rule 6): not_stated with the reason

269 constats Concorde 259 Non mentionné 10

syntax_edges.relation vs Shamela 86 «إعراب القرآن الكريم» (Dār al-Ṣaḥāba) word analyses and clause statements (syntax Phase 1)

comparer syntax/phase1/compare_relation.py + conventions_syntax.v1.tsv (plan S2 rules 1-9: ok rows measured >=98%/n>=5 in this run, equal_only mismatches uncertain, book alternatives uncertain, structural conventions never fixes); quotes verbatim at their page offset; 50 random agrees hand-checked 50/50; differs / both-books-disagree cells read by hand (1 decision file(s), 1774 cells folded in); not_stated/not_found not recorded (D12; Phase 4 sweep)

39,555 constats Concorde 37,413 Diffère 279 Incertain 1,863

syntax_edges.relation vs Shamela 22916 «الجدول في إعراب القرآن» «الإعراب» sections and clause statements (syntax Phase 1)

comparer syntax/phase1/compare_relation.py + conventions_syntax.v1.tsv (plan S2 rules 1-9: ok rows measured >=98%/n>=5 in this run, equal_only mismatches uncertain, book alternatives uncertain, structural conventions never fixes); quotes verbatim at their page offset; 50 random agrees hand-checked 50/50; differs / both-books-disagree cells read by hand (1 decision file(s), 1774 cells folded in); not_stated/not_found not recorded (D12; Phase 4 sweep)

35,559 constats Concorde 32,944 Diffère 271 Incertain 2,344

ayah_word_segments cells corrected from the audit review list (follow-up Step 2)

each corrected cell cites the page its decision rests on; decisions + verifier: followups/step2/workflow_result.json

35 constats Concorde 35

ayah_word_segments after the row-level reanalysis of أربى 16:92, يعفون 2:237, يرجون 24:60 (follow-up Step 1)

new noun lemma 4845 and stems 12114/12115 via corrections.insert_row (ADR 0001); each cell cites the page its reanalysis rests on

8 constats Concorde 8

ayah_word_segments Ruling 31 verbs after correction (Phase 5b)

five verbs whose form the books give as II/IV, moved to the form-II/IV lemma rows the importer had created (plan D19)

10 constats Concorde 10

ayah_word_segments cells no source states — final sweep (Phase 5)

not_stated with a reason per class (not_stated_classes.json); source_book_id = the book read for the word when its entry was read (plan D17/D18)

130,220 constats Concorde 0 Non mentionné 130,220

ayah_word_segments cells corrected by the Phase 5 whole-word reanalysis (post-fix state)

each corrected cell cites the page its reanalysis was grounded on (plan D18)

42 constats Concorde 42

ayah_word_segments voice, particle lemmas and remaining gaps vs Shamela 86 (Phase 5)

new cells grounded from the Phase 1 alignment (voice from the named subject, particle lemmas, POS/pronoun/special-group gaps); mappings measured ≥98%; conflicts read by hand; 84 fixes verified (plan D18)

45,558 constats Concorde 45,361 Diffère 13 Incertain 184

ayah_word_segments voice, particle lemmas and remaining gaps vs Shamela 22916 al-Jadwal (Phase 5)

new cells grounded from the Phase 1 alignment (voice from the named subject, particle lemmas, POS/pronoun/special-group gaps); mappings measured ≥98%; conflicts read by hand; 84 fixes verified (plan D18)

23,265 constats Concorde 23,130 Diffère 3 Incertain 132

ayah_word_segments person/gender/number/is_indefinite vs Shamela 86 «إعراب القرآن الكريم» (hidden subject, named pronouns, جمع/مثنى) (Phase 4)

reuses the Phase 1 alignment; mappings <98% give uncertain; conflicts read by hand, 83 fixes verified (plan D16)

69,143 constats Concorde 66,536 Diffère 33 Incertain 2,574

ayah_word_segments person/gender/number/is_indefinite vs Shamela 22916 al-Jadwal «الإعراب» (hidden subject, named pronouns, جمع/مثنى) (Phase 4)

reuses the Phase 1 alignment; mappings <98% give uncertain; conflicts read by hand, 83 fixes verified (plan D16)

39,431 constats Concorde 38,488 Diffère 7 Incertain 936

ayah_word_segments verb_form and derivation vs ṣarf statements (al-Jadwal 22916 «الصرف», Jabal 21017, Lisān 1687, al-ʿUkbarī 22928, al-Durra 30271, al-Mujtabā 9617, al-Zuḥaylī 22915) (Phase 3)

per lemma: a verb's form belongs to its lemma; 729 lemmas without a mechanical statement and all conflicts read by hand; fixes/fills verified adversarially (plan D15); jabal-root-only matches give no finding

24,929 constats Concorde 24,673 Diffère 15 Incertain 241

ayah_word_segments root_id and lemma_id vs Shamela 21017 (Jabal) root entries (Phase 2)

root: the word is quoted/named in its root entry; lemma: only when the entry quotes the ayah or names the lemma form (plan D13/D14); 688 items read by hand, 7 fixes verified adversarially

96,116 constats Concorde 95,069 Diffère 20 Incertain 1,027

ayah_word_segments grammatical cells vs Shamela 86 «إعراب القرآن الكريم» (Dār al-Ṣaḥāba) (Phase 1)

comparer + conventions table (plan D10/D11); every differs read by hand, fixes adversarially verified; not_stated/not_found not recorded (D12)

160,585 constats Concorde 154,385 Diffère 83 Incertain 6,117

ayah_word_segments grammatical cells vs Shamela 22916 «الجدول في إعراب القرآن» «الإعراب» sections (Phase 1)

comparer + conventions table (plan D10/D11); every differs read by hand, fixes adversarially verified; not_stated/not_found not recorded (D12)

100,385 constats Concorde 93,603 Diffère 32 Incertain 6,750

ayah_word_segments (POS, root, verb form, derivation) vs Shamela 22916 «الجدول في إعراب القرآن» «الصرف» sections

4,457 ṣarf entries parsed and matched by script; every disagreement (193) read by hand with the raw corpus line and, for suspected errors, Shamela 86 / 30271 / 22915; 40 random agreements hand-checked. 13 cells corrected (data_corrections 1027–1039); أربى 16:92 fix rolled back (1040–1051) pending a noun lemma. Verdicts: agrees / differs / uncertain.

9,432 constats Concorde 9,282 Diffère 52 Incertain 98

ayah_irab.text_arabic vs Shamela 150578 «الإعراب المفصل لكتاب الله المرتل», all rows

Script alignment (whitespace + NFC, presentation conventions folded) plus every differing row read by hand; word spans checked against the book's ayah sections; 62 corrections applied the same day as user_review (verdicts delegated by the user).

30,486 constats Concorde 30,021 Diffère 465