{
  "documentType": "specimen-speech-data-card",
  "specimen": true,
  "schemaVersion": "0.1",
  "notice": "This is a sanitized example of a possible documentation structure. It is not a description of an available dataset, a universal delivery promise, or legal advice.",
  "dataset": {
    "title": "Specimen: consented Danish telephony role-play",
    "status": "illustrative-only",
    "commercialAvailability": false,
    "intendedModelJob": "Example structure for ASR training or held-out telephony evaluation",
    "language": "da-DK",
    "domain": "simulated customer-service conversations"
  },
  "provenance": {
    "source": "Purpose-recorded role-play with recruited participants",
    "productionCustomerCallsUsed": false,
    "collectionMethod": "Two-speaker moderated scenarios",
    "collectionWindow": "[project-specific date range]",
    "collectionMarkets": ["[approved market]"],
    "contributorNotice": "[approved participant-facing project and model-use explanation]",
    "consentRecord": "[project-specific consent record reference]",
    "withdrawalProcess": "[project-specific process and limits]"
  },
  "capture": {
    "sourceAudio": "[wideband source format]",
    "deliveryAudio": "[buyer-approved format and channel layout]",
    "telephonyPath": "[direct call, re-recording path, or codec simulation]",
    "speakerChannels": "[mono mix or synchronized separate channels]",
    "conditions": ["[device]", "[codec]", "[noise band]", "[scenario]"],
    "knownLimitations": ["[documented coverage gap]"]
  },
  "coverage": {
    "usableHours": "[accepted total]",
    "speakerCount": "[unique accepted speakers]",
    "quotaFields": ["locale", "accent", "ageBand", "role", "channel", "scenario"],
    "quotaSummaryReference": "[manifest or report path]"
  },
  "annotation": {
    "transcript": "[verbatim, normalized, or both]",
    "speakerLabels": true,
    "timestamps": "[utterance, segment, word, or none]",
    "optionalFields": ["overlap", "nonSpeechEvent", "intent", "entity"],
    "guidelineVersion": "[version-controlled guideline]",
    "reviewMethod": "[review and adjudication process]"
  },
  "splits": {
    "unitOfSeparation": "speaker",
    "trainingPercent": "[agreed percentage]",
    "validationPercent": "[agreed percentage]",
    "evaluationPercent": "[agreed percentage]",
    "leakageControls": ["speaker-disjoint split", "split manifest", "duplicate-content check"],
    "nonReuseCommitment": "[state whether and how held-out audio may be reused]"
  },
  "quality": {
    "acceptanceRules": [
      "[file readability and format validation]",
      "[transcript accuracy threshold and sampling plan]",
      "[quota tolerance]",
      "[metadata completeness]",
      "[duplicate and silence checks]"
    ],
    "rejectionReasons": ["[controlled rejection-code list]"],
    "replacementPolicy": "[agreed replacement and re-review terms]",
    "qualityReportReference": "[delivery report path]"
  },
  "rightsAndDelivery": {
    "permittedUse": "[contracted training, evaluation, research, or other use]",
    "restrictedUse": ["[project-specific restriction]"],
    "license": "[contract reference]",
    "exclusivity": "[exclusive, non-exclusive, or limited]",
    "retentionAndDeletion": "[contracted terms]",
    "deliveryContents": ["audio", "transcripts", "metadata", "split manifest", "quality report", "license documentation"]
  },
  "exampleManifestRow": {
    "utteranceId": "specimen-call-0001-turn-0007",
    "speakerId": "SPECIMEN_SPK_A",
    "split": "evaluation",
    "language": "da-DK",
    "channel": "speaker-a",
    "scenario": "billing-question",
    "audioPath": "audio/specimen-call-0001_speaker-a.wav",
    "transcriptPath": "transcripts/specimen-call-0001.json",
    "checksum": "[sha256 supplied at delivery]"
  }
}