README.md

metadata

task_categories:
  - token-classification
language:
  - vi

Dataset Card for BKEE

1. Dataset Summary

BKEE is the first Vietnamese Event Extraction dataset, containing 1,066 fully annotated documents covering:

• 33+ event types
• 28 argument roles
• Entity mentions, event mentions, and relation mentions

This resource addresses the lack of Vietnamese‑specific EE datasets and supports tasks from event detection to argument extraction.

2. Supported Tasks and Metrics

• Tasks

  • Entity Mention Detection
  • Event Detection
  • Event Argument Extraction

• Metrics

  • Precision / Recall / F1 (per subtask)
  • Overall document‑level accuracy

3. Languages

• Vietnamese

4. Dataset Structure

All splits have been merged into one CSV. Each row corresponds to one sentence mention:

Column	Type	Description
doc_id	string	Unique document identifier (e.g. “train-00001”).
sent_id	string	Unique sentence identifier within a document.
tokens	list	Tokenized words.
sentence	string	Original sentence text.
event_types	list	List of event type labels occurring in this sentence.
argument_roles	list	List of argument-role labels aligned to tokens.
entity_mentions	list	List of entity mention spans (start, end, label).
event_mentions	list	List of event trigger spans (start, end, label).
relation_mentions	list	List of relations between arguments and events.
type	string	Split: train / dev / test.
dataset	string	Always BKEE (for provenance).

5. Data Fields

• doc_id, sent_id (str): Document and sentence IDs.
• tokens (List[str]): Tokenized text.
• sentence (str): Raw sentence.
• event_types, argument_roles, entity_mentions, event_mentions, relation_mentions: Annotations per sentence.
• type (str): Data split.
• dataset (str): Always BKEE.

6. Usage

from datasets import load_dataset

ds = load_dataset("visolex/BKEE")

# Filter by split
train = ds.filter(lambda ex: ex["type"] == "train")
val   = ds.filter(lambda ex: ex["type"] == "dev")
test  = ds.filter(lambda ex: ex["type"] == "test")

# Inspect one example
print(train[0])

7. Source & Links

• GitHub (original data & code): https://github.com/nhungnt/BKEE ([github.com][2])
• Paper: Thi‑Nhung Nguyen et al. (2024), “BKEE: Pioneering Event Extraction in the Vietnamese Language,” LREC‑COLING 2024. ([aclanthology.org][1])
• Hugging Face (this unified version): https://huggingface.co/datasets/visolex/BKEE

8. License

Released under CC BY‑NC 4.0. See the GitHub repo for full terms.

9. Citation

@inproceedings{nguyen-etal-2024-bkee,
  title     = {BKEE: Pioneering Event Extraction in the Vietnamese Language},
  author    = {Nguyen, Thi-Nhung and Tran, Bang Tien and Luu, Trong-Nghia and 
               Nguyen, Thien Huu and Nguyen, Kiem-Hieu},
  booktitle = {Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)},
  month     = may,
  year      = {2024},
  address   = {Torino, Italia},
  publisher = {ELRA and ICCL},
  url       = {https://aclanthology.org/2024.lrec-main.217},
  pages     = {2421--2427}
}

@misc{nhungnt_bkee,
  title        = {BKEE: Pioneering Event Extraction in the Vietnamese Language},
  author       = {{nhungnt}},
  howpublished = {\url{https://github.com/nhungnt/BKEE}},
  year         = {2024}
}