splits/split_summary.json

{
  "input_file": "dialog_1-1152.xlsx (source workbook not included)",
  "input_sha256": "24c359d2d5135b2a420ab42aaf87b2b3c8175a1fcd8f0f6cf2d92f70239c8562",
  "selected_sheet": "dialog_all",
  "image_root": "not distributed (copyright-restricted)",
  "dataset_fingerprint": "c7c87855c0c7a15c468dc93076be3dc938c14596d2045a2274d7a010c375e0e0",
  "split_seed": 42,
  "split_method": "exact_dialog_subset_sum",
  "requested_label_targets": {
    "train": {
      "negative": 856,
      "neutral": 845,
      "positive": 661
    },
    "validation": {
      "negative": 107,
      "neutral": 106,
      "positive": 82
    },
    "test": {
      "negative": 107,
      "neutral": 106,
      "positive": 82
    }
  },
  "total_samples": 2952,
  "total_dialogs": 1152,
  "extra_image_count": 0,
  "splits": {
    "train": {
      "samples": 2362,
      "dialogs": 931,
      "label_counts": {
        "negative": 856,
        "neutral": 845,
        "positive": 661
      }
    },
    "validation": {
      "samples": 295,
      "dialogs": 113,
      "label_counts": {
        "negative": 107,
        "neutral": 106,
        "positive": 82
      }
    },
    "test": {
      "samples": 295,
      "dialogs": 108,
      "label_counts": {
        "negative": 107,
        "neutral": 106,
        "positive": 82
      }
    }
  },
  "release_scope": "original Tibetan/Chinese text, sentiment labels, and split metadata"
}