What this matrix really documents is not a single artifact but a transformation: the slow shift in how epigraphers have produced, preserved, and published their evidence over five centuries. Every page below instantiates a particular strategy of documentation and publication that worked at some moment and was eventually superseded, or absorbed, by the next. Ligorio's 1500s manuscript catalogue, Mommsen's 1847 master plan for CIL, the 1932 Leiden conventions, the 1990s digital databases, EpiDoc's 2000s standardisation, EFES's 2017 turnkey publishing, the 2021 reproducible Journal of Digital History paper, and the 2022 / 2025 neural-network experiments each represent a different answer to the same recurring question: how do you record a stone, share that record with other scholars, and let later readers test and revise it?
本矩阵真正记录的并非任何单一器物，而是一种转变：五个世纪以来铭文学者制作、保存与发布证据方式的缓慢更迭。下方每一页都体现了某一时期行得通、终被后来者所取代或吸纳的文献记录与出版策略。利戈里奥 16 世纪的手稿目录、蒙森 1847 年为 CIL 拟定的总规划、1932 年的莱顿规约、1990 年代的数字数据库、2000 年代的 EpiDoc 标准化、2017 年 EFES 的交钥匙出版、2021 年可复现的《数字史》论文、2022 / 2025 年的神经网络实验，每一者都是对同一反复出现之问题的不同回答：如何记录一块石头、如何把记录传给其他学者、又如何让后人能够检验并修订它？
The four clusters mark four moments in that shift. The Primary sources cluster shows pre-print and early-print strategies: Ligorio's hand-drawn manuscript catalogue (with its characteristic forgeries) and Mommsen's plan for a state-academy print corpus that would standardise editorial criticism. The SDAM ETL suite cluster shows what those strategies look like when ported to the early 21st century: open-access databases, bilingual portals, code-level reproducibility, a Journal of Digital History paper that publishes its data and pipeline alongside its argument. The EpiDoc reference & editors cluster gives you the encoding standard that makes today's modern publications interchangeable across projects, plus two editors to try the encoding yourself. The Chinese entry-point page provides the first such map of the field for readers who would otherwise have no Chinese-language survey of this territory.
四个群组对应转变中的四个时刻。原典语料库与目录群组展示前印本与早期印本的策略：利戈里奥手绘的手稿目录（连同其典型的伪刻），与蒙森为国家学院级印本语料拟定、欲将编辑批判标准化的规划。SDAM ETL 套件群组展示这些策略移植到 21 世纪初的形态：开放访问的数据库、双语入口、代码级可复现、一篇连数据与流水线一并发表的《数字史》论文。EpiDoc 参考与编辑器群组提供使当代出版可跨项目互换的编码标准，并附两款可亲自尝试的编辑器。中文入口一页为本无中文综述的读者提供该领域的第一张地图。
Notice what each transition gains and loses. Ligorio could draw a letterform with the eye of an artist, sometimes more eloquently than any photograph, but he could also fabricate an inscription wholesale; Mommsen's CIL accepted neither risk and traded visual fidelity for editorial rigour. Print corpora could fix a text but could not link it to its parallels; digital databases multiplied parallels but ended up disagreeing on identifiers, dates, and even readings of the same stone. EpiDoc's XML solved the cross-database identification problem at the cost of asking every project to learn a schema; AI restorers trained on the resulting corpora can suggest text supplements at scale but inherit the editorial decisions baked into their training data. Each cluster in this matrix lets you inspect one of these trade-offs in detail, and each page below documents the workflow, the assumptions, and the publishing practice that a given strategy entails.
留心每次转变所得与所失。利戈里奥能以艺术家之眼勾勒字形 ， 时或比任何照片都更传神 ， 却也可整篇杜撰一块铭文；蒙森的 CIL 不愿承担二者之险，遂以视觉传神之失换取编辑之严谨。印本语料能固定文本，却无法链接其平行例；数字数据库使平行例倍增，却在同一块石头的标识、年代、甚至读法上各执一词。EpiDoc 的 XML 以要求每个项目学习一套模式 (schema) 为代价，解决了跨数据库识别问题；以这些语料训练的 AI 修复器能在规模上提出补字建议，却继承了训练数据中已固化的编辑判断。本矩阵每个群组都让你详细察验其中一项权衡，每一页都记录某一策略所涉的工作流、假设与出版实践。
One case study runs through the matrix as a worked example. A small bilingual stonecutter's sign from Roman Palermo, CIL X 7296 = IG XIV 297 = EDR140617 = EDCS-22000882 = PHI 175744 = ISic000470, happens to have travelled the full distance from manuscript through print to five digital databases and a modern EpiDoc edition. Eight of the pages name it; the rest provide the scaffolding (institutional, methodological, technical) that lets you read those eight more carefully. Treat it as the matrix's diagnostic specimen, not its organising centre. Other case studies would do as well: an Aphrodisias inscription would foreground the early EpiDoc-Aphrodisias pilot strategy; a Telamon stone would foreground EFES-based publishing for Greek inscriptions in Bulgaria. We picked CIL X 7296 because it sits at the densest crossroads of the strategies above, not because it is the only road through them.
有一项案例研究 (case study) 作为范例贯穿全矩阵。一块来自罗马时期巴勒莫 (Palermo) 的小型双语石匠招牌 ， CIL X 7296 = IG XIV 297 = EDR140617 = EDCS-22000882 = PHI 175744 = ISic000470 ， 恰好走完从手稿到印本、再到五个数字数据库与一份现代 EpiDoc 版本的全程。十四页中有八页指名提及；其余则提供使这八页可被更细读的脚手架（机构的、方法论的、技术的）。请将它视为矩阵的诊断样本 (diagnostic specimen)，而非组织中枢。其他案例研究同样可行：一块阿芙洛狄西亚 (Aphrodisias) 铭文将凸显早期 EpiDoc-Aphrodisias 试点策略；一块 Telamon 石头将凸显基于 EFES 的保加利亚希腊铭文出版策略。我们选择 CIL X 7296 不是因为它是穿越上述策略的唯一道路，而是因为它恰好处在最密集的路口。
Below: cluster cards for the four documentation/publication moments; a 15-week course alignment that pairs each strategy with its modern teaching context; five reading paths for different goals; a clickable concept index that surfaces every recurring term; a 12-row theme matrix showing where each editorial or technical theme is treated; and a relationship graph that visualises the cross-cluster bridges.
下方依次为：四个文献记录与出版时刻的群组卡片；将每种策略与现代教学情境配对的 15 周课程对应表；面向不同目标的五条阅读路径；可点击的概念索引；展示每个编辑或技术主题归属的 12 行主题矩阵；以及把跨群组桥梁可视化的关系图。