{
  "schema_version": 1,
  "acquired_at_utc": "2026-07-21T04:24:12Z",
  "status": "complete",
  "sample": {
    "id": "aishell4-test-S_R003S01C01",
    "source_dataset": "AISHELL-4",
    "source_split": "test",
    "source_id": "S_R003S01C01",
    "language": "zh",
    "scenario": "meeting_overlap",
    "duration_seconds": 2220.529,
    "speaker_count": 5
  },
  "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/S_R003S01C01.flac",
      "url": "https://huggingface.co/datasets/AISHELL/AISHELL-4/resolve/aada72727856313b19d4a030383c426364931dbf/test/wav/S_R003S01C01.flac",
      "expected_size_bytes": 255177109,
      "expected_lfs_sha256": "e4a8a76315b7dabe63f43e2364486d9dee989ddcf111d73dc3eae47d86838cd0",
      "local_path": "audio.multichannel.flac",
      "size_bytes": 255177109,
      "sha256": "e4a8a76315b7dabe63f43e2364486d9dee989ddcf111d73dc3eae47d86838cd0",
      "checksum_verified": true
    },
    {
      "role": "reference_textgrid",
      "repository_path": "test/TextGrid/S_R003S01C01.TextGrid",
      "url": "https://huggingface.co/datasets/AISHELL/AISHELL-4/resolve/aada72727856313b19d4a030383c426364931dbf/test/TextGrid/S_R003S01C01.TextGrid",
      "repository_git_oid": "2be87dead5e0095dfcb9cfd988a1e99a76bf12d0",
      "local_path": "reference.TextGrid",
      "size_bytes": 198779,
      "sha256": "fe1ac420affae1e377e1a254c72c992b560ea02d3a5d3c7e22b89f5faacf283b"
    },
    {
      "role": "reference_rttm",
      "repository_path": "test/TextGrid/S_R003S01C01.rttm",
      "url": "https://huggingface.co/datasets/AISHELL/AISHELL-4/resolve/aada72727856313b19d4a030383c426364931dbf/test/TextGrid/S_R003S01C01.rttm",
      "repository_git_oid": "4185e0e3e2092774cfe9efaf97de3ec909f9d6d9",
      "local_path": "reference.rttm",
      "size_bytes": 46995,
      "sha256": "1fe23bd4aef6575bfe61f0bc4875bc54fb39044ec013c6383546217ccf3e56e5"
    }
  ],
  "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": 71057006,
    "sha256": "352d72f24c6e98f54897a15fe47b162bf6f1672ad7f070a125369b43f9ac3edb"
  },
  "ffprobe": {
    "source_audio": {
      "format_name": "flac",
      "codec_name": "flac",
      "sample_format": "s16",
      "sample_rate_hz": 16000,
      "channels": 8,
      "channel_layout": "7.1",
      "duration_seconds": 2220.529,
      "size_bytes": 255177109,
      "bit_rate_bps": 919338
    },
    "evaluation_audio": {
      "format_name": "wav",
      "codec_name": "pcm_s16le",
      "sample_format": "s16",
      "bits_per_sample": 16,
      "sample_rate_hz": 16000,
      "channels": 1,
      "duration_seconds": 2220.529,
      "size_bytes": 71057006,
      "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": 2220.529,
    "textgrid_tier_count": 5,
    "rttm_file_ids": [
      "S_R003S01C01"
    ],
    "rttm_speaker_ids": [
      "001-M",
      "002-F",
      "003-M",
      "004-F",
      "005-M"
    ],
    "rttm_segment_count": 716,
    "rttm_max_speech_end_seconds": 2209.7518,
    "speech_union_seconds": 2042.7017,
    "overlap_seconds": 80.6979,
    "overlap_percent_of_speech_union": 3.9505,
    "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/S_R003S01C01/reference.TextGrid --plain-output data/samples/meetings/aishell-4/S_R003S01C01/reference.txt --speaker-output data/samples/meetings/aishell-4/S_R003S01C01/reference.speaker.txt --segments-output data/samples/meetings/aishell-4/S_R003S01C01/reference.segments.jsonl --metadata-output data/samples/meetings/aishell-4/S_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": 717,
    "segment_count_scoreable": 716,
    "excluded_event_only_interval_count": 1,
    "ambiguous_overlap_group_count": 122,
    "artifacts": [
      {
        "role": "plain_reference",
        "local_path": "reference.txt",
        "size_bytes": 32031,
        "sha256": "a6a426e6ea55998c2089fe4b83df8d587370a92a54cd6e961ab6c0a305c57b52"
      },
      {
        "role": "speaker_tagged_reference",
        "format": "[start_seconds][Sxx]raw_text[end_seconds]",
        "local_path": "reference.speaker.txt",
        "size_bytes": 52789,
        "sha256": "f2b63edf798d0b20bdbc32ffc119ecf59b8532f6a4855e141676daaecc72950f"
      },
      {
        "role": "lossless_segment_ledger",
        "local_path": "reference.segments.jsonl",
        "size_bytes": 206455,
        "sha256": "27ae152d2b14487ebf308ce82b1eee3cc172512a8aeec08ff6bb0e9dda47662d"
      },
      {
        "role": "ordering_and_overlap_metadata",
        "local_path": "reference.ordering.json",
        "size_bytes": 36713,
        "sha256": "c5cc3c41a781be037b170f224d815e6ecee264187d084107ec19d70e57aa7a47"
      }
    ]
  },
  "retention": {
    "original_multichannel_audio_retained": true,
    "source_annotations_retained": true,
    "download_urls_recorded": true,
    "source_revision_recorded": true
  }
}