dataset_manifest.json

{
  "created_at_utc": "2026-07-22T11:23:46.879214Z",
  "records": 2495,
  "unique_source_words": 732,
  "error_type_counts": {
    "extra_letter": 470,
    "missing_diacritic": 132,
    "case_ending_error": 505,
    "c_q_confusion": 51,
    "missing_letter": 493,
    "letter_transposition": 588,
    "digraph_reduction": 228,
    "drop_final_vowel": 28
  },
  "covered_digraphs": [
    "dh",
    "gj",
    "ll",
    "nj",
    "rr",
    "sh",
    "th",
    "xh",
    "zh"
  ],
  "format": "jsonl_and_csv",
  "files": {
    "jsonl": "/Users/Apple/Desktop/Phd 08:02:2026/backend/huggingface_publish/dataset_albanian_error_augmentation/train.jsonl",
    "csv": "/Users/Apple/Desktop/Phd 08:02:2026/backend/huggingface_publish/dataset_albanian_error_augmentation/train.csv"
  },
  "description": "Controlled Albanian orthographic error augmentation pairs (incorrect -> correct) for spelling correction and GEC research."
}