Datasets:

Modalities:
Text
Formats:
parquet
Languages:
Tagalog
Size:
< 1K
ArXiv:
Libraries:
Datasets
pandas
License:
kalahi / README.md
raileymontalan's picture
Update README.md
bfe9c48 verified
metadata
license: cc-by-4.0
license_name: cc-by-4.0
license_link: LICENSE
language:
  - tl

Kalahi

Kalahi is a high-quality, manually-crafted cultural dataset that is part of SEA-HELM. It was collaboratively created by native Filipino speakers and designed to determine LLMs’ abilities to provide relevant responses to culturally-specific situations that Filipinos face in their day-to-day lives.

Dataset Details

Kalahi is composed of 150 situationally-enriched prompts and culturally relevant and irrelevant responses that cover shared Filipino cultural knowledge and values. The cultural topics covered in Kalahi are:

Cultural Topic # of prompts
beauty and clothing 16
beliefs and practices 4
career and livelihood 20
communication and body language 5
dating and courtship 6
family and marriage 16
food and gatherings 18
friendship 7
health and wellness 13
local know-how 19
social etiquette 26

Limitations

While Kalahi is the result of the consensus views of the involved native Filipino speakers, the Filipino culture in this study refers only to cultural values acquired by Filipino speakers who were born and grew up in or at least spent most of their lives in Metro Manila. Individuals who have had different upbringings may have different perspectives on Filipino culture, such that the consensus view arrived at in this study does not fully represent the opinions of all Filipino individuals. Additionally, while Kalahi is designed to accurately represent Filipino culture, it is not intended to encompass all possible aspects of Filipino culture.

License

This dataset is made available under Creative Commons Attribution 4.0 International (CC BY 4.0).

References

@misc{montalan2024kalahihandcraftedgrassrootscultural,
      title={Kalahi: A handcrafted, grassroots cultural LLM evaluation suite for Filipino}, 
      author={Jann Railey Montalan and Jian Gang Ngui and Wei Qi Leong and Yosephine Susanto and Hamsawardhini Rengarajan and William Chandra Tjhi and Alham Fikri Aji},
      year={2024},
      eprint={2409.15380},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2409.15380}, 
}