entity_tokens.json

{
  "entity_tokens": [
    "<|B_DT|>",
    "<|B_LC|>",
    "<|B_OG|>",
    "<|B_PS|>",
    "<|B_QT|>",
    "<|B_TI|>",
    "<|E_DT|>",
    "<|E_LC|>",
    "<|E_OG|>",
    "<|E_PS|>",
    "<|E_QT|>",
    "<|E_TI|>"
  ],
  "num_tokens": 12,
  "conversion_info": {
    "original_format": "<entity:label> (with malformed cases)",
    "converted_format": "<|B_LABEL|>entity<|E_LABEL|> (normalized)",
    "source_dataset": "klue/klue (ner subset)",
    "improvements": [
      "Only valid NER labels (DT, LC, OG, PS, QT, TI) are used",
      "Malformed labels are normalized",
      "Entity content is preserved exactly",
      "Duplicate label prefixes are removed"
    ]
  },
  "valid_labels": [
    "DT",
    "LC",
    "OG",
    "PS",
    "QT",
    "TI"
  ]
}