{
  "comparison": "Nemotron3 Diarization vs pyannote Community-1",
  "inputs": [
    "nemotron/results.json",
    "pyannote/results.json"
  ],
  "method": "Speaker labels are arbitrary. Labels are mapped per clip by maximum temporal overlap. Boundary matching is one-to-one within 0.5 seconds. This measures output difference, not DER or correctness without human reference.",
  "clips": {
    "mid-meeting-turns.wav": {
      "nemotron_segments": 44,
      "pyannote_segments": 45,
      "nemotron_speakers": [
        "0",
        "1",
        "2",
        "3",
        "4"
      ],
      "pyannote_speakers": [
        "SPEAKER_00",
        "SPEAKER_01",
        "SPEAKER_02",
        "SPEAKER_03",
        "SPEAKER_04"
      ],
      "label_mapping_by_maximum_temporal_overlap": {
        "4": "SPEAKER_00",
        "2": "SPEAKER_01",
        "3": "SPEAKER_02",
        "1": "SPEAKER_03",
        "0": "SPEAKER_04"
      },
      "nemotron_speech_seconds_union": 76.30999999999997,
      "pyannote_speech_seconds_union": 78.45187500000002,
      "both_active_seconds": 75.36071875000005,
      "mapped_speaker_agreement_seconds": 71.89168750000005,
      "mapped_speaker_agreement_rate_when_both_active": 0.953967646440474,
      "nemotron_only_speech_seconds": 0.9492812499999275,
      "pyannote_only_speech_seconds": 3.0911562499999787,
      "nemotron_boundary_count": 88,
      "pyannote_boundary_count": 83,
      "boundary_matches_within_0_5_seconds": 73,
      "boundary_match_rate_vs_nemotron": 0.8295454545454546,
      "boundary_match_rate_vs_pyannote": 0.8795180722891566,
      "longest_nonagreement_runs": [
        {
          "start": 17.98596875,
          "end": 20.37,
          "category": "speaker-label-disagreement"
        },
        {
          "start": 2.7815937500000003,
          "end": 3.2,
          "category": "speaker-label-disagreement"
        },
        {
          "start": 56.82,
          "end": 57.19,
          "category": "pyannote-only-speech"
        },
        {
          "start": 91.07,
          "end": 91.42,
          "category": "pyannote-only-speech"
        },
        {
          "start": 58.91,
          "end": 59.22,
          "category": "pyannote-only-speech"
        },
        {
          "start": 74.75,
          "end": 75.06,
          "category": "speaker-label-disagreement"
        },
        {
          "start": 9.82,
          "end": 10.09,
          "category": "speaker-label-disagreement"
        },
        {
          "start": 11.29,
          "end": 11.48,
          "category": "pyannote-only-speech"
        }
      ]
    },
    "opening-multispeaker.wav": {
      "nemotron_segments": 55,
      "pyannote_segments": 39,
      "nemotron_speakers": [
        "0",
        "1",
        "2",
        "3",
        "4",
        "5"
      ],
      "pyannote_speakers": [
        "SPEAKER_00",
        "SPEAKER_01",
        "SPEAKER_02",
        "SPEAKER_03",
        "SPEAKER_04"
      ],
      "label_mapping_by_maximum_temporal_overlap": {
        "1": "SPEAKER_00",
        "4": "SPEAKER_01",
        "5": "SPEAKER_02",
        "2": "SPEAKER_03",
        "0": "SPEAKER_04"
      },
      "nemotron_speech_seconds_union": 85.91999999999999,
      "pyannote_speech_seconds_union": 93.96000000000004,
      "both_active_seconds": 84.84184375000002,
      "mapped_speaker_agreement_seconds": 82.6139375,
      "mapped_speaker_agreement_rate_when_both_active": 0.9737404781470226,
      "nemotron_only_speech_seconds": 1.0781562499999486,
      "pyannote_only_speech_seconds": 9.118156250000006,
      "nemotron_boundary_count": 109,
      "pyannote_boundary_count": 75,
      "boundary_matches_within_0_5_seconds": 69,
      "boundary_match_rate_vs_nemotron": 0.6330275229357798,
      "boundary_match_rate_vs_pyannote": 0.92,
      "longest_nonagreement_runs": [
        {
          "start": 38.29,
          "end": 39.45,
          "category": "pyannote-only-speech"
        },
        {
          "start": 28.41,
          "end": 29.24159375,
          "category": "speaker-label-disagreement"
        },
        {
          "start": 53.07,
          "end": 53.88,
          "category": "pyannote-only-speech"
        },
        {
          "start": 47.53,
          "end": 48.3,
          "category": "pyannote-only-speech"
        },
        {
          "start": 30.44,
          "end": 31.08,
          "category": "speaker-label-disagreement"
        },
        {
          "start": 63.46,
          "end": 64.03784375000001,
          "category": "pyannote-only-speech"
        },
        {
          "start": 43.71,
          "end": 44.28,
          "category": "pyannote-only-speech"
        },
        {
          "start": 49.87,
          "end": 50.42,
          "category": "pyannote-only-speech"
        }
      ]
    },
    "pretrim-tail.wav": {
      "nemotron_segments": 0,
      "pyannote_segments": 0,
      "nemotron_speakers": [],
      "pyannote_speakers": [],
      "label_mapping_by_maximum_temporal_overlap": {},
      "nemotron_speech_seconds_union": 0.0,
      "pyannote_speech_seconds_union": 0.0,
      "both_active_seconds": 0.0,
      "mapped_speaker_agreement_seconds": 0.0,
      "mapped_speaker_agreement_rate_when_both_active": null,
      "nemotron_only_speech_seconds": 0.0,
      "pyannote_only_speech_seconds": 0.0,
      "nemotron_boundary_count": 0,
      "pyannote_boundary_count": 0,
      "boundary_matches_within_0_5_seconds": 0,
      "boundary_match_rate_vs_nemotron": null,
      "boundary_match_rate_vs_pyannote": null,
      "longest_nonagreement_runs": []
    },
    "signoff.wav": {
      "nemotron_segments": 39,
      "pyannote_segments": 32,
      "nemotron_speakers": [
        "0",
        "1",
        "2",
        "3",
        "4",
        "5"
      ],
      "pyannote_speakers": [
        "SPEAKER_00",
        "SPEAKER_01"
      ],
      "label_mapping_by_maximum_temporal_overlap": {
        "2": "SPEAKER_00",
        "0": "SPEAKER_01"
      },
      "nemotron_speech_seconds_union": 44.55999999999999,
      "pyannote_speech_seconds_union": 46.423125,
      "both_active_seconds": 43.516156249999995,
      "mapped_speaker_agreement_seconds": 38.54553124999999,
      "mapped_speaker_agreement_rate_when_both_active": 0.8857751826369085,
      "nemotron_only_speech_seconds": 1.043843749999991,
      "pyannote_only_speech_seconds": 2.9069687500000065,
      "nemotron_boundary_count": 77,
      "pyannote_boundary_count": 58,
      "boundary_matches_within_0_5_seconds": 54,
      "boundary_match_rate_vs_nemotron": 0.7012987012987013,
      "boundary_match_rate_vs_pyannote": 0.9310344827586207,
      "longest_nonagreement_runs": [
        {
          "start": 18.762218750000002,
          "end": 19.84,
          "category": "speaker-label-disagreement"
        },
        {
          "start": 25.782218750000002,
          "end": 26.76096875,
          "category": "speaker-label-disagreement"
        },
        {
          "start": 61.85,
          "end": 62.56,
          "category": "speaker-label-disagreement"
        },
        {
          "start": 2.09,
          "end": 2.67,
          "category": "nemotron-only-speech"
        },
        {
          "start": 21.32721875,
          "end": 21.81,
          "category": "speaker-label-disagreement"
        },
        {
          "start": 57.75,
          "end": 58.21,
          "category": "pyannote-only-speech"
        },
        {
          "start": 0.9,
          "end": 1.29,
          "category": "speaker-label-disagreement"
        },
        {
          "start": 35.61,
          "end": 35.91,
          "category": "pyannote-only-speech"
        }
      ]
    }
  }
}