Records with ≤ 20% restored text. Closest to original-state evidence; supplements are minimal.补字 ≤ 20% 之文书。最接原状之证，增补极少。
Corpus at a glance语料库概况
Distribution by institution按机构分布
Top formulae by attestation按出现次数排序的公式
Formula frequency over time公式在时间维度的分布
Decree-class inscriptions only (n=). Filter by slot or language; hover a bar segment for exact counts.仅显示法令类铭文（共 件）. 可按位置或语言过滤；鼠标悬停可见精确数值.
Provenance map出土地图
Geocoded decree-class inscriptions (n=). Greek (blue), Latin (terracotta), bridge inscriptions (purple). Click a marker for inscription detail.已定位的法令类铭文（共 件）. 希腊语（蓝色），拉丁语（赭色），桥接铭文（紫色；指同碑并存希腊拉丁两套公式、连接两个系统的铭文）. 点击标记查看详情.
Bridge inscriptions only仅桥接铭文
Institutional hierarchy机构层级
Greek civic (boule + demos / boule / demos by polity) · Roman (senatus consultum / decretum decurionum by era) · Other / unclassified. Click ▶ to expand. Click any node to see its formula bundle, century distribution, and sample inscriptions.希腊市民（boule + demos / boule / demos 按城邦） · 罗马（元老院决议 / 市政决议 按时代） · 其他 / 未分类. 点击 ▶ 展开. 点击任意节点查看其公式集、年代分布和示例铭文.
Tree树状结构
Selected node detail选中节点详情
Click a node in the tree.请在左侧树中点击一个节点.
Diff two nodes对比两个节点
Cross-influence network跨语言转译网络
Force-directed graph of formula relations. Purple thick edges = explicit Greek↔Latin calque pairs. Terracotta edges = co-occurrence in bridge inscriptions (≥2 bridges). Node size reflects total attestations; colour reflects language.公式关系力导向图. 紫色粗线 = 希腊↔拉丁明确转译对. 赭色线 = 桥接铭文共现（≥2件）. 节点大小反映出现次数；颜色反映语言.
What distance means: the layout is force-directed and reflects an aesthetic balance, not a literal corpus metric. Linked nodes are pulled together by edge-springs (rest length = 90px for calque / loan-translation edges, 140px for structural-parallel edges); all nodes repel each other (charge −260). Outer-orbit isolates: unconnected nodes (edge degree 0) ride a radial force at 0.46 × min(W,H) from centre with strength 0.55, so they sit on the OUTER ring of the layout — outside the dense connected cluster, never crowding into it. Weakly-linked nodes (degree 1) sit on a mid-ring (0.32 × min(W,H), strength 0.22); well-connected nodes (degree ≥ 2) are free in the centre with only mild centring (strength 0.03). Closer pairs in the centre therefore share a stronger formal-evidence link, but absolute distance remains an aesthetic balance, not a measured similarity. Drag any node to reposition; the simulation will re-balance. Labels: only formulae with ≥ 3 attestations show their name by default; hover any unlabeled node to reveal it. Zoom + pan: wheel zoom, click-drag pan (the chart uses d3-zoom). 距离之意： 此为力导布局，距离反映美学平衡，非语料度量。相连之节点由边之弹簧相牵（calque/借译边静止长度 90 像素，结构平行边 140 像素）；所有节点彼此相斥（电荷 −260）。外环之孤者： 无边相连之节点（边数 0）受径向力牵于 0.46 × min(W,H) 处，强度 0.55，故居布局之最外环，出主簇密集区之外，不挤其内。仅一边相连者（边数 1）居中环（0.32 × min(W,H)，强度 0.22）；多边相连者（边数 ≥ 2）自由居中，仅受微弱归中之力（强度 0.03）。故中区相近者形式证据较强，然绝对距离仍为美学之平衡，非测度之相似度。可拖拽任一节点以重布；模拟将重平衡。标签： 默认仅显示 ≥3 次出现之公式名；悬停未标节点以显示其名。缩放与平移： 滚轮缩放，单击拖动平移（使用 d3-zoom）。
Greek formula
Latin formula
Calque edge
Bridge-inscription co-occurrence
Decree-class inscriptions法令类铭文
Click any row for full text + formula highlights.点击任意行查看完整文本与公式高亮.
Bridge inscriptions only仅桥接铭文
| UID | Lang语言 | Date年代 | Place地点 | Institution机构 | Genre类型 | Formulae公式 |
|---|
Evidentiary Profile — HT (historical) × ET (epigraphic) × RT (restoration)证据画像 — HT（历史） × ET（铭文） × RT（修复）
The "Evidentiary Profile" is the umbrella name for the three independent tiers that together describe each inscription's value as evidence: HT (content as evidence), ET (survival as evidence), RT (inferential mediation as evidence). See §16 of the typology plan for the rationale.“证据画像” 是三个独立层级之总称，共同描述每条铭文之证据价值：HT（内容证据）, ET（留存证据）, RT（推断中介证据）. 见 类型学计划 §16.
Three independent tiers per record (M15 redesign)
每条记录三个独立层级（M15 重新设计）
HT — Historical Tier
density of significant terms: named persons + institutional / magisterial markers + date markers + socio-political, cultural, economic, religious concepts.重要术语密度：具名人物 + 机构/官职标记 + 纪年标记 + 社会政治，文化，经济，宗教概念。
High HT → the text is content-rich.HT 高 → 文本内容丰富。
ET — Epigraphic Tier
preservation completeness: long unbroken runs of intact text, low lacuna density, low dotted-letter density.保存完整度：长段无中断文本，低残缺密度，低加点（存疑）字母密度。
High ET → primary-witness reliability.ET 高 → 原始证据可靠。
RT — Restoration Tier
density of editorial restoration: many […] brackets, much scholarly inferential investment.编辑修复密度：多 […] 括号，学者推断投入甚重。
High RT → much hypothesis built into the surviving readings.RT 高 → 现存读法多承编者推断。
The three are independent. A record can be HT5 / ET5 / RT1 (content-rich AND pristine AND minimal restoration — the SC de Bacchanalibus case) or HT2 / ET1 / RT4 (content-thin AND fragmentary AND heavily restored — the Asklepieion decree case). The scatter below plots HT (Y) × ET (X); RT shows in the type label and the distribution table. 三者独立。 一条记录可为 HT5 / ET5 / RT1（内容丰富 + 完整 + 极少修复，如 SC Bacchanalibus）亦可为 HT2 / ET1 / RT4（内容稀薄 + 残缺 + 大量修复，如 Asklepieion 法令）。下方散点图绘 HT（Y） × ET（X）; RT 则见类型标签与分布表。
Interactive zoom (Main + pairwise + faceted panels): wheel scrolls toward the cursor (up to 6×); click-drag pans within the zoomed region; double-click resets. Each scatter zooms by shrinking its SVG viewBox, so axes, gridlines, data circles, and the four-quadrant overlay all share one coordinate space — every data point stays exactly aligned with its true axis-tick value at any zoom level, and the quadrant boundaries always cut along HT = 0.5 and ET = 0.5. The Main scatter also offers an "↗ open in new window" button (per-scatter toolbar) for a standalone popup with the same zoom controls. 交互式缩放（主图，两两投影，分面小图皆同）: 滚轮按光标方向放大（至 6 倍）；单击拖拽于已放大之区域内平移；双击还原。各散点图以收缩其 SVG viewBox 实现缩放，故坐标轴，网格线，数据点，与四象限叠层共处同一坐标系 — 各数据点于任一缩放层级皆与其轴上之真实数值精准对齐，象限分界恒沿 HT = 0.5 与 ET = 0.5 之线划分。主散点图另设 “↗ open in new window” 按钮（各图工具栏内），可于独立窗口中以同等缩放控制查看。
2D view: HT (Y) × ET (X). Tier bins (T1 → T5 in 0.20 steps): T5 ≥ 0.80, T4 ≥ 0.60, T3 ≥ 0.40, T2 ≥ 0.20, T1 < 0.20. 2D 视图：HT（Y） × ET（X）. 层级（T1 → T5 按 0.2 步距）: T5 ≥ 0.80, T4 ≥ 0.60, T3 ≥ 0.40, T2 ≥ 0.20, T1 < 0.20。
Pairwise projections — three 2D scatters两两投影 — 三张 2D 散点图
Each pair of tiers exposes a different analytic question (M19.2). Use the same sample-size + highlight selectors above; click any record to open its detail modal.每对层级提示不同分析问题（M19.2）。使用以上抽样数量与突出选择器；单击任意记录打开详情。
HT × ET: where do content-rich records cluster — top-right (well-preserved) or bottom-left (damaged)?
ET × RT: does well-preserved ↔ less-restored hold? Strong negative correlation predicted.
RT × HT: are content-rich records restored more? Tests the "scholars restore what matters" hypothesis.
Faceted small multiples — HT × ET by RT tier分面小图 — 按 RT 层级之 HT × ET 散点
Five 2D scatters, each filtered to one RT tier. Reading across the panels — from RT1 (≤ 20% restored) to RT5 (> 80% restored) — the third axis (RT) progresses from minimal to maximal restoration. The point cloud should drift from the lower-right (well-preserved, lightly restored) into the upper-left (damaged, heavily restored), giving you the joint 3D distribution without any rotation. Panels are fixed-size (360 px) and reflow into one of three symmetric arrangements depending on viewport width: 3+2 (the most common on a wide screen), 2+2+1 on medium screens, or a single stacked column on narrow ones. Each row stays centred. Each panel zooms independently (wheel = zoom, drag = pan, dblclick = reset), with axes, gridlines, and the quadrant overlay all kept in lock-step with the data via viewBox-based zoom. 五张 2D 散点图，各仅留一个 RT 层级。自 RT1（补字 ≤ 20%）渐至 RT5（补字 > 80%）跨读各面板，第三轴（RT）由极少补字渐至极多补字。点云宜由右下（完好少补）漂往左上（残损多补），即不旋转而呈三维联合分布。面板尺寸固定（360 像素），依视窗宽窄重布为三种对称布局之一： 至宽则 3+2（常态），中宽则 2+2+1，至窄则单列堆叠；各行皆居中以保平衡。各面板独立缩放（滚轮 = 缩放，拖拽 = 平移，双击 = 还原），坐标轴，网格线，与象限叠层皆由 viewBox 缩放与数据同步。
Records with 20–40% restored text. Light editorial intervention; main argument still on stone.补字 20–40% 之文书。编者所补尚轻，大义仍存于石。
Records with 40–60% restored text. Moderate restoration; structure recoverable, formulae often supplied.补字 40–60% 之文书。中度补字，结构可复，程式常由编者增益。
Records with 60–80% restored text. Heavy editorial intervention; reading depends heavily on the editor.补字 60–80% 之文书。补字甚重，通读多藉编者之力。
Records with > 80% restored text. Mostly reconstruction; use with caution as primary evidence.补字 > 80% 之文书。多为重构，作主证当慎。
X axis = ET (epigraphic preservation, 0 left → 1 right). Y axis = HT (historical content density, 0 bottom → 1 top). Each panel uses the current sample-size and highlight selectors above. Panels are 360 px wide and reflow to fit the viewport — each row is centred so the layout stays balanced as you resize. X 轴 = ET（铭文留存度，0 左 → 1 右）。Y 轴 = HT（历史内容密度，0 下 → 1 上）。各面板共用上文之抽样数量与突出选择器。面板宽 360 像素，随视窗宽窄重布，各行居中以保布局平衡。
Formula bank (closed vocabulary)公式库（封闭词汇）
Cross-language calques shown with → links.跨语言转译以 → 标识.
| UID | Label名称 | Slot位置 | Tier层级 | Attestations出现次数 | Calque转译 | Anchor文献依据 |
|---|