📄 中文摘要
Nemotron-Personas-新加坡项目专注于为新加坡的主权AI战略协同设计数据。该项目旨在解决在开发主权AI模型时，数据集可能存在的文化、语言和地域偏差问题。通过与新加坡本地专家和机构紧密合作，项目团队共同构建了一个高度定制化和代表性的数据集，以确保训练出的AI模型能够准确理解并反映新加坡独特的社会文化语境。数据集的协同设计过程涉及多个阶段，包括需求分析、数据收集、标注规范制定、数据清洗与验证以及持续迭代优化。在数据收集阶段，项目特别关注了新加坡多民族、多语言的特点，力求覆盖不同族裔、语言群体的表达习惯和知识体系。标注规范的制定也充分考虑了本地化的细微差别，以避免引入外部偏见。数据清洗和验证环节则利用了先进的AI技术和人工审核相结合的方式，确保数据质量和一致性。最终形成的数据集将用于训练定制化的Nemotron大型语言模型，使其在处理与新加坡相关的任务时，能够展现出更高的准确性、相关性和文化敏感性。该项目的成功实施将为新加坡构建自主可控、符合本地需求的主权AI能力奠定坚实基础，并为其他国家或地区在发展主权AI方面提供宝贵的经验和方法论。此外，项目还探索了数据隐私和伦理问题，确保在数据协同设计过程中，严格遵守当地的数据保护法规和伦理准则，构建可信赖的AI系统。
📄 English Summary
Nemotron-Personas-Singapore: Co-Designed Data for Sovereign AI
The Nemotron-Personas-Singapore project is dedicated to co-designing data for Singapore's sovereign AI strategy. This initiative addresses the challenges of cultural, linguistic, and geographical biases that can arise in datasets used for developing sovereign AI models. By collaborating closely with local Singaporean experts and institutions, the project team collaboratively constructs a highly customized and representative dataset. This ensures that the trained AI models accurately comprehend and reflect Singapore's unique socio-cultural context. The co-design process for the dataset involves multiple stages, including requirements analysis, data collection, annotation guideline formulation, data cleaning and validation, and continuous iterative optimization. During the data collection phase, the project specifically emphasizes Singapore's multi-ethnic and multi-lingual characteristics, striving to cover the expression patterns and knowledge systems of diverse ethnic and linguistic groups. The formulation of annotation guidelines also fully considers localized nuances to avoid introducing external biases. Data cleaning and validation leverage a combination of advanced AI techniques and human review to ensure data quality and consistency. The resulting dataset will be used to train customized Nemotron large language models, enabling them to exhibit higher accuracy, relevance, and cultural sensitivity when handling Singapore-specific tasks. The successful implementation of this project will lay a solid foundation for Singapore to build autonomous, controllable, and locally relevant sovereign AI capabilities, and will provide valuable experience and methodology for other countries or regions in developing sovereign AI. Furthermore, the project explores data privacy and ethical considerations, ensuring strict adherence to local data protection regulations and ethical guidelines throughout the data co-design process, thereby building trustworthy AI systems.