N1 2012-07 — 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: Clean retypeset digital text with no scan artifacts; watermarks are confined to page headers and never overlap content. Docked one point because pages 3 and 5–11 are rendered at much lower resolution than the rest, making the dense reading passages small, and third-party ads/watermarks (日语轻松考, Yuuki Bùi) are present throughout.

Every section was transcribed twice independently (a full-page pass and a double-resolution half-page pass) and reconciled. Of 13 sections, 5 agreed on all content between the two passes (reconciled automatically, differing only in transcriber notes) and 8 had at least one content difference resolved by a third adjudication pass that re-read the page images. 119 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: no human/agent spot-check was performed for this section (A/B byte-agreement on content was treated as sufficient)
• 問題4: Question 24 option 2 (3人分満たない, no に) and question 21 option 3 (頭がブランクだったよ) are faithful to this retypeset but may differ from the original JLPT booklet; cross-check against another source if one becomes available.
• 問題4: The source is a third-party retypeset, so underline extents (e.g. 怠ら vs 怠らず in question 22) reflect the retypesetter's choices, not necessarily the original exam's.
• 問題5: The Q34 half-width '?' resolution rests on interpreting the conventions' punctuation clause as non-normalizing; if the project intends ？ to always be full-width, $.data[8].question and the related note must be flipped to B's value.
• 問題5: Q33 answers 1-2 are transcribed with the printed doubled とする (引き出したとするとして…); if this section is later cross-checked against another source of the original exam, expect a mismatch there (original likely 引き出したとして／引き出したのに対して).
• 問題5: Whitespace-only layout artifacts (wider gaps after the blank's ） in Q31/33/35) were normalized away per convention; the exact printed spacing is not recoverable from the transcription.
• 問題6: no human/agent spot-check was performed for this section (A/B byte-agreement on content was treated as sufficient)
• 問題7: The phrase 「語るには情をもってし」 looks like a possible retypeset corruption of the original essay; transcribed as printed but unverifiable against the original 1992 newspaper source.
• 問題7: The exact width of the stray space in the instruction line (1、2、 3、4) is a judgment call; transcribed as one half-width space, consistent with the convention of recording printed spacing as-is.
• 問題8: no human/agent spot-check was performed for this section (A/B byte-agreement on content was treated as sufficient)
• 問題9: Underline extents on the ① / ② referenced spans are taken from the full-page p07.png (underlines are visible there) and applied per convention; the exact start/end character of each <u> span on the dense page is moderately confident, not pixel-certain.
• 問題9: The formula glyph '=?' in passage(1) and the Q50 stem: read as ASCII '=' followed by '?'. The '=' is small and could in principle be a long dash or 'ニ'-like artifact, but it most consistently reads as '=?'.
• 問題9: Passage(1) '今やさぞかし①「ワシも族」がはびこっていることだろうと思いきや' is small dense print; high-confidence but not pixel-certain on every kana.
• 問題10: Colon glyph in the 注 glosses: rendered narrow in the dense retypeset; read as half-width ':' per A's note, but full-width '：' cannot be 100% excluded at available resolution.
• 問題10: The passage is a flawed Chinese retypeset; some phrasings (e.g. 喜びを支援する) are awkward but are transcribed as printed and confirmed against the high-res halves.
• 問題11: The retypeset uses a half-width colon in 注1/注3 but a full-width colon in 注2; this inconsistency is as printed (faithfully transcribed) rather than an OCR artifact.
• 問題11: correctAnswerIndex left null for both questions per convention; not adjudicated here.
• 問題12: Citation bracket type around the title 家 is hard to read at this resolution; kept 「家」(article/column title) over 『家』 on convention and legibility, but cannot be 100% certain.
• 問題12: （中略）shares a printed line with 動物は序列に敏感なのだ。 (large trailing gap); treated as its own <p>. If a reviewer reads the gap as mere intra-line spacing, it could instead belong to the same <p>.
• 問題12: Faint kanji in para 9 (髯をはやし, 頻発する) are legible but small; transcribed as printed.
• 問題13: Field-separator spacing on the bullet list lines (time / 回数 / 料金) is wider than one space in the print due to column alignment; normalized to a single full-width space for determinism, so exact printed gap widths are not preserved.
• 問題13: The （平日9:00〜17:00 ）line has a stray space before the closing paren in the print; kept as printed. The parens around bracketed Japanese segments are full-width per convention even though the retypeset renders them narrow (could be ASCII in the source).
• 問題13: Whether the original (non-retypeset) exam placed the entire course box on a single facing page is unknown; transcription reflects the retypeset's two-page split, merged into one logical passage.

Source-text quirks the OCR transcribed verbatim (printed typos/recall artifacts; the ones judged to be retype errors were corrected in the layer below) — 27 noted at OCR time:

• 問題1: The printed instruction line has a stray half-width space between 選び and なさい (「一つ選び なさい。」), likely a retypeset artifact; transcribed as printed.
• 問題1: Question 5 option 1 is printed with a full-size よ (とうしよう), unlike option 3's small ょ (としょう); suspected retypeset typo for とうしょう, transcribed as printed.
• 問題1: The printed 問題1 instruction line has a visible space between 選び and なさい (一つ選び なさい。), almost certainly a retypeset artifact; transcribed with a half-width space as printed.
• 問題1: Question 5 option 1 is printed with a full-size よ (とうしよう), verified at high zoom against the clearly small ょ in option 3 (としょう); likely a retypeset typo for とうしょう, transcribed as printed.
• 問題2: Question 11 prints 「ソ」と「リ」; the official exam is commonly cited as 「ソ」と「ン」, so 「リ」 may be a retypeset typo. Transcribed as printed.
• 問題2: Question 11: the print clearly reads 「ソ」と「リ」 (verified at high zoom: the second katakana is リ with two vertical strokes). This is a suspected retypeset typo for 「ン」, since ソ/ン is the classic confusable pair and option まぎらわしい fits that pairing; transcribed as printed.
• 問題4: Question 22 option 3: the underline covers only 怠ら, with ずに outside the underline, as printed (verified against the double-resolution image during adjudication).
• 問題4: Question 24 option 2 reads 「会議資料が3人分満たないから」 as printed (no に before 満たない); plausibly the intended wrong-usage sentence rather than a retype error, transcribed as printed (verified against the image).
• 問題4: Question 21 option 3 reads 頭がブランクだったよ in this retypeset; verified against the double-resolution image, transcribed as printed.
• 問題5: The printed 問題5 instruction has a stray space between 選び and なさい（…から一つ選び なさい。）; transcribed as printed with a half-width space — likely a retypeset typo.
• 問題5: Question 34: the question mark in B「どんな話?」 is printed half-width; transcribed as printed (adjudicated against the double-resolution image: the glyph occupies half a character cell, immediately followed by 」).
• 問題5: Question 33: answers 1 and 2 print the suspicious doubled pattern 引き出したとするとして／引き出したとするのに対して (the genuine exam likely had 引き出したとして／引き出したのに対して); verified character-by-character against the double-resolution image and transcribed as printed.
• 問題7: Author name is printed as 小林善彦; confirmed against the double-resolution render (the second kanji is 善, not 源). Transcribed as printed.
• 問題7: The 注 gloss is printed with a half-width colon and no space after the label: 「（注）傾聴:真剣に聞くこと。」 — transcribed as printed (the colon is the narrow half-width ':', confirmed on the high-res render).
• 問題7: Instruction line has a stray extra space printed between 「2、」 and 「3、」 (1、2、 3、4); the 3 starts a gap late, so it is transcribed as a single half-width space, as printed.
• 問題7: 「【44】わが国では、語るには情をもってし、つねに控え目で…」 — the phrase 語るには情をもってし reads awkwardly and may be a retypeset corruption of the original essay text; transcribed as printed.
• 問題8: Passage (4) citation prints the year and date in full-width kanji numerals (二〇一〇年三月八日付); kept as printed since they are not Latin/ASCII digits.
• 問題8: In passage (1), the source prints キノコ in katakana once and 肉の味のしみた野菜 — transcribed as printed.
• 問題8: Passage (4): printed without commas in 「遭遇しそれを」 and 「都市を築き文明を」 where commas might be expected; transcribed as printed (possible retypeset omission).
• 問題9: Q50 stem reads in full '筆者の友人が、＜「自分」―「仕事」=?＞という設問に「ゼロや」と答えたそうだが、それはどのような意味か。' — verified against p07-t. (Transcriber A dropped the 'という設問に「ゼロや」' clause; transcriber B kept it but mis-read '=?' as '？'.)
• 問題9: Passage (1) repeats the noun-form '想像・想像する' for the verb form in passage (3) too; both '空想・空想する' and '想像・想像する' duplicate noun/verb in the source. Transcribed verbatim as a retypeset quirk.
• 問題9: Question 50 option 2 reads 'これからも家族と一緒に' in the source; transcribed as printed.
• 問題9: Adjudication: the double-resolution half-page renders (p06-b, p07-t, p07-b, p08-t, p08-b) are fully legible. Transcriber A's readings matched the page throughout; transcriber B's passage prose and option text were largely garbled best-guesses and were rejected except where B independently confirmed A. The B file's worry that the Q58 stem was 'inferred' is unfounded: the stem is clearly printed on p08-b.
• 問題10: Transcriber B's passage and most options are heavily garbled/hallucinated (e.g. '理論や目的', '三木亮逸', '③肌に届かない', '姿勢すずに'); the printed page matches A almost everywhere, so A was taken as the baseline and B rejected wherever it diverged.
• 問題10: The 注 glosses use a narrow half-width colon (コンペ:ここでは…) rather than the full-width ：; transcribed as printed with half-width :.
• 問題11: Inconsistent colon glyphs in the 注 glosses as printed in this retypeset: 注1 and 注3 use a half-width colon (保全:ここでは / 保全生物学:生態学), while 注2 uses a full-width colon with surrounding spacing (生息する：生きる). Transcribed faithfully as printed.
• 問題12: Transcriber B worked from the half-page renders and produced an almost entirely garbled passage and questions (e.g. 'うちの夫は訓練に仕込まれた犬である', '①おじいちゃんありますか'); A's reading is coherent and matches the high-resolution p10-b/p11-t/p11-b renders, so every content diff was resolved to A.

Corrections applied (7)

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

• 問1 q5 answers[0]: 「とうしよう」 → 「とうしょう」
  • Mondai 1 (kanji reading) option 1 is printed with a full-size よ (とうしよう), a non-word. The answer-key analysis in two independent witnesses lists option 1 as '1.とうしょう 凍傷', a real word (frostbite) with small ょ, paralleling option 3's としょう (渡渉). The full-size よ is an OCR/typeset artifact; the same witness line also garbles option 2 as とうしゅ2う, confirming the exam-transcription line is noisy.
  • evidence: input/N1_2012-07/5Mcv4He-jlpt-freq-exam-and-answers.txt line 650 ('1.とうしょう 凍傷') vs noisy line 19 ('1 とうしよう ... 2 とうしゅ2う'); independently confirmed by input/N1_2012-07/5Mcv4He-jlpt-freq-exam-and-answers.txt lines 19 and 650. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問2 q11 question: 「カタカナの「ソ」と「リ」は（　　）ので」 → 「カタカナの「ソ」と「ン」は（　　）ので」
  • ソ/ン is the classic confusable katakana pair, and the correct answer まぎらわしい (easily confused) fits ソ/ン, not ソ/リ which are not visually confusable. The retype miswrote ン as リ.
  • evidence: Category typo, confidence high. [retype-corrections-pass2 (propose + adversarial verify), 2026-06-20]
• 問5 q33 answers[0]: 「引き出したとするとして」 → 「引き出したとして」
  • Doubled とする is a retype artifact; options 3/4 parallel with 引き出そうとしたとして/のに対して, so option 1 should be 引き出したとして to match the parallel structure.
  • evidence: Category typo, confidence high. [retype-corrections-pass2 (propose + adversarial verify), 2026-06-20]
• 問5 q33 answers[1]: 「引き出したとするのに対して」 → 「引き出したのに対して」
  • Doubled とする is a retype artifact; parallels option 4 引き出そうとしたのに対して, so option 2 should be 引き出したのに対して.
  • evidence: Category typo, confidence high. [retype-corrections-pass2 (propose + adversarial verify), 2026-06-20]
• 問10 q59 passage: 「それでは（注3）無味乾燥の教科書文章となり」 → 「それでは無味乾燥（注3）の教科書文章となり」
  • （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6 q36 question/correctOrder: 「 _ _ ★ _ 良心は残っているはずだ。; correctOrder null」 → 「 _ ★ _ _ 良心は残っているはずだ。; correctOrder [3,1,2,0]」
  • The assembled order is どんな悪人であろうとどこかに良心は残っているはずだ (4231), and the answer key gives q36 answer 2. Those can both be true only if ★ is the second slot, not the third.
  • evidence: input/N1_2012-07/5Mcv4He-jlpt-freq-exam-and-answers.txt line 810: q36 正解:2 and 正秦捧席/排序 4231. [order-conflict cleanup, 2026-06-24]
• 問6 q39 correctOrder: 「null」 → 「[0,3,2,1]」
  • The answer explanation gives the accepted order デザートがただっていうだけじゃないか (1432). With ★ in the third slot, that confirms answer 3 and resolves the stale solver-order conflict.
  • evidence: input/N1_2012-07/5Mcv4He-jlpt-freq-exam-and-answers.txt lines 822-827: 正秦捧序/排序 1432 and assembled sentence デザートがただっていうだけじゃないか. [order-conflict cleanup, 2026-06-24]