{
  "title": "Public speaking practice excerpt data for AI",
  "source_page": "https://fiund.com/industries/public-speaking-practice-excerpt-data",
  "authorship": "Original fictional Fiund editorial exercise. No actual operating records or third-party source assets are included.",
  "example": {
    "title": "Attribute a timing note to the take that contains the pause",
    "context": "Dummy timing records describe two takes of an original adult-authored practice script. No real learner assessment, personal story or audience recording is included.",
    "records": [
      [
        "PS1",
        "Take A timing card",
        "Excerpt 00:08–00:14 contains a two-second pause after the word harbor."
      ],
      [
        "PS2",
        "Take B timing card",
        "Same script; excerpt 00:08–00:12 has no marked pause after harbor."
      ],
      [
        "PS3",
        "Coach's practice note",
        "Names Take A at 00:10 and asks to compare the marked pause. File was also copied into Take B's folder."
      ]
    ],
    "question": "Which take supports the coach's specific note, and what would be needed to claim a broader improvement between takes?",
    "answer": "PS1 and PS3 support the note about Take A's marked pause around 00:10. Copying the file into Take B's folder does not change that reference, and PS2 records no corresponding marked pause in its excerpt. The packet supports a narrow timing difference, not a general improvement in delivery. A broader evaluation would need an explicit rubric, suitable recordings and an authorized reviewer.",
    "failureModes": [
      "Attributing the pause to Take B from folder location.",
      "Turning a shorter excerpt or absent pause marker into a broad quality judgment."
    ],
    "fields": [
      [
        "Take and excerpt boundaries",
        "Connects commentary to the exact source span."
      ],
      [
        "Script anchor and timing",
        "Identifies which pause is being discussed."
      ],
      [
        "Note origin versus copied location",
        "Prevents duplicate filing from becoming a new observation."
      ]
    ],
    "holdout": "Move the marked pause to a different word in a third take. Ask the model to locate the new anchor rather than matching time alone.",
    "boundary": "Use dummy timing cards or original scripts recorded by consenting adults. Exclude actual learner feedback, personal narratives, student records and third-party scripts."
  }
}