N1 2019-12 — audit report

69 questions (問題1–13, the non-listening half). Generated from the OCR and correction audit trails.

OCR confidence

Source: retypeset-digital — quality 4/5: Clean retypeset digital document (not a true scan): crisp horizontal type throughout, a diagonal '成都帝京日语学校' watermark on every page, and no scan speckle or skew. Strong plus: a complete answer key (page 37) covering both the reading half and listening. Minuses: third-party retypeset rather than the official booklet, so furigana and original vertical layout are largely lost, and the reading questions are renumbered to end at 69 instead of the official 70 (the cover itself warns it is 'not the final/perfect version'). Mondai 12 long passage (p26) is the densest/smallest type but still fully legible.

Every section was transcribed twice independently (a full-page pass and a double-resolution half-page pass) and reconciled. Of 13 sections, 3 agreed on all content between the two passes (reconciled automatically, differing only in transcriber notes) and 10 had at least one content difference resolved by a third adjudication pass that re-read the page images. 63 leaf-level differences were examined in total.

Residual OCR risks (29):

• 問題1: no human/agent spot-check was performed for this section (A/B byte-agreement on content was treated as sufficient)
• 問題2: no human/agent spot-check was performed for this section (A/B byte-agreement on content was treated as sufficient)
• 問題3: The （16） marker's parenthesis width is read as full-width from the render; if intended as half-width ASCII parens it would instead be (16), but it visually matches full-width 全角 parens.
• 問題3: Question numbers are printed boxed on the page (contrary to the 'plain unboxed' source description in the brief); this does not affect the integer number fields but is noted in case it signals a different source variant.
• 問題4: no human/agent spot-check was performed for this section (A/B byte-agreement on content was treated as sufficient)
• 問題5: None material. Q30's printed 「承りします」 and the duplicated 毎日 are confirmed retype artifacts in the source, not OCR errors; transcribed verbatim per conventions.
• 問題5: Question numbers are printed in boxes in this retype (vs. the 'plain unboxed' expectation in the task brief), but this does not affect transcribed content.
• 問題6: The instruction marker is a printed underlined blank containing a ★; rendered as ＿★＿. The exact underline-length is not byte-preservable, so a 3-glyph ＿★＿ is used as the closest convention-consistent form.
• 問題6: （1*5） is an unusual school-added points annotation; kept verbatim per 'transcribe exactly as printed'. If a later normalization pass strips score annotations, this would need removal.
• 問題7: Full-width spacing after ？ in '好き'なの？どう好きなの？ (para3): B inserted full-width spaces; final omits them, treating the gap as ？-glyph kerning. Could not confirm a literal space character.
• 問題7: 好きなの? in paragraph 4 is printed with what looks like a half-width ?; normalized to full-width ？ per the punctuation convention. If the convention intends strict glyph fidelity here, this single mark could differ.
• 問題7: （注n） gloss lines show a wide tab gap before the term; final uses no space per the convention's canonical example, but the print does show alignment whitespace.
• 問題8: The 注 gloss parentheses/colon are rendered half-width '(注)…位置' as printed; this overrides both transcribers' full-width （注）. If downstream tooling normalizes Japanese parens, revisit.
• 問題8: Furigana on 真似 is printed glyph-by-glyph (ま over 真, ね over 似); rendered as whole-word ruby per convention rather than per-kanji.
• 問題8: Stray full-width/half-width spaces (before ゴール, before 「必要」, inside 「A42-C4」 について, 'FH エンジニア', and before the センター colon) are retype artifacts retained verbatim; widths inferred from print and may be approximate.
• 問題9: Passage(3) これ underline: the printed underline in the body visually appears to run under これに気付 to the line end, but Q56 quotes only これ, so the span was set to これ. If the source intended a longer span this could differ.
• 問題9: Question numbers are printed boxed (not the plain style noted for this source family); no text-field impact but noted in case it signals a different layout than expected.
• 問題10: 杭(くい) furigana is small and slightly faint in p22-b; reading is confident but the lowest-confidence of the restored furigana.
• 問題10: 注1 gloss colon read as full-width '：' (matching 注2/3/5/6/7); only 注4 uses a half-width ':'. Cross-checked but full-/half-width colon distinctions are inherently hard at this resolution.
• 問題10: 悩む was confirmed on the double-res half, but the source is a retyped document, so '悩む' could itself be the retypist's substitution for an intended 脳; transcribed exactly as printed regardless.
• 問題11: The 叱 furigana glyph is small and slightly faint in the renders; read as しか, which is the standard reading for 叱られる, so confidence is high.
• 問題11: The instruction line's trailing "(2*3)" is a school annotation deliberately omitted; if a downstream consumer expected it preserved, this is a judgment call (treated as branding per conventions).
• 問題12: Watermarks (成都帝京日语学校 / 日语轻松考) cross some passage lines; readings were taken from regions where the glyphs were clear, but a stray dakuten/handakuten under a watermark could in principle be missed.
• 問題12: さかのほれる is transcribed as printed (a probable source retype typo for さかのぼれる); kept verbatim per the exact-transcription convention.
• 問題12: The half-width ASCII '?' at the end of para2 is transcribed as printed; if a downstream consumer expects full-width ？ for Japanese punctuation, this is a deliberate as-printed choice.
• 問題12: Question numbers 64-67 are printed inside thin boxes on this retypeset; numbers are stored in the number field (not the text), consistent with both transcribers.
• 問題13: Title gap width (交 流): represented as one full-width space; the printed gap is ~one character but could be a single half-width space. Both A and B disagreed on width and the rendering is ambiguous.
• 問題13: Table cell role (<th> vs <td>): the upper grid's left labels are bold (clear headers); the schedule grid's left labels are not bold but were also given <th> as row headers for consistency. A reviewer preferring styling-faithful <td> could differ on the schedule grid.
• 問題13: The 趣旨 row body had no bullets and stays a single line; no risk, noted for completeness.

Source-text quirks the OCR transcribed verbatim (printed typos/recall artifacts; the ones judged to be retype errors were corrected in the layer below) — 17 noted at OCR time:

• 問題1: Question 4 has no trailing 。 in the printed stem (it ends at 確かめる); transcribed as printed.
• 問題2: Q13 option 3 is printed as 続々 and option 4 as 堂々 using the iteration mark 々, as printed.
• 問題5: Q30 stem repeats 毎日 oddly (「毎日を迎え、お忙しい毎日をお過ごしのことと…」), likely a retype error where a 〜の候/時節 phrase was dropped; transcribed as printed and confirmed against p07-b.png.
• 問題6: 問題6 instruction line prints the slot marker as an underlined blank with an embedded ★; transcribed as ＿★＿ to preserve the printed ★. The printed score annotation （1*5） follows the sentence and is kept as printed.
• 問題6: Q36 has ★ as the first of the four slots ( ★ _ _ _ ); Q37 has ★ as the second slot ( _ ★ _ _ ); Q38, Q39, Q40 have ★ as the third slot ( _ _ ★ _ ). All verified against the double-resolution half-page renders.
• 問題7: Choices 44-1 '生まれそうだが.' and 44-3 '生まれるからだが.' are printed with a trailing half-width period (visible in the double-resolution half-page); transcribed as printed (likely retype artifacts).
• 問題7: Cloze markers are rendered as 【41】…【45】 inline where the boxed numbers sit; passage paragraphs join across the markers as printed.
• 問題8: Passage (1): the print contains stray spaces retained as printed — '「A42-C4」 について', 'FH エンジニア', and 'お客様センター ：0120-333-4585' (space before the full-width colon). These also appear in question 46's stem ('「A42-C4」 の再修理について').
• 問題9: Q52 option 1 自分の「考え」が出来上がる uses kanji 出来 (confirmed p17). Q52 option 3 「考え」が存在して初めて、書き進められる。 ends with a full-width 。 unlike sibling options — transcribed as printed.
• 問題10: （注4）gloss uses a half-width colon '鷲掴みにする:ここでは…' and carries furigana わしづか, while the other 注 glosses use full-width '：'; transcribed as printed.
• 問題10: （注5）gloss reads 'フィードバッ：還元' (missing ク) although the passage body has フィードバック（注5）; likely a retype typo, transcribed as printed.
• 問題10: Q60 stem prints 'メデイア' (メ-デ-イ-ア, full-size イ) where the body and answer options use 'メディア'; likely a retype typo, transcribed as printed.
• 問題10: （注7）'〜に裏打ちされる' but the passage body has '裏打ちされない'; the gloss gives the dictionary form. Transcribed as printed.
• 問題12: Single passage; questions 64-67 as printed, matching the expected 4 questions.
• 問題12: Para 1 prints (中略) with half-width parentheses, inline at the end of the first paragraph (not on its own line); transcribed exactly as printed.
• 問題12: Para 1: the source prints さかのほれる (ほ without dakuten); both transcribers silently corrected it to さかのぼれる. Transcribed exactly as printed (さかのほれる) per the no-correction rule; likely a retype typo for さかのぼれる.
• 問題13: Title prints as 十条市国際交　流事業助成制度 with a roughly full-width gap between 交 and 流; a retype spacing artifact, transcribed as printed with a full-width space.

Corrections applied (11)

Evidence-based reconstructions of the original exam text. This section is the canonical correction log for the sitting.

• 問3 q14 question: 「<u>異例</u>のこと」 → 「<u>異例の</u>こと」
  • The choices are complete adnominal forms, including the correct 珍しい. Leaving の outside the tested span produces 珍しいのこと; the full target is 異例の.
  • evidence: structural: 問題3 choices replace the underlined span.
• 問3 q16 question: 「<u>ルーズ</u>な人」 → 「<u>ルーズな</u>人」
  • The choices are complete adnominal forms, including the correct だらしない. Leaving な outside the tested span produces だらしないな人; the full target is ルーズな.
  • evidence: structural: 問題3 choices replace the underlined span.
• 問10 q60 question: 「メデイア」 → 「メディア」
  • Q60 stem prints 'メデイア' with a full-size イ, but the passage body and all answer choices consistently use 'メディア' (small ィ). The full-size イ is a retype typo for the small kana.
  • evidence: Confirmed in output/N1_2019-12.json: the mondai10 passage uses メディア, メディア対応, メディアの要求 etc. throughout, and every answer choice for Q60/Q61 uses メディア; only the Q60 stem has 'メデイア'. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問10 q60 passage: 「（注5）フィードバッ：還元」 → 「（注5）フィードバック：還元」
  • The （注5）gloss headword is truncated to 'フィードバッ' but the passage body has 'フィードバック（注5）'; the gloss headword must match the body term, so the dropped ク is a retype omission.
  • evidence: Confirmed in output/N1_2019-12.json: body reads '知見のフィードバック（注5）をしなければ'; the gloss reads '（注5）フィードバッ：還元'. 'フィードバッ' is a non-word. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q64 passage: 「何万年も前までさかのほれる」 → 「何万年も前までさかのぼれる」
  • 'さかのほれる' (ほ without dakuten) is a non-word; the intended verb is 遡れる/さかのぼれる. A missing dakuten is a classic retype typo.
  • evidence: Confirmed exact substring in output/N1_2019-12.json. Google Books returns standard hits for 'さかのぼれるようだが' (with dakuten); 'さかのほれる' is a non-word. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問7 q44 answers[0]: 「生まれそうだが.」 → 「生まれそうだが」
  • This is a mid-sentence blank-filler option (そこに本来的なギャップが【44】、逆に…); a trailing ASCII full stop '.' is a retype/OCR artifact. Parallel options 'generated as 生まれるのはいいが' / '生まれるかどうかだが' carry no terminal punctuation.
  • evidence: Category added-char, confidence high. [retype-corrections-pass2 (propose + adversarial verify), 2026-06-20]
• 問7 q44 answers[2]: 「生まれるからだが.」 → 「生まれるからだが」
  • Same stray ASCII full stop '.' artifact on a mid-sentence blank-filler option; the correct-answer option for Q44 should not carry terminal punctuation, matching its parallel options.
  • evidence: Category added-char, confidence high. [retype-corrections-pass2 (propose + adversarial verify), 2026-06-20]
• 問7 q0 passage: 「（注3）な部」 → 「な（注3）部」
  • （注N）note marker repositioned to follow its glossed term (the retype placed the marker before the word).
  • evidence: Marker-only move; the marker-stripped text is byte-identical. [note-marker relocation pass, 2026-06-20]
• 問9 q0 passage: 「（注3）の断」 → 「の（注3）断」
  • （注N）note marker repositioned to follow its glossed term (the retype placed the marker before the word).
  • evidence: Marker-only move; the marker-stripped text is byte-identical. [note-marker relocation pass, 2026-06-20]
• 問10 q0 passage: 「（注3）の知覚の裏をかかれるよ」 → 「の知覚の裏をかかれる（注3）よ」
  • （注N）note marker repositioned to follow its glossed term (the retype placed the marker before the word).
  • evidence: Marker-only move; the marker-stripped text is byte-identical. [note-marker relocation pass, 2026-06-20]
• 問12 q0 passage: 「（注2）の生」 → 「の（注2）生」
  • （注N）note marker repositioned to follow its glossed term (the retype placed the marker before the word).
  • evidence: Marker-only move; the marker-stripped text is byte-identical. [note-marker relocation pass, 2026-06-20]