metadata

task_categories:
  - token-classification
language:
  - km

KHEED Dataset Card

Summary

The Khmer Health Event Extraction Dataset (KHEED) is designed for named entity recognition in the Khmer language, specially in the health domain. The data was crawled from Khmer news websites. This dataset includes eight entity types:

Disease (DIS)
Location (LOC)
Organization (ORG)
HumanCount (HUM)
Pathogen (PAT)
Date (DAT)
Symptom (SYM)
Medication (MED)

Please note that the dataset should be used with caution, as it may contain mislabeled data due to human error. Additionally, the classes could be expanded or further refined, and some boundaries may be inaccurate due to non-linguistic expertise.

Dataset Structure

All splits of the dataset share the same data fields:

tokens: A list of string features.
ner_tags: A list of classification labels (integers).

The full tagset with indices is as follows:

{
  "O": 0,
  "B-DAT": 1,
  "I-DAT": 2,
  "B-DIS": 3,
  "I-DIS": 4,
  "B-HUM": 5,
  "I-HUM": 6,
  "B-LOC": 7,
  "I-LOC": 8,
  "B-MED": 9,
  "I-MED": 10,
  "B-ORG": 11,
  "I-ORG": 12,
  "B-PAT": 13,
  "I-PAT": 14,
  "B-SYM": 15,
  "I-SYM": 16
}