metadata

pretty_name: SEA Abstractive Summarization
license:
  - cc-by-nc-sa-4.0
task_categories:
  - text-generation
language:
  - id
  - ta
  - th
  - vi
dataset_info:
  features:
    - name: id
      dtype: string
    - name: label
      dtype: string
    - name: prompts
      list:
        - name: text
          dtype: string
    - name: prompt_templates
      sequence: string
    - name: metadata
      struct:
        - name: language
          dtype: string
        - name: url
          dtype: string
        - name: title
          dtype: string
  splits:
    - name: id
      num_bytes: 322112
      num_examples: 100
    - name: id_fewshot
      num_bytes: 5963
      num_examples: 5
    - name: ta
      num_bytes: 1075514
      num_examples: 100
    - name: ta_fewshot
      num_bytes: 10198
      num_examples: 5
    - name: th
      num_bytes: 1201794
      num_examples: 100
    - name: th_fewshot
      num_bytes: 8735
      num_examples: 5
    - name: vi
      num_bytes: 395697
      num_examples: 100
    - name: vi_fewshot
      num_bytes: 9092
      num_examples: 5
  download_size: 1258846
  dataset_size: 3029105
configs:
  - config_name: default
    data_files:
      - split: id
        path: data/id-*
      - split: id_fewshot
        path: data/id_fewshot-*
      - split: ta
        path: data/ta-*
      - split: ta_fewshot
        path: data/ta_fewshot-*
      - split: th
        path: data/th-*
      - split: th_fewshot
        path: data/th_fewshot-*
      - split: vi
        path: data/vi-*
      - split: vi_fewshot
        path: data/vi_fewshot-*
size_categories:
  - n<1K

SEA Abstractive Summarization

SEA Abstractive Summarization evaluates a model's ability to read a document, identify the key points within, and summarize them into a coherent and fluent text while paraphrasing the document. It is sampled from XL-Sum for Indonesian, Tamil, Thai, and Vietnamese.

Supported Tasks and Leaderboards

SEA Abstractive Summarization is designed for evaluating chat or instruction-tuned large language models (LLMs). It is part of the SEA-HELM leaderboard from AI Singapore.

Languages

Indonesian (id)
Tamil (ta)
Thai (th)
Vietnamese (vi)

Dataset Details

SEA Abstractive Summarization is split by language, with additional splits containing fewshot examples. Below are the statistics for this dataset. The number of tokens only refer to the strings of text found within the prompts column.

Split	# of examples	# of GPT-4o tokens	# of Gemma 2 tokens	# of Llama 3 tokens
id	100	61628	55485	77016
ta	100	114275	156476	457559
th	100	155203	151988	176985
vi	100	86305	78285	82269
id_fewshot	5	1124	1050	1430
ta_fewshot	5	964	1339	3905
th_fewshot	5	925	869	1062
vi_fewshot	5	2396	2170	2282
total	420	422820	447662	802508

Data Sources

Data Source	License	Language/s	Split/s
XL-Sum	CC BY-NC-SA 4.0	Indonesian, Tamil, Thai, Vietnamese	id, id_fewshot, ta, ta_fewshot, th, th_fewshot, vi, vi_fewshot

License

For the license/s of the dataset/s, please refer to the data sources table above.

We endeavor to ensure data used is permissible and have chosen datasets from creators who have processes to exclude copyrighted or disputed data.

References

@inproceedings{hasan-etal-2021-xl,
    title = "{XL}-Sum: Large-Scale Multilingual Abstractive Summarization for 44 Languages",
    author = "Hasan, Tahmid  and
      Bhattacharjee, Abhik  and
      Islam, Md. Saiful  and
      Mubasshir, Kazi  and
      Li, Yuan-Fang  and
      Kang, Yong-Bin  and
      Rahman, M. Sohel  and
      Shahriyar, Rifat",
    booktitle = "Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021",
    month = aug,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.findings-acl.413",
    pages = "4693--4703",
}

@misc{leong2023bhasaholisticsoutheastasian,
      title={BHASA: A Holistic Southeast Asian Linguistic and Cultural Evaluation Suite for Large Language Models}, 
      author={Wei Qi Leong and Jian Gang Ngui and Yosephine Susanto and Hamsawardhini Rengarajan and Kengatharaiyer Sarveswaran and William Chandra Tjhi},
      year={2023},
      eprint={2309.06085},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2309.06085}, 
}