{ "method": "Case-folded word tokens; curved apostrophe normalized; punctuation/whitespace ignored. Python difflib.SequenceMatcher, autojunk=False. No audio or ASR evaluation.", "created": "2026-10-31", "authorship": "All reference sentences and deliberately altered drafts were written for this page. Names and scenarios are fictional.", "total": 10, "wording_difference": 8, "no_wording_difference": 2, "cases": [ { "id": "01", "title": "A missing negative reverses permission", "kind": "changed", "reference": "I do not publish client names.", "draft": "I do publish client names.", "check": "Replay the verb together with the word immediately before it. The missing “not” reverses the statement; adding polished punctuation would not repair it. Keep the negative in your checked quotation and do not infer permission from the shorter draft.", "result": { "wording_difference": true, "changes": [ { "operation": "delete", "reference_tokens": [ "not" ], "draft_tokens": [] } ], "reference_token_count": 6, "draft_token_count": 5 } }, { "id": "02", "title": "A similar sound changes the budget", "kind": "changed", "reference": "Budget fifteen dollars for the draft.", "draft": "Budget fifty dollars for the draft.", "check": "Listen specifically to the amount, then write it beside the currency. Here “fifteen” and “fifty” are different instructions. A readable sentence cannot decide which amount the speaker intended. Do not silently choose the more plausible budget.", "result": { "wording_difference": true, "changes": [ { "operation": "replace", "reference_tokens": [ "fifteen" ], "draft_tokens": [ "fifty" ] } ], "reference_token_count": 6, "draft_token_count": 6 } }, { "id": "03", "title": "The value stays, but the unit changes", "kind": "changed", "reference": "Leave a gap of fifteen milliseconds.", "draft": "Leave a gap of fifteen seconds.", "check": "Audit the quantity and its unit as one phrase. Both lines retain “fifteen,” so checking numbers alone misses the problem. Replay the full instruction before copying it into a tutorial, brief, or caption. Preserve the unit actually spoken.", "result": { "wording_difference": true, "changes": [ { "operation": "replace", "reference_tokens": [ "milliseconds" ], "draft_tokens": [ "seconds" ] } ], "reference_token_count": 6, "draft_token_count": 6 } }, { "id": "04", "title": "A plausible name still needs a source", "kind": "changed", "reference": "Ask Mira Vale before posting.", "draft": "Ask Meera Vale before posting.", "check": "Mira Vale is a fictional name written for this exercise. The similar-looking draft is not a verified alternative spelling. In a real clip, check the speaker’s own written identification when available; if you cannot establish the spelling, mark it unresolved instead of guessing.", "result": { "wording_difference": true, "changes": [ { "operation": "replace", "reference_tokens": [ "mira" ], "draft_tokens": [ "meera" ] } ], "reference_token_count": 5, "draft_token_count": 5 } }, { "id": "05", "title": "The speaker corrects the instruction", "kind": "changed", "reference": "Send three drafts—sorry, two drafts.", "draft": "Send three drafts.", "check": "Replay beyond the first complete sentence or number. The correction changes the requested delivery. Keep the correction in a verbatim transcript; if you write a summary, report the final instruction and label it as your summary rather than presenting the shorter draft as an exact quote.", "result": { "wording_difference": true, "changes": [ { "operation": "delete", "reference_tokens": [ "sorry", "two", "drafts" ], "draft_tokens": [] } ], "reference_token_count": 6, "draft_token_count": 3 } }, { "id": "06", "title": "The words move to the wrong speaker", "kind": "changed", "reference": "A: Can I publish it?\nB: Not yet.", "draft": "A: Can I publish it? Not yet.", "check": "Compare speaker turns, not just the combined words. The draft assigns the reply to A. Listen through the change of voice and retain separate turns when attribution matters. A word comparison can flag a missing label, but it cannot identify a person from the audio.", "result": { "wording_difference": true, "changes": [ { "operation": "delete", "reference_tokens": [ "b" ], "draft_tokens": [] } ], "reference_token_count": 8, "draft_token_count": 7 } }, { "id": "07", "title": "A visual note becomes invented speech", "kind": "changed", "reference": "[On screen: Step 2]\nSpoken: Wait until tomorrow.", "draft": "Step two. Wait until tomorrow.", "check": "The bracketed line is an author-written visual annotation, not part of the spoken sentence. Keep visible text in a separate note. For a real Reel, replay with sound before putting a phrase inside quotation marks; a readable overlay does not prove someone said it.", "result": { "wording_difference": true, "changes": [ { "operation": "delete", "reference_tokens": [ "on", "screen" ], "draft_tokens": [] }, { "operation": "replace", "reference_tokens": [ "2", "spoken" ], "draft_tokens": [ "two" ] } ], "reference_token_count": 8, "draft_token_count": 5 } }, { "id": "08", "title": "A short extract drops its condition", "kind": "changed", "reference": "Use the clip only if the speaker has approved the edit.", "draft": "Use the clip.", "check": "Replay the end of the sentence, including any “only if,” “unless,” or exception. This shortened draft removes a condition. Save the complete checked passage before writing a shorter quotation. If the audio ends before you can hear the condition, leave the claim unverified.", "result": { "wording_difference": true, "changes": [ { "operation": "delete", "reference_tokens": [ "only", "if", "the", "speaker", "has", "approved", "the", "edit" ], "draft_tokens": [] } ], "reference_token_count": 11, "draft_token_count": 3 } }, { "id": "09", "title": "Control: straight and curved apostrophes", "kind": "format-control", "reference": "I won't post it.", "draft": "I won’t post it.", "check": "This control uses the same words. The comparison deliberately treats straight and curved apostrophes alike. It is a formatting control, not proof that either line matches a recording.", "result": { "wording_difference": false, "changes": [], "reference_token_count": 4, "draft_token_count": 4 } }, { "id": "10", "title": "Control: a line break becomes a space", "kind": "format-control", "reference": "Check the name.\nThen replay the line.", "draft": "Check the name. Then replay the line.", "check": "This control changes layout only. The comparison ignores whitespace and punctuation between tokens. It does not check subtitle timing, sentence meaning, speaker identity, or audio completeness.", "result": { "wording_difference": false, "changes": [], "reference_token_count": 7, "draft_token_count": 7 } } ] }