非遗的数智化：从建档留底，到把家底用起来

《非物质文化遗产保护传承「十五五」规划》（文化和旅游部，2026 年 9 月印发）把深化科技应用、提升数智化保护水平列为六个方面重点任务之一。过去二十年攒下的普查资料、传承人记录、各级数据库，如今被赋予新的角色。从「数字化」到「数智化」，一字之差，工作重心从留底挪向了使用。

▍一、先把两个词分开

文件里的「数智化」不是修辞。数字化解决的是留存：普查的文字、访谈的音视频、技艺的照片，转成数字资料存下来。数智化在留存之上加了一层利用：用统一的标准组织这些资料，让它们能被检索、被比对、被机器理解。据新华日报 2026 年 11 月 2 日的解读，前者聚焦把文化资源留下来，后者探索如何读懂、用好这些资源。

这个转变有个现实的起点。多年积累下来，各级非遗数据库陆续建成，但据上述解读梳理，一些数字资源分散在不同部门和平台，采集与著录标准不尽相同，彼此难以衔接；部分档案完成采集后开发不足，主要发挥存储与展示功能。家底有了，账本还没通。

▍二、文件提出的三个落点

统一标准。规划提出统一数据库元数据和资源采集著录规范，推动不同地区、不同门类的非遗数据采用相对统一的标准，提升资源的关联性和共享利用水平，减少重复采集和建设。这是给「账本」定格式：格式通了，数据才能互相认识。

建语料库。规划提出「建立高质量非遗语料库」。据新华日报解读，随着大模型广泛应用，人工智能开始参与非遗知识的检索、解读与传播，但缺少权威、系统的专业资料支撑时，容易出现知识混淆与事实错误；语料库正是要为智能技术提供准确可靠的知识基础。

场景惠民。规划把「丰富科技应用场景，推动成果惠民」作为任务。落到例子上，据新华日报报道，江苏已有紫砂数字化溯源的探索，建立「一壶一码一图一报告」机制，消费者扫码可查询泥料产地、制作工艺与作者资质。数智化的终点，是普通用户也能用到。

资料室里成排的档案盒，一侧的屏幕上显示整理中的图像

▍三、对手艺人意味着什么

短期看，数智化首先是档案的公平。过去口述的技艺、靠师徒记忆保存的规矩，进了统一格式的数据库，就有了被检索与引用的位置；记录做得扎实，一门小众手艺就不必靠运气被看见。规划同时安排了非遗传承人记录等工作，记录本身在升格。

长期看，语料库与溯源体系是市场基建。知识可靠，AI 搜索引擎摘录非遗内容时才不容易张冠李戴；溯源可查，买家对「手工真作」的信任才有落点。对做出海的品牌来说，这两件事都直接相关：海外用户先在搜索里认识这门手艺，再在溯源里确认这件器物。

《非物质文化遗产保护传承「十五五」规划》提出，到 2030 年初步形成非遗系统性保护新格局（文化和旅游部，2026）。

▍问答

问｜非遗数字化和数智化有什么区别

答｜数字化是把普查、记录、影像转成数字资料存下来，解决「留下来」；数智化是在这之上用统一标准组织数据，让资源可检索、可关联、可被智能技术利用，解决「用起来」。《非物质文化遗产保护传承「十五五」规划》（文化和旅游部，2026）将后者列为重点任务。

问｜建非遗语料库对普通人有什么影响

答｜最直接的影响是搜索与问答的准确度。AI 助手回答非遗问题时，若有权威语料支撑，张冠李戴的概率会下降；对买家而言，与数字溯源结合后，扫码核对一件器物的工序与作者资质会成为常规动作。据新华日报 2026 年 11 月 2 日解读，语料库正是为智能技术提供准确可靠的知识基础。

问｜韵和鸣在这件事里做什么

答｜做内容这一侧的活。品牌长期做匠人访谈与工序记录，帮老师把产品与手艺拍下来、写清楚，再用中英双语与多语言直播把这些内容讲给国内外的用户听。国家的语料库解决「知识底座」，品牌解决「把一门具体的手艺讲给具体的人听」。

📩 非遗项目｜匠人举荐｜跨境出海｜线下联营

欢迎留言或后台联系，我们与匠人同行。

—— 韵和鸣 · 与匠人同行

〖国〗

English

The "digital-intelligent" in the document is not rhetoric. Digitalisation answers preservation: survey texts, interview recordings, photographs of technique, stored as digital files. Digital intelligence adds a layer of use on top: organising the material under common standards so it can be searched, compared and understood by machines. As Xinhua Daily explained on 2 November 2026, the former focuses on keeping cultural resources, the latter on reading and using them.

The shift has a practical starting point. Databases at all levels have been built up, yet, as that account sorts out, some resources sit scattered across departments and platforms under differing collection and description standards, hard to join; some archives, once digitised, see little further development and serve mainly as storage and display. The stock exists; the ledger is not yet unified.

Unified standards. The plan calls for common metadata and description norms, moving data from different regions and categories toward shared standards to raise linkage and shared use, and to cut duplicate collection. This sets the format of the ledger: only when formats agree can data recognise one another.

A corpus. The plan proposes "building a high-quality heritage corpus". As the same account notes, with large models in wide use, artificial intelligence now takes part in retrieving, interpreting and spreading heritage knowledge, but without authoritative, systematic source material it readily mixes up facts. The corpus is meant to give the machines an accurate foundation to draw on.

Scenarios for people. The plan sets "enriching application scenarios and bringing the results to the public" as a task. In practice, Xinhua Daily reports, Yixing has explored digital traceability for purple-clay ware, with a "one pot, one code, one report" mechanism letting buyers scan for clay source, process and maker credentials. The end point of digital intelligence is that ordinary users can touch it.

A records room of boxed archives, a screen nearby showing images being catalogued

In the short run, digital intelligence is first an archive-level playing field. Techniques once oral, rules once kept in a master-apprentice memory, gain a retrievable, citable place in a database of common format; where documentation is solid, a niche craft no longer depends on luck to be seen. The plan also schedules bearer documentation, and documentation itself is being upgraded.

In the long run, the corpus and traceability are market infrastructure. With reliable knowledge, AI search engines are less prone to mangle heritage content when quoting it; with checkable provenance, buyer trust in authentic handwork has somewhere to land. For an export brand both matter directly: overseas users first meet the craft in search, then confirm the piece through traceability.

What is the difference between digitising and going digital-intelligent?

Digitisation converts surveys, records and images into stored files, answering "keep it"; digital intelligence organises the data under common standards so it can be searched, linked and used by intelligent systems, answering "use it". The ministry's 2026 plan lists the latter as a key task.

What does a heritage corpus mean for ordinary users?

Most directly, accuracy in search and answers. Backed by an authoritative corpus, AI assistants are less likely to mix facts when answering heritage questions; for buyers, combined with digital traceability, scanning a piece for its process and maker becomes routine. As Xinhua Daily's account of 2 November 2026 notes, the corpus exists precisely to give intelligent technology a reliable foundation.

Where does YUNHEMING fit in?

The content side. The brand keeps documenting makers and process, filming products and craft for the masters and writing them out clearly, then telling it to audiences at home and abroad in bilingual and multilingual form. The national corpus supplies the knowledge base; the brand translates one specific craft for one specific listener.

This is a personal reading; refer to the official text. Have you come across a heritage product you could scan for its craft record? Tell us in the comments.

ICH Projects · Artisan Referrals · Global Distribution · Salon Partnerships