{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-23-018",
    "source_story_id": "tmp-brief-tts-verifier-floor",
    "edition_id": "mp-2026-10-23-morning-0105",
    "edition_url": "https://themachinepress.com/edition/2026-10-23",
    "position": 20,
    "story_type": "ticker",
    "section": "frontier-models",
    "editorial_classification": "editorial",
    "headline": "The Speech Evaluator Was Mishearing the Rare Words",
    "slug": "the-speech-evaluator-was-mishearing-the-rare-words",
    "dek": "Evaluation recognizers missed 33 to 67 percent of very rare words even in human recordings used as the control.",
    "summary": "Evaluation recognizers missed 33 to 67 percent of very rare words even in human recordings used as the control.",
    "body_text": "Adding two recognizer families cut apparent very-rare-word TTS error from roughly 10 to 13 percent down to 3 to 5 percent on words at least one system heard. Yet masking recognizer misses introduced its own bias, and the authors conclude that listening tests remain necessary.",
    "why_it_matters": "Evaluation recognizers missed 33 to 67 percent of very rare words even in human recordings used as the control.",
    "limitations": [
      "Yet masking recognizer misses introduced its own bias, and the authors conclude that listening tests remain necessary."
    ],
    "importance": 7,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-23-018/the-speech-evaluator-was-mishearing-the-rare-words",
    "json_url": "https://themachinepress.com/story/mp-2026-10-23-018.json",
    "first_published_at": "2026-10-23T09:00:00.000-04:00",
    "modified_at": "2026-10-23T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-23-018-001",
        "text": "Evaluation recognizers missed 33 to 67 percent of very rare words even in human recordings used as the control.",
        "source_ids": [
          "source-2026-10-23-020"
        ],
        "qualification": "Yet masking recognizer misses introduced its own bias, and the authors conclude that listening tests remain necessary."
      }
    ],
    "source_ids": [
      "source-2026-10-23-020"
    ],
    "tags": [
      "speech",
      "evaluation",
      "ASR"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-23-020",
      "title": "arXiv preprint 2610.31470, original v1",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.31470v1",
      "canonical_url": "https://arxiv.org/abs/2610.31470v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-22T11:59:54.000-04:00",
      "accessed_at": "2026-10-23T08:21:38.024-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-23-018-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "The Speech Evaluator Was Mishearing the Rare Words",
    "publisher": "The Machine Press",
    "published_at": "2026-10-23T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-23-018/the-speech-evaluator-was-mishearing-the-rare-words"
  }
}