摘要:On May 31st, the "Digital Library Cave" platform developed by the Dunhuang Academy (https://cave17.e-dunhuang.com) officially laun
On May 31st, the "Digital Library Cave" platform developed by the Dunhuang Academy (https://cave17.e-dunhuang.com) officially launched, opening access to over 9,900 volumes of manuscripts and more than 60,000 images to global users. The Library Cave covers Buddhist scriptures, documents and artworks, marking a major step forward as Dunhuang culture enters the digital era.
Jointly developed by the Dunhuang Academy and Tencent, the platform allows users to perform multilingual translation, translations from classical to modern Chinese, intelligent Q&A, and AI-powered summarization by leveraging large AI models and intelligent search technologies. Take the Diamond Sutra as an example, users can not only view high-definition images but also extract its core ideas through the "AI-powered summary" function and connect them with related murals from the same period.
To address the features of ancient texts such as traditional Chinese characters, lack of punctuation, and vertical layout, the platform incorporates a specialized dictionary for ancient texts and an optimized segmentation system to improve AI accuracy. According to abig data expert from Tencent Cloud, the platform adopts a dual-model driven RAG (Retrieval-Augmented Generation) architecture and integrates two leading domestic large models Tencent Hunyuan and DeepSeek, both of which rank highly on the global Chatbot Arena list for their advanced reasoning and retrieval capabilities. Tested by Dunhuang experts, the system achieves up to 95% accuracy in interpreting historical manuscripts.
Empowered by cutting-edge technologies such as AI, the global release of the "Digital Library Cave" will showcase the achievements of the "Digital Restoration Project for the Dunhuang Relics Scattered Overseas" to offer the international academic community and the public an authoritative, accurate, convenient and intelligent international public cultural service platform.
Source: Lingnan on the Cloud
AI陪你逛敦煌千年数字图书馆
5月31日,敦煌“数字藏经洞”数据库平台(https://cave17.e-dunhuang.com)正式上线,向全球用户开放9900多卷敦煌文书和逾6万幅图像资源,内容涵盖佛经、律典、契约、绢画等,敦煌文化迈入“云端时代”。
该平台由敦煌研究院与腾讯联合打造,借助AI大模型与智能检索技术,用户可实现古籍多语言翻译、文白互译、智能问答、AI归纳等操作。以《金刚经》为例,用户不仅可浏览高清图像,还能通过“AI归纳”功能提取核心思想,进一步关联同期壁画内容。
针对繁体、无标点、竖排等古籍特点,平台引入“古籍专属词典”并优化分词系统,提升AI处理准确性。腾讯云大数据专家介绍,平台采用“双模驱动”的RAG(检索增强生成)技术架构,融合混元与DeepSeek两大国产大模型,在Chatbot Arena全球榜单中排名靠前,具备强大推理与搜索能力。经敦煌专家测试,该系统在解读文献时准确率高达95%。
借助AI等前沿技术,“数字藏经洞”数据库平台全球上线,也将集中展示“流失海外敦煌文物数字化复原项目”成果,为全球学术界和社会公众了解、获取敦煌文化艺术资源贡献一项权威准确、便捷智能的国际性公共文化服务产品。
文 | 记者 王丹阳
图|受访者供图
译 | 盛嘉
英文审校|林佳岱
来源:羊城派