N1 2013-07 — audit report

70 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: Cleanly retypeset digital text at uniform ~1782x2400 resolution, no scan artifacts. Docked one point because the vocab/grammar pages 3-11 use a very small, dense font that requires zooming to read comfortably, and every page carries publisher watermarks (header text plus a script-style margin watermark), though they never overlap question text.

Every section was transcribed twice independently (a full-page pass and a double-resolution half-page pass) and reconciled. Of 13 sections, 2 agreed on all content between the two passes (reconciled automatically, differing only in transcriber notes) and 11 had at least one content difference resolved by a third adjudication pass that re-read the page images. 113 leaf-level differences were examined in total.

Residual OCR risks (25):

• 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: Source is a Chinese retypeset; no furigana and no obvious typos found in this section, but printed wording was transcribed as-is without correction.
• 問題3: Underline-scope judgement on Q14 (従来の vs 従来) is based on a continuous rule under proportional kana; rendered high-res crop is clear but the boundary at の is the only subtle call.
• 問題4: No image-processing tools (ImageMagick/PIL) were available for pixel-level zoom; bracket adjudication and spot-checks relied on the provided full-page and double-resolution half-page renders, which were legible at the needed level.
• 問題4: Headword for Q22 is printed in hiragana かばう; underline on options 2 and 4 covers only the inflected stem かばっ, consistent with the print — verified but worth noting as an unusual underline span.
• 問題5: Q30 options 1 (取れるって) and 4 (取っとけなくて) read oddly in context; transcribed exactly as printed in this Chinese retypeset, flagged as possible retype artifacts but the glyphs are crisp and unambiguous.
• 問題5: All differences resolved to B; A appears to have systematically dropped/misread a few characters (どうぞ/どうか, る in 取れるって, しようと in Q34, が好き in Q35). No remaining ambiguity in the imaged text.
• 問題6: The instruction-line blank glyph (＿＿＿★＿＿) is a borderline case: conventions specify ＿＿＿ (three full-width low lines) for instruction blanks but do not explicitly cover an embedded ★ in the instruction. Chose full-width low lines preserving the star as printed; exact count of low lines on either side of the star is not separately significant.
• 問題6: Source is a Chinese retypeset; no furigana present and no obvious typos found in this section, so nothing was flagged as suspicious.
• 問題7: Option 43-1 「朝だとはお思いのはずだ」 is grammatically odd but is printed exactly as transcribed (verified at high zoom); flagged only as a possible source-retype quirk, not an OCR error.
• 問題7: ひらと (Q44 sentence) is printed as ひらと (single occurrence), not the more idiomatic ひらひら; verified against the original render. Transcribed as printed.
• 問題8: The 2-em dash after 裏人生 renders as two touching horizontal strokes; I read it as two U+2015 per the long-2-em-dash convention. If the source intended a single wide dash glyph this is a one-character judgment call, but it is visually two strokes.
• 問題8: No furigana, tables, or 問題6/7 slot patterns occur in this section, so those conventions were not exercised here.
• 問題9: Underline extent in question stems (e.g. exactly which characters of ②「時代の流れ」 carry the rule) is taken from A and is plausible but the thin underline is near the resolution limit; A and the passage are internally consistent so risk is low.
• 問題9: The 2-em dash in passage (2)’s citation is rendered ―― (two U+2015) per convention; the printed glyph is a single long dash and could alternatively be one em-dash, but convention mandates the two-U+2015 form.
• 問題10: Source is a dense Chinese retypeset with no furigana; tiny glyphs like 簗 (青樹簗一) and 撲滅/種属 read at half-page resolution carry small residual uncertainty.
• 問題10: Q60 options 2 and 3 are genuinely identical (化学薬品の有用性) in print — preserved as a likely retypeset duplication rather than corrected.
• 問題11: 'どの店調も' in passage B contains the kanji 調 (言 radical) which is almost certainly a retypeset typo for 店舗; it is transcribed as printed per convention, and could trip downstream consumers expecting valid Japanese.
• 問題11: Passages were resolved from the double-resolution half renders; a handful of kana okurigana (e.g. 楽々/邪魔) are confident but the original source is a third-party retype, so latent typos elsewhere cannot be fully excluded.
• 問題12: The retypeset is itself flawed; the passage diverges from the published Yamada Taichi essay in wording, but it is transcribed as printed and the printed glyphs were read at high resolution.
• 問題12: Para-2 opening 「親とのつながりを」 uses brackets around 対話 elsewhere that render very close to ［ ］ vs 「 」; standard 「」 used for consistency with the rest of the passage.
• 問題13: 料金2行の列揃えは原文では複数の全角スペースで実現されているが、列揃えをスペース列で再現するのは非決定的なため、先頭インデントを除去し場合と金額の間のスペースのみBの形で残した。厳密なスペース数は採点に無関係と判断。
• 問題13: 全角カンマ（9，450）と全角ハイフン（01－1234）の判別は倍解像度でもグリフが小さく、全角/半角の最終判定はレンダラ依存の見た目に基づく。規約に従い数字・Latinは半角化、日本語文脈の読点様カンマは全角保持とした。
• 問題13: タイトル行のボールド判定はA・B一致のため再検証していないが、中央寄せの見出しとして両者ボールドを採用。

Source-text quirks the OCR transcribed verbatim (printed typos/recall artifacts; the ones judged to be retype errors were corrected in the layer below) — 22 noted at OCR time:

• 問題3: Q14 underline covers 従来の including the trailing の (verified on p02-b zoom); it ends right before 経営.
• 問題4: Q22 option 4: the quoted speech is printed with full-width square brackets ［…］ rather than 「…」; transcribed as printed.
• 問題4: For 問題22 (かばう), options 2 and 4 inflect the headword as かばっ(ている/てくれた); the underline covers かばっ as printed.
• 問題5: Q30 option 1 prints 「取れるって」 and option 4 「取っとけなくて」; transcribed as printed though they read oddly in context (likely intentional distractors or possible retype artifacts).
• 問題6: Instruction line on the page prints the blank as full-width low lines with an embedded star: ＿＿＿★＿＿. Rendered with full-width low lines (instruction-line-blank glyph normalization), preserving the star as printed.
• 問題6: In-sentence blank runs all use the four-slot pattern with the star in the third position ( _ _ ★ _ ) as printed; no question deviated from that layout.
• 問題7: Option 43-1 「朝だとはお思いのはずだ」 reads awkwardly (mixing 〜のはずだ with the honorific お思い); transcribed as printed, possibly a source retype quirk.
• 問題8: Passage (1) citation prints the magazine title as 「AERA MooK」 with mixed case "MooK" and corner brackets; transcribed as printed.
• 問題9: Passage (3) uses angle brackets 〈　〉 around 私 and full-width square brackets ［同類］ as printed.
• 問題9: The retypeset text on pp.6-8 is dense but cleanly legible at double resolution; the printed Japanese in all three passages and all options reads as coherent natural text. Transcriber B reported many spans as 'garbled/low-confidence', but those were OCR failures on B's side, not source corruption — every disputed leaf was resolved to A's coherent reading against the images.
• 問題10: 問題10 has a single passage with no （1）（2） marker; questions 59-62 as printed.
• 問題10: The passage begins at the bottom of p08-b and continues onto p09-t; it is a translation of Rachel Carson の 沈黙の春. Transcriber A (full-page renders) read it accurately; transcriber B (half-page renders) produced a heavily garbled passage and several wrong answer options, all of which were rejected against the images.
• 問題10: Q60 options 2 and 3 are printed identically as 化学薬品の有用性; transcribed as printed (apparent retypeset duplication, not corrected).
• 問題11: Source is a dense Chinese retypeset (日语轻松考). The two boxed passages A and B were re-read against the double-resolution half-page renders (p09-b for A, p10-t for B), which made nearly all characters legible; both transcribers' passage bodies contained many independent reading errors that were corrected here.
• 問題11: Passage B has a genuine retypeset typo: 'どの店調も' uses 調 (with the 言 radical) where context wants 店舗. The image unambiguously shows 調, so it is transcribed as printed.
• 問題11: Q63 option 4 reads '客の要望を品ぞろえに反映していること' (verified at high zoom); neither transcriber had this wording.
• 問題12: The source is a dense Chinese retypeset (日语轻松考) of the Yamada Taichi essay and the passage body contains several apparent garblings/typos relative to the original; transcribed as printed without correction.
• 問題12: Passage paragraph 2 opens 「といって、親とのつながりを、まったくのぞんでいなかったわけではない。」 (verified against p10-b high-res); both transcribers misread this opening clause.
• 問題12: Q66 stem and the underlined phrase in the passage body both read 「気持は安定していた」 (は); verified against p10-b and p11-t.
• 問題12: Passage uses 「分らない」 (not 「分からない」) in paragraph 5; transcribed as printed.
• 問題12: Citation reads 『誰かへの手紙のように』 (への, not 〜の); verified against p10-b high-res.
• 問題13: Money amounts are printed with a full-width comma as the thousands separator (9，450円 / 8，925円); the full-width comma is kept as printed since it is Japanese-context punctuation, not a Latin/digit normalization target. The leading column-alignment indents before the two money lines are dropped per the leading-indent convention.

Corrections applied (4)

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

• 問5 q26 question: 「竹山鉄道」 → 「位山鉄道」
  • Fabricated railway name in a 問題5 grammar stem; answer unaffected. Our retype/OCR reads 竹山 but the jlpt-freq 真题+答案 full retype reads 位山鉄道 cleanly ("…続いていた位山鉄道が、今年の3月31日（ ）廃止された"). No witness supports 竹山; 位 is a plausible scan misread as 竹.
  • evidence: input/N1_2013-07/5Mcv4He-jlpt-freq-exam-and-answers.txt
• 問11 q63 passageB: 「どの店調も広々としていて」 → 「どの店舗も広々としていて」
  • 店調 is a non-word OCR/retype corruption of 店舗 (visually similar left radicals). The same passage uses 店舗 twice in the following sentence (「ある店舗が混雑するようになると、近く新店舗を作ってしまう」), and 「どの店舗も広々としていて」 is the only coherent reading. This is a reading-comparison passage (not a 問題4 usage distractor), so it reflects genuine exam content that the official exam printed as 店舗.
  • evidence: from string exact-matches passageB in output/N1_2013-07.json; same passage contains 店舗 twice; 調 forms no word with 店 [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問10 q59 passage: 「複雑な形の自然石かつみかさなってできた石垣から」 → 「複雑な形の自然石が積み重なってできた石垣から」
  • 「自然石かつみかさなって」is not a possible reading: the intended phrase is 「自然石が積み重なって」(natural stones piled up). The retype voiced/garbled the subject particle (が→か) and rendered 積み重なって in bare kana; が is the certain correction, with 積み重なって the natural orthography for this kanji-using passage.
  • evidence: Category typo, confidence medium. [retype-corrections-pass2 (propose + adversarial verify), 2026-06-20]
• 問10 q0 passage: 「（注1）に、」 → 「に（注1）、」
  • （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]