N1 2020-12 — audit report

68 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: Retypeset digital text (Chengdu Teikyo school recall edition), so glyphs are perfectly crisp at 1407x1990 with no scan noise; a light gray 帝京日語 watermark and header/footer ads sit on every page without harming legibility. Docked from 5 for recall-grade fidelity: cover itself warns it is not the final version; typos include 言葉知識 for 言語知識 on the cover, misordered option numbers in listening 問題2 7番, a duplicated 質問1 label and mislabeled 2番 in 問題5, and a stray editorial phrase fused into the 問題9(2) passage; two written questions and at least two listening items are missing; no furigana. The embedded text layer is reliable only for headers/answer key — body text extracts as mojibake (broken ToUnicode), so OCR from the page images is required.

Every section was transcribed twice independently (a full-page pass and a double-resolution half-page pass) and reconciled. Of 13 sections, 1 agreed on all content between the two passes (reconciled automatically, differing only in transcriber notes) and 12 had at least one content difference resolved by a third adjudication pass that re-read the page images. 70 leaf-level differences were examined in total.

Residual OCR risks (32):

• 問題1: Options 3/4 of q1 (がんぽ/かんぽ) are nonsense distractors, so disambiguation rests solely on the glyph shape; handakuten vs dakuten was confirmed under high zoom but small-mark OCR is inherently the fragile spot here.
• 問題1: Instruction-line blank rendered as ＿＿＿ per convention; the printed underscore run width is not independently verifiable but is conventionally normalized.
• 問題2: Q12 reads grammatically incomplete (開催が（　　）いたが); this is a defect in the Chinese prep-school retype, preserved verbatim per conventions, but a downstream consumer may flag it.
• 問題2: The blank widths printed in the source vary; all normalized to （　　） per convention, so the original printed blank size is not recoverable from the final.
• 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: q21 answer 1 reads '藤田さんはいつもの外国語を円滑に話すそうだ' (いつもの rather than いつも) — verified as printed in both transcribers and the render; flagged only because the wording is slightly unnatural, but it is what the source prints (likely a retype quirk).
• 問題4: Watermark/branding overlaps several lines on p07-b.png; underline endpoints were judged on the cleanest available render but faint half-tick artifacts near watermarks cannot be fully excluded.
• 問題4: Answer keys/correctAnswerIndex intentionally left null per task; not adjudicated here.
• 問題5: Q33 typo パンプレット is faithful to the render but is a likely source retype error; flagged in notes rather than corrected.
• 問題5: Q29 headword 山下春好 reading is uncertain (no furigana printed); transcribed exactly as printed kanji.
• 問題5: Question numbers are boxed in this render, contrary to the source's usual plain-number style; this does not affect transcribed content but is worth noting for the section-level assumptions.
• 問題6: Instruction-line ★ spacing (half- vs full-width) is not explicitly covered by the conventions; chose half-width for determinism/consistency with the slot-pattern rule, but the printed box padding could be read either way.
• 問題6: Slot-pattern star position read from low-resolution underline runs; all five questions read as ★ in the 3rd of 4 slots, consistent between both transcribers.
• 問題7: Paragraph 5 has several uncommon wordings (庭先に生動する, 風ともつかぬ, 鶯や雀の声, 仔細を美しい言葉にのせて綴った) that read as retypeset oddities; transcribed exactly as printed but could be source typos.
• 問題7: Dash runs in paras 2/5/6 are rendered as four U+2015 (――――) matching the printed glyph width; if the house style intends two 2-em dashes this would need normalization, but printed form was preserved.
• 問題7: 七、八ページ uses kanji numerals (not Arabic), so the half-width-digit normalization does not apply; verified kanji in the image.
• 問題8: Question numbers are printed inside boxes in this scan (contrary to the 'plain/unboxed' expectation), but this does not affect the text fields (numbers live in the 'number' field).
• 問題8: The full-width-space alignment after （注n） is a layout artifact; transcribed as printed (one full-width space) per 'exactly as printed', but it could be interpreted as mere alignment whitespace.
• 問題9: The source is a low-fidelity Chinese prep-school recall re-typeset; passage bodies almost certainly diverge from the genuine 2020-12 exam wording. The final reproduces what is printed, not the original essays.
• 問題9: Passage (2) para 1's restored stray phrase 筆者が言いたいことは何か。 is a source defect; if a downstream consumer expects clean prose this will look wrong, but it is faithful to the page.
• 問題9: Fine furigana/period-level marks are at the resolution limit of the half-page crops; the missing 。 after 煮つめられた模範ではない and the 煮つめ/煮詰 split were the only sub-character ambiguities and both were resolved against the highest-zoom crops.
• 問題10: The passage is a Chinese prep-school retype with multiple awkward wordings (e.g. 'その速度はほとんど速くなっている') that are transcribed as printed; these are faithful to this source but may diverge from the original published essay.
• 問題10: No source-citation line is printed for this passage in the leaked document; if the genuine exam included one, it is absent here.
• 問題10: The figure '二万七千種' in paragraph 2 was legible at double resolution and matches A, but the original essay's number is not independently verifiable from this document alone.
• 問題11: The watermark over passage B partially overlaps the text region; however the relevant line 「取り残されてしまう」 is unobstructed and legible in p18-b.png, so confidence is high.
• 問題11: Q62 option 4's 「持っこと」 small っ is a genuine source typo retained per the transcribe-as-printed convention; downstream consumers may want to flag it.
• 問題12: メディアカ vs メディア力: the cram-school answer key for this essay (源 タイトル likback to 池田晶子-style 'メディア力') normally uses kanji 力, so the katakana カ is almost certainly a Chinese-retype glyph error in the source; transcribed exactly as printed per conventions, but downstream consumers may want to flag it.
• 問題12: Q66 answer 1 '伝われる' and body '1％異常' / 'わかかった' are likely source retype typos; preserved verbatim.
• 問題12: Passage adjudicated almost entirely to B because A's full-page render was too degraded to compare leaf-by-leaf; remaining sub-character uncertainty is low given the clean digital text in p19-t/p19-b.
• 問題13: Circle-marker code point: image cannot reliably distinguish ○ (U+25CB) from 〇 (U+3007); standardized to ○ (U+25CB) per convention. If downstream expects U+3007, the legend and table circles would need a sweep.
• 問題13: Phone-hours tilde: both table occurrences render as the same short tilde; normalized to 〜 (U+301C) per convention rather than ASCII ~; the exact intended code point is not certain from the render.
• 問題13: Bullet indentation: the gap after ● was rendered as a full-width space (●　); it could be Word list indentation rather than a literal space character.

Source-text quirks the OCR transcribed verbatim (printed typos/recall artifacts; the ones judged to be retype errors were corrected in the layer below) — 29 noted at OCR time:

• 問題2: Q12: printed text reads "開催が（　　）いたが" — likely a retype dropoff of "〜て" before "いた" (e.g. あやぶまれていたが); transcribed exactly as printed.
• 問題4: Q24 answer 2 uses transitive 収容する with intransitive-style 例文が subject marking ('例文がたくさん収容することができます'), which reads oddly; transcribed exactly as printed.
• 問題5: Q29 stem begins with anonymized company name printed as 「X自動車」 (Latin capital X normalized to half-width). Headword 山下春好 (はるよし?) printed without furigana; transcribed as printed.
• 問題5: Q33 scene label prints 「パンプレット」 (missing フ; standard spelling パンフレット) and the body reads をおしらせ in hiragana — both transcribed as printed.
• 問題7: Passages 3-6 were badly garbled in both retyped OCR drafts; re-read from the double-resolution half-page renders. Correct readings: 3 ends 「主人公として描いた小説や映画の中に出てくるそいつの方が、ずっと、…ということがある。」; 4 has 「小品」「荷風がたった七、八ページの文章の中で描写した」「実感したことがないということに気づいたのである。」; 5 reads 「庭先に生動する楓や野菊の様子、流れる渡る日の光、風ともつかぬゆるやかな大気の動き、鶯や雀の声、それらの仔細を美しい言葉にのせて綴った文章を読んで、はじめて、…かつ知ったような気がした。大自然を凌駕する文芸の力がそこにあった。」; 6 reads 「【42】言った。――――「昔、文学をやろうと思ったこともあったが、永井荷風がいるんでやめたよ。…」.
• 問題7: Paragraphs 2, 5, and 6 print a long dash run of four 2-em dashes (――――, four U+2015); transcribed as printed.
• 問題7: Source inconsistency: the passage body and answer choices use 永井荷風 / 荷風, but 注1 glosses the name as 永井和風. Transcribed exactly as printed.
• 問題7: Final sentence ends with 「見当らないようであるか？」 — the trailing か？ on a -ようである clause reads oddly but is printed as-is; kept as printed.
• 問題8: q45 answer 3 reads '材料が加工されて内容に深み増やしたもの' as printed; '深み増やした' is awkward (likely a retype omitting に/を), left as-is.
• 問題8: q46 answer 1 reads '街全体が脈やかに感じられること'; '脈やか' is a suspicious wording (likely a retype error for '賑やか', which appears in the passage body), transcribed as printed.
• 問題8: q47 answer 1 reads '文化がかわりに' as printed; phrasing is awkward, left uncorrected.
• 問題8: In q48 the term is printed in 「」 quotes in the stem, but the same phrase '夢を売る' appears underlined (without quotes) in the passage body; both transcribed as printed.
• 問題9: This is a Chinese prep-school re-typeset recall version. Passage bodies were read directly from the rendered page images (the PDF text layer is scrambled custom-font mojibake), cross-checked against the double-resolution half-page crops.
• 問題9: Passage (1) verified character-by-character against p13-t crops; transcriber B's passage (1) was heavily garbled and discarded.
• 問題9: Passage (2) para 1: the printed source itself contains a stray inserted phrase. The printed text reads "...自分の考え筆者が言いたいことは何か。を率直に述べるということを...". This retype garble is transcribed verbatim per the exact-transcription rule (transcriber A silently dropped it; transcriber B captured the garble but corrupted surrounding text).
• 問題9: Passage (2) para 1: "気楽に自分の考えを話すような習慣できる。" reads as printed; "習慣できる" (rather than 習慣ができる) is unusual but kept verbatim.
• 問題9: Passage (3): the sentence "...煮つめられた模範ではない「教師と生徒」は近代のものだが、" is missing a full-width period (。) after "ではない"; transcribed exactly as printed (likely a retype typo).
• 問題9: Passage (3): the body prints 煮つめられた (hiragana つ) while the （注2） gloss prints 煮詰められた (kanji 詰); both kept exactly as printed in their respective places (transcriber A had normalized the body to 詰).
• 問題9: Q54 answer 4 "否定的な思考が直接な原因となる。" — "直接な" kept as printed.
• 問題9: Q55 answer 2 "弟子が師に選べれば、師の決定に従うことになる。" — "師に選べれば" odd particle kept as printed.
• 問題9: Q55 answer 1 "弟子が師から優れたものを学ぶ個別なつながりである。" — "個別な" kept as printed.
• 問題10: p16 line: 'その速度はほとんど速くなっている' reads oddly (one would expect 確実に/ますます速くなっている), but it is printed with ほとんど; transcribed as printed.
• 問題11: 問題11 instruction is redundantly worded in the source (「…最も良いものを1・2・3・4から最もよいものを一つえらびなさい。」), with 「最も良いもの」 and 「最もよいもの」 both appearing; transcribed as printed.
• 問題11: Q62 option 4 prints 「独創的なアイデアを持っこと」 with a small っ rather than the expected 「持つこと」; confirmed against the double-resolution image and transcribed exactly as printed (likely a retype typo).
• 問題12: Throughout the section the printed glyph for what should be the kanji 力 (chikara, as in メディア力) is actually the katakana カ — both passage and all Q66 stems/answers read 「メディアカ」; transcribed exactly as printed.
• 問題12: Retype typos transcribed as printed: 'わかかった' in 「10年間過ごしたことがわかかった」; '1％異常' in 「1％異常よりも」 (likely intended 異状/異常); Q66 answer 1 「正確に伝われるように」 (likely intended 伝わる).
• 問題13: Passage 「陶芸器の生地の製造」 likely intends 陶磁器 (cf. the instruction's 陶磁器); transcribed as printed (陶芸器).
• 問題13: Passage 「七日前までにお申し込みをお願います」 — 「お願います」 appears to be a typo for 「お願いします」; transcribed exactly as printed.
• 問題13: Availability grid verified against the double-resolution p22-b half-page: ① 9時 = X ○ ○ X △ ○ X, ② 10時 = △ X X ○ X X X, ③ 14時 = △ △ ○ ○ X ○ △ for dates 15-21.

Corrections applied (20)

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

• 問5 q33 question: 「旅館のパンプレットで」 → 「旅館のパンフレットで」
  • パンプレット is a non-word (missing フ); the standard loanword spelling is パンフレット. The scene label was meant to read パンフレット.
  • evidence: Clear non-word. jlpt-freq witness reads パンプレット; teikyo OCR pass A read パンフレット while pass B read パンプレット, confirming the intended form is パンフレット. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問7 q41 passage: 「永井和風」 → 「永井荷風」
  • 注1 glosses the author's name, but the passage body uses 永井荷風 (Nagai Kafū, a real author) consistently. 和風 is a wrong-kanji retype error for 荷風; the gloss must match the name it explains.
  • evidence: Both witnesses confirm body = 永井荷風 (2x) and note = 永井和風; teikyo OCR passes A/B/final all preserve this internal inconsistency, and 永井荷風 is a documented real author. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問8 q46 answers[0]: 「脈やかに」 → 「賑やかに」
  • 脈やか is a non-word; the intended word is 賑やか (lively), which appears in the passage body. 脈/賑 is a retype/OCR confusion. This is a wrong-answer choice but should still print a real word.
  • evidence: Passage body contains 賑やか (both witnesses); jlpt-freq and teikyo OCR both render the answer as 脈やか, a clear non-word. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問9 q52 passage: 「自分の考え筆者が言いたいことは何か。を率直に述べる」 → 「自分の考えを率直に述べる」
  • A question-stem fragment (筆者が言いたいことは何か。) was accidentally spliced into the middle of the sentence; it duplicates question 51's stem and breaks the grammar. Removing it restores the sentence 自分の考えを率直に述べる, confirmed by answer[2] of this same question (自分の考えを率直に述べること).
  • evidence: Note: the proposal's original evidence is INACCURATE — the jlpt-freq witness DOES contain the fragment (自分の考え筆者が言いたいことは何か。 を率直に述べる), and teikyo OCR shows it too (in varying positions). However the splice is unambiguously spurious and answer[2] of #52 reads the clean 自分の考えを率直に述べること, fully grounding the clean reconstruction. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問11 q62 answers[3]: 「持っこと」 → 「持つこと」
  • 持っこと with small っ is a non-word; the intended form is 持つこと. This is the correct answer.
  • evidence: Clear non-word (small っ for つ); both witnesses (jlpt-freq and teikyo OCR A/B/final) propagate 持っこと. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q63 passage: 「ふさわしい「メディアカ」を相手」 → 「ふさわしい「メディア力」を相手」
  • Katakana カ is a glyph confusion for kanji 力 (ちから); the intended term is メディア力 (media power), a meaningful coinage, whereas メディアカ is meaningless.
  • evidence: カ/力 retype confusion; teikyo OCR rendered the term as メディア力 once and メディアカ otherwise, confirming the kanji 力 is intended; output has exactly 8 メディアカ occurrences. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q63 passage: 「たあなたの「メディアカ」が決ま」 → 「たあなたの「メディア力」が決ま」
  • Katakana カ is a glyph confusion for kanji 力; the intended term is メディア力 (media power).
  • evidence: カ/力 retype confusion; teikyo OCR recovered メディア力 in one pass; from string is an exact substring. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q63 passage: 「実像に近い「メディアカ」が形成」 → 「実像に近い「メディア力」が形成」
  • Katakana カ is a glyph confusion for kanji 力; the intended term is メディア力 (media power).
  • evidence: カ/力 retype confusion; teikyo OCR recovered メディア力 in one pass; from string is an exact substring. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q66 question: 「自分の「メディアカ」について」 → 「自分の「メディア力」について」
  • Katakana カ is a glyph confusion for kanji 力; the intended term is メディア力 (media power).
  • evidence: カ/力 retype confusion; teikyo OCR recovered メディア力 in one pass; from string is an exact substring. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q66 answers[0]: 「「メディアカ」が正確に伝われるように」 → 「「メディア力」が正確に伝わるように」
  • Two retype artifacts: katakana カ for kanji 力 (メディア力, media power), and the awkward potential form 伝われる for the natural intransitive 伝わる. 正確に伝わるように ('so that it is conveyed accurately') is the natural phrasing; this is the correct answer.
  • evidence: カ/力 confusion confirmed by OCR; teikyo OCR pass A read 正確に伝わるように (pass B read 伝われる), so 伝わる is the intended natural intransitive. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q66 answers[1]: 「「メディアカ」を伝える」 → 「「メディア力」を伝える」
  • Katakana カ is a glyph confusion for kanji 力; the intended term is メディア力 (media power).
  • evidence: カ/力 retype confusion; teikyo OCR recovered メディア力 in one pass; from string is an exact substring. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q66 answers[2]: 「「メディアカ」を気にせず」 → 「「メディア力」を気にせず」
  • Katakana カ is a glyph confusion for kanji 力; the intended term is メディア力 (media power).
  • evidence: カ/力 retype confusion; teikyo OCR recovered メディア力 in one pass; from string is an exact substring. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問12 q66 answers[3]: 「「メディアカ」を下げない」 → 「「メディア力」を下げない」
  • Katakana カ is a glyph confusion for kanji 力; the intended term is メディア力 (media power).
  • evidence: カ/力 retype confusion; teikyo OCR recovered メディア力 in one pass; from string is an exact substring. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問13 q67 passage: 「陶芸器の生地」 → 「陶磁器の生地」
  • 陶芸器 is not a standard word; the company manufactures 陶磁器 (ceramics). The exam's mondai13 instruction line describing this same page reads 陶磁器を作っている会社. 芸/磁 is a retype error.
  • evidence: Output instruction line: 右のページは、陶磁器を作っている会社; teikyo OCR mixes 陶磁器 and 陶芸器 across passes; 陶磁器 is the standard term. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問13 q67 passage: 「お申し込みをお願います」 → 「お申し込みをお願いします」
  • お願います is a non-word (missing し); the standard polite form is お願いします.
  • evidence: Clear non-word; jlpt-freq and most teikyo OCR passes read お願います, but OCR pass A recovered お願いします, confirming the intended form. [correct-retypes workflow (propose + adversarial verify), 2026-06-17]
• 問3 q16 question: 「<u>架空</u>の人物」 → 「<u>架空の</u>人物」
  • 問題3 underline boundary: the synonym options are adnominal (素敵な/思い出の/想像の/変な), so the underlined span that they substitute for must be 架空の, not bare 架空 (otherwise 想像の leaves a dangling の). Matches the clean substitution pattern of the sibling questions q14/q15/q18.
  • evidence: structural: 問題3 options replace the underlined span; the の must be inside the underline for grammatical substitution.
• 問12 q63 passage: 「過ごしたことがわかかった」 → 「過ごしたことがわかった」
  • 「わかかった」 is a non-word; the doubled か is a retype error for 「わかった」 (it became clear that you spent 10 years). Same passage uses 「わかった」 correctly elsewhere.
  • evidence: Category typo, confidence high. [retype-corrections-pass2 (propose + adversarial verify), 2026-06-20]
• 問9 q52 passage: 「習慣できる」 → 「習慣ができる」
  • 「習慣ができる」 requires the particle が (習慣 + が + できる). 「習慣できる」 drops が; a dropped-character retype error in the passage body.
  • evidence: Category dropped-char, confidence medium. [retype-corrections-pass2 (propose + adversarial verify), 2026-06-20]
• 問9 q55 passage: 「模範ではない「教師と生徒」 → 「模範ではない。「教師と生徒」
  • Two distinct sentences are run together with no period: 「…煮つめられた模範ではない」 ends a sentence, and 「「教師と生徒」は近代のものだが…」 begins the next. A 。 was dropped in the retype.
  • evidence: Category typo, confidence medium. [retype-corrections-pass2 (propose + adversarial verify), 2026-06-20]
• 問8 q45 answers[2]: 「内容に深み増やしたもの」 → 「内容に深みを増やしたもの」
  • 「深み増やした」 is ungrammatical; the object particle を is missing (深みを増やした). This is a comprehension distractor (not a 問題4 deliberate-misuse sentence), so the broken grammar is a retype dropped-char rather than intended.
  • evidence: Category dropped-char, confidence medium. [retype-corrections-pass2 (propose + adversarial verify), 2026-06-20]