{
  "schema_version": 1,
  "acquired_at_utc": "2026-07-21T04:24:12Z",
  "status": "complete",
  "sample": {
    "id": "aishell4-test-M_R003S01C01",
    "source_dataset": "AISHELL-4",
    "source_split": "test",
    "source_id": "M_R003S01C01",
    "language": "zh",
    "scenario": "meeting_overlap",
    "duration_seconds": 2282.21,
    "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/M_R003S01C01.flac",
      "url": "https://huggingface.co/datasets/AISHELL/AISHELL-4/resolve/aada72727856313b19d4a030383c426364931dbf/test/wav/M_R003S01C01.flac",
      "expected_size_bytes": 252564853,
      "expected_lfs_sha256": "26b4cd2045077e3db7170c376db3cb63b469d208e1d474155721911b3fe9866a",
      "local_path": "audio.multichannel.flac",
      "size_bytes": 252564853,
      "sha256": "26b4cd2045077e3db7170c376db3cb63b469d208e1d474155721911b3fe9866a",
      "checksum_verified": true
    },
    {
      "role": "reference_textgrid",
      "repository_path": "test/TextGrid/M_R003S01C01.TextGrid",
      "url": "https://huggingface.co/datasets/AISHELL/AISHELL-4/resolve/aada72727856313b19d4a030383c426364931dbf/test/TextGrid/M_R003S01C01.TextGrid",
      "repository_git_oid": "43011b30256fd0c718beade037a1a245a8b86bea",
      "local_path": "reference.TextGrid",
      "size_bytes": 178461,
      "sha256": "35eaaab7f13e594393adacdabbacac5779e2850992277d72dfbcfd12320c5782"
    },
    {
      "role": "reference_rttm",
      "repository_path": "test/TextGrid/M_R003S01C01.rttm",
      "url": "https://huggingface.co/datasets/AISHELL/AISHELL-4/resolve/aada72727856313b19d4a030383c426364931dbf/test/TextGrid/M_R003S01C01.rttm",
      "repository_git_oid": "e9681ca479781671eb1f269bd3c2cf6b90eb7798",
      "local_path": "reference.rttm",
      "size_bytes": 42175,
      "sha256": "bf67a4df2f2d2ae78cccfeefc47ca4244db043e6bbc7a44f83dc9b3d3ee3af57"
    }
  ],
  "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 -v error -xerror -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": 73030798,
    "sha256": "724df9fc8f4210359f6c2296932f4a07db78121bde85bded7044074209d3b483"
  },
  "ffprobe": {
    "source_audio": {
      "format_name": "flac",
      "codec_name": "flac",
      "sample_format": "s16",
      "sample_rate_hz": 16000,
      "channels": 8,
      "channel_layout": "7.1",
      "duration_seconds": 2282.21,
      "size_bytes": 252564853,
      "bit_rate_bps": 885334
    },
    "evaluation_audio": {
      "format_name": "wav",
      "codec_name": "pcm_s16le",
      "sample_format": "s16",
      "bits_per_sample": 16,
      "sample_rate_hz": 16000,
      "channels": 1,
      "duration_seconds": 2282.21,
      "size_bytes": 73030798,
      "bit_rate_bps": 256000
    }
  },
  "decode_validation": {
    "tool": "ffmpeg 6.1.1-3ubuntu5",
    "source_audio": {
      "command": "ffmpeg -hide_banner -nostdin -v error -xerror -i audio.multichannel.flac -f null -",
      "status": "passed",
      "stderr_bytes": 0
    },
    "evaluation_audio": {
      "command": "ffmpeg -hide_banner -nostdin -v error -xerror -i audio.wav -f null -",
      "status": "passed",
      "stderr_bytes": 0
    }
  },
  "reference_validation": {
    "textgrid_xmax_seconds": 2282.21,
    "textgrid_tier_count": 6,
    "rttm_file_ids": [
      "M_R003S01C01"
    ],
    "rttm_speaker_ids": [
      "001-M",
      "002-M",
      "003-F",
      "004-F",
      "005-F",
      "006-F"
    ],
    "rttm_segment_count": 641,
    "rttm_max_speech_end_seconds": 2247.8433,
    "speech_union_seconds": 2058.6923,
    "overlap_seconds": 127.6305,
    "overlap_percent_of_speech_union": 6.1996,
    "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/M_R003S01C01/reference.TextGrid --plain-output data/samples/meetings/aishell-4/M_R003S01C01/reference.txt --speaker-output data/samples/meetings/aishell-4/M_R003S01C01/reference.speaker.txt --segments-output data/samples/meetings/aishell-4/M_R003S01C01/reference.segments.jsonl --metadata-output data/samples/meetings/aishell-4/M_R003S01C01/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": 644,
    "segment_count_scoreable": 641,
    "excluded_event_only_interval_count": 2,
    "ambiguous_overlap_group_count": 100,
    "artifacts": [
      {
        "role": "plain_reference",
        "local_path": "reference.txt",
        "size_bytes": 29226,
        "sha256": "b3c42f5b2a1853ba959ce9c34f5dfc088e2686794f90b9ddd03fc6a0b266c5ab"
      },
      {
        "role": "speaker_tagged_reference",
        "format": "[start_seconds][Sxx]raw_text[end_seconds]",
        "local_path": "reference.speaker.txt",
        "size_bytes": 47870,
        "sha256": "343eee920cd4d2f3bd2e0490e717172030c5817fb012626871d2e8973ea5ff69"
      },
      {
        "role": "lossless_segment_ledger",
        "local_path": "reference.segments.jsonl",
        "size_bytes": 185777,
        "sha256": "3687e60aace55005894d02de3cb1358c9823010c00dd681dbbd7cd6ce84f55bf"
      },
      {
        "role": "ordering_and_overlap_metadata",
        "local_path": "reference.ordering.json",
        "size_bytes": 31448,
        "sha256": "e4c3a20ef29e085409faf4dd931dae7f0cfde5c67388eba1fccf9007d2c4e6c7"
      }
    ]
  },
  "retention": {
    "original_multichannel_audio_retained": true,
    "source_annotations_retained": true,
    "download_urls_recorded": true,
    "source_revision_recorded": true
  }
}