DIWALI -Diversity and Inclusivity aWare cuLture specific Items for India
收藏资源简介:
DIWALI是一个涵盖印度文化多样性的高质量数据集,包含了来自印度36个子地区的约8k个文化概念,涵盖了17个文化层面。该数据集的创建旨在为文化文本适应任务提供评估指标,并解决现有数据集在文化概念表示上的不足。数据集的内容包括从GPT-4和网页搜索中提取的文化概念,并经过两阶段的验证过程确保其可靠性和文化有效性。DIWALI数据集为文化文本适应任务提供了宝贵的资源,有助于评估大型语言模型的文化适应能力,并促进跨文化交际领域的进一步研究。
DIWALI is a high-quality dataset encompassing the cultural diversity of India. It contains approximately 8,000 cultural concepts sourced from 36 sub-regions across the country, covering 17 cultural dimensions. Developed to provide evaluation metrics for cultural text adaptation tasks and address the shortcomings of existing datasets in representing cultural concepts, the dataset consists of cultural concepts extracted from GPT-4 and web searches, and has undergone a two-stage verification process to ensure its reliability and cultural validity. As a valuable resource for cultural text adaptation tasks, the DIWALI dataset aids in evaluating the cultural adaptation capabilities of large language models and promotes further research in the field of cross-cultural communication.
DIWALI 数据集概述
数据集简介
DIWALI(Diversity and Inclusivity aWare cuLture specific Items for India)是一个针对印度文化的文化特定项目(CSIs)数据集,专注于文化文本适应在印度语境中的应用。
核心特征
- 文化概念数量:8817个
- 文化国家:印度
- 覆盖子区域:36个(印度境内)
- 文化背景:区域和国家特定
文化维度覆盖
数据集涵盖17个文化维度:食物、舞蹈、节日、姓名、珠宝、地点、传统、语言、服装、游戏、仪式、建筑、饮料、艺术、纺织品、宗教和邦。
引用信息
bibtex @misc{sahoo2025diwalidiversityinclusivity, title={DIWALI - Diversity and Inclusivity aWare cuLture specific Items for India: Dataset and Assessment of LLMs for Cultural Text Adaptation in Indian Context}, author={Pramit Sahoo and Maharaj Brahma and Maunendra Sankar Desarkar}, year={2025}, eprint={2509.17399}, archivePrefix={arXiv}, primaryClass={cs.CL}, url={https://arxiv.org/abs/2509.17399}, }
联系方式
- Maharaj Brahma:cs23resch01004@iith.ac.in
- Pramit Sahoo:pramitsahoo.gnipst@gmail.com
相关资源
- 论文地址:https://arxiv.org/abs/2509.17399
- 数据集地址:https://huggingface.co/datasets/nlip/DIWALI
- 博客地址:https://nlip-lab.github.io/nlip/publications/diwali/




