{
  "schema_version": 1,
  "acquired_at_utc": "2026-07-21T03:40:45Z",
  "status": "complete",
  "sample": {
    "id": "aishell4-test-L_R003S01C02",
    "source_dataset": "AISHELL-4",
    "source_split": "test",
    "source_id": "L_R003S01C02",
    "language": "zh",
    "scenario": "meeting_overlap",
    "duration_seconds": 2362.945,
    "speaker_count": 6
  },
  "license": {
    "spdx_like": "CC-BY-SA-4.0",
    "name": "Creative Commons Attribution-ShareAlike 4.0 International",
    "authoritative_source": "https://www.openslr.org/111/",
    "note": "The Hugging Face repository metadata has advertised Apache-2.0; this acquisition uses the upstream OpenSLR/AISHELL CC BY-SA 4.0 declaration as authoritative."
  },
  "source": {
    "provider": "huggingface",
    "repository": "AISHELL/AISHELL-4",
    "repository_type": "dataset",
    "revision": "aada72727856313b19d4a030383c426364931dbf",
    "revision_pinned": true,
    "upstream_distribution": "https://www.openslr.org/111/",
    "upstream_specification": "https://aishell-4.oss-cn-hangzhou.aliyuncs.com/AISHELL-4%20Data-Specification.pdf",
    "official_baseline": {
      "repository": "https://github.com/felixfuyihui/AISHELL-4",
      "commit": "bad82b77c3753df1b232c5c6491cd3e2f2e32d24"
    }
  },
  "downloads": [
    {
      "role": "source_audio_multichannel",
      "repository_path": "test/wav/L_R003S01C02.flac",
      "url": "https://huggingface.co/datasets/AISHELL/AISHELL-4/resolve/aada72727856313b19d4a030383c426364931dbf/test/wav/L_R003S01C02.flac",
      "expected_size_bytes": 286036324,
      "expected_lfs_sha256": "93125b2c8f9f73c042ccd63178d1fec6db2d3047af4078834da7c6c1aa118e72",
      "local_path": "audio.multichannel.flac",
      "size_bytes": 286036324,
      "sha256": "93125b2c8f9f73c042ccd63178d1fec6db2d3047af4078834da7c6c1aa118e72",
      "checksum_verified": true
    },
    {
      "role": "reference_textgrid",
      "repository_path": "test/TextGrid/L_R003S01C02.TextGrid",
      "url": "https://huggingface.co/datasets/AISHELL/AISHELL-4/resolve/aada72727856313b19d4a030383c426364931dbf/test/TextGrid/L_R003S01C02.TextGrid",
      "repository_git_oid": "a000813567cc1d09cb5431b54ce3f647b167251d",
      "local_path": "reference.TextGrid",
      "size_bytes": 148034,
      "sha256": "a7e060c87f76f6ad51ca1bcd6ae7243942757f25bdcbfc9a628c552d03dcbc4c"
    },
    {
      "role": "reference_rttm",
      "repository_path": "test/TextGrid/L_R003S01C02.rttm",
      "url": "https://huggingface.co/datasets/AISHELL/AISHELL-4/resolve/aada72727856313b19d4a030383c426364931dbf/test/TextGrid/L_R003S01C02.rttm",
      "repository_git_oid": "29ca40077633af3377ed7e9f6bb1e158a6f5ce37",
      "local_path": "reference.rttm",
      "size_bytes": 33636,
      "sha256": "9dfd301c5b96f60813aaddec2549edc007851699917d9326e4db7e13c6b05ed9"
    }
  ],
  "derived_audio": {
    "local_path": "audio.wav",
    "source_path": "audio.multichannel.flac",
    "source_channel_index_zero_based": 0,
    "selection_rationale": "AISHELL-4 uses an eight-channel circular array. The upstream specification says transcription alignment used the first array channel; the official diarization baseline also selects the first channel.",
    "command": "ffmpeg -hide_banner -nostdin -y -i audio.multichannel.flac -map 0:a:0 -af 'pan=mono|c0=c0' -ar 16000 -c:a pcm_s16le audio.wav",
    "tool": "ffmpeg 6.1.1-3ubuntu5",
    "size_bytes": 75614318,
    "sha256": "b8536df94481a65cd27d9213541e7832582590e09792804ab1be4b6b0d269a70"
  },
  "ffprobe": {
    "source_audio": {
      "format_name": "flac",
      "codec_name": "flac",
      "sample_format": "s16",
      "sample_rate_hz": 16000,
      "channels": 8,
      "channel_layout": "7.1",
      "duration_seconds": 2362.945,
      "size_bytes": 286036324,
      "bit_rate_bps": 968406
    },
    "evaluation_audio": {
      "format_name": "wav",
      "codec_name": "pcm_s16le",
      "sample_format": "s16",
      "bits_per_sample": 16,
      "sample_rate_hz": 16000,
      "channels": 1,
      "duration_seconds": 2362.945,
      "size_bytes": 75614318,
      "bit_rate_bps": 256000
    }
  },
  "reference_validation": {
    "textgrid_xmax_seconds": 2362.945,
    "textgrid_tier_count": 6,
    "rttm_file_ids": [
      "L_R003S01C02"
    ],
    "rttm_speaker_ids": [
      "002-F",
      "003-F",
      "004-M",
      "005-M",
      "006-M",
      "007-M"
    ],
    "rttm_segment_count": 512,
    "rttm_max_speech_end_seconds": 2266.755,
    "speech_union_seconds": 2114.3912,
    "overlap_seconds": 188.3593,
    "overlap_percent_of_speech_union": 8.9084,
    "duration_matches_source_audio": true,
    "evaluation_audio_matches_source_duration": true
  },
  "reference_conversion": {
    "converter": "asr_eval.textgrid_reference",
    "converter_version": "1.0.0",
    "converter_source": "src/asr_eval/textgrid_reference.py",
    "command": "uv run asr-eval convert-textgrid-reference --input data/samples/meetings/aishell-4/L_R003S01C02/reference.TextGrid --plain-output data/samples/meetings/aishell-4/L_R003S01C02/reference.txt --speaker-output data/samples/meetings/aishell-4/L_R003S01C02/reference.speaker.txt --segments-output data/samples/meetings/aishell-4/L_R003S01C02/reference.segments.jsonl --metadata-output data/samples/meetings/aishell-4/L_R003S01C02/reference.ordering.json",
    "deterministic_sort_key": [
      "start_seconds",
      "end_seconds",
      "source_tier_index",
      "source_interval_index"
    ],
    "is_unique_semantic_order": false,
    "ordering_note": "Cross-speaker intervals overlap. The serialized references are deterministic but do not claim a uniquely correct word order inside overlap regions.",
    "segment_count_nonempty": 515,
    "segment_count_scoreable": 512,
    "excluded_event_only_interval_count": 3,
    "ambiguous_overlap_group_count": 103,
    "artifacts": [
      {
        "role": "plain_reference",
        "local_path": "reference.txt",
        "size_bytes": 28824,
        "sha256": "af79593c93c5d2af3f3edff5cb3ccc4ad93afb6bc3b198e737c864b83ea7506b"
      },
      {
        "role": "speaker_tagged_reference",
        "format": "[start_seconds][Sxx]raw_text[end_seconds]",
        "local_path": "reference.speaker.txt",
        "size_bytes": 43112,
        "sha256": "4e961433a2a1b63e7e42985f674a716422a495ff8b5868fb448ccad69f792dbb"
      },
      {
        "role": "lossless_segment_ledger",
        "local_path": "reference.segments.jsonl",
        "size_bytes": 154682,
        "sha256": "739916be57936136f12768c7e39a612ffa20f3bf0b553110ee1d02cb51d37a55"
      },
      {
        "role": "ordering_and_overlap_metadata",
        "local_path": "reference.ordering.json",
        "size_bytes": 31661,
        "sha256": "a3f5753643dfcf5e8541626747594c0402663b585f96699eb7569115a5adfe26"
      }
    ]
  },
  "retention": {
    "original_multichannel_audio_retained": true,
    "source_annotations_retained": true,
    "download_urls_recorded": true,
    "source_revision_recorded": true
  }
}