Lemmatisation: 1,346 tokens, every one searchable词形标注: 1,346 个词标记, 皆可检索

<w lemma="ἀρχιερεύς">ἀρχιερεὺς</w>
<w lemma="δημαρχικός">δημαρχικῆς</w>
<w lemma="ἑπτακαιδέκατος">ἑπτακαιδέκατον</w>
<w lemma="ἐξουσία">ἐξουσίας</w>

Every Greek word in C.101's 152 lines — all 1,346 of them — sits inside a <w> wrapper carrying its dictionary form on @lemma. A naïve search for "δημαρχικός" against the raw text would miss δημαρχικῆς and δημαρχικὴν; with lemmatisation, all three forms surface together. The lemmatisation pass is timestamped 2020-11-24 by editor "Irene" — even the digital scholarly labour is preserved.C.101 在 152 行中的 1,346 个希腊词, 无一例外, 都包在一个带 @lemma 属性的 <w> 中, 上挂词典原形。若以原文 "δημαρχικός" 搜索, 必漏掉 δημαρχικῆς、δημαρχικὴν 等屈折形; 一加词形标注, 三形齐显。本次标注的时间戳为 2020-11-24, 经手者为编者 "Irene"，连数字学术之劳作, 亦被妥为存档。

The killer corpus query. Across all of IRCyr2020 (~2,300 inscriptions), find every attestation of ἐξουσία in any case form and rank by date. Without lemmatisation, this would mean grep + a Greek-paradigm checklist + manual disambiguation, taking days. With it, one query, two seconds.一招制胜的语料库查询。在整个 IRCyr2020(约 2,300 件铭文)中, 找出 ἐξουσία 之一切格变化的出现, 并按年代排序。若无词形标注, 须 grep + 希腊词形对照表 + 人工消岐, 旷日费时; 一旦标注到位, 一条查询, 两秒之间。

Editions, translations, and the file's own biography版本、译文, 与文件本身之传记

Why this matters for accessibility. A reader who can't read Greek can still grasp every edict via four translations. A reader who can read Greek can compare Reynolds's 2020 text against Oliverio 1927's editio princeps via the apparatus. A reader who wants to cite this monument has 28 prior editions at hand. And a reader who wants to know how the digital edition itself was built can read the file's own change log. One file, four readerships, one hundred years of scholarship.这何以关乎可读性。不通希腊文者, 仍可借四份译文通晓六诏全意; 通希腊文者, 可借校勘记将 Reynolds 2020 之校本与 Oliverio 1927 之祖本逐字比对; 欲引此碑者, 28 部前人版本一掌可握; 欲考此电子版之造法者, 文件自附沿革日志可读。一份文件, 四类读者, 百年学问。

Before any text: the metadata machine未及一字, 先有元数据机制

<msIdentifier><repository ref="institution.xml#db933">Cyrene Museum</repository></msIdentifier>
<physDesc>
  <objectDesc>
    <supportDesc><support>
      <p><material>Marble</material> <objectType>stele</objectType>, tapering toward the top
      (<dimensions><width>0.61-0.54</width><height>2.045</height><depth>0.36</depth></dimensions>).</p>
    </support></supportDesc>
    <layoutDesc><layout><rs type="execution" key="scalpro">Inscribed</rs> on one face.</layout></layoutDesc>
  </objectDesc>
  <handDesc><handNote>Augustan; ave. <height>0.01</height>.</handNote></handDesc>
</physDesc>

Every descriptive sentence a museum catalogue might write is here split into typed elements with measurable values. <material> isn't the word "marble" but a typed reference; <dimensions> carries three child elements with measurements in metres; <rs type="execution" key="scalpro"> means the inscription was chisel-cut (the key is from a controlled vocabulary of inscribing techniques).凡博物馆目录可能写出的描述, 此处皆被拆为带可测量值的有类型元素。<material> 不是"大理石"这个词, 而是一个类型引用; <dimensions> 内含三个子元素, 各以米为单位; <rs type="execution" key="scalpro"> 表示此铭为凿刻而成(键名取自一份刻法受控词表)。

Caveat. Don't treat the teiHeader as "the page before the text". It IS the text, for any search that asks "show me all marble stelae over 1.5 m" or "all inscriptions in Augustan letterforms". The encoding is what makes those queries possible.编者所诫。不可将 teiHeader 视作"正文前的扉页"。对任何"列出所有高过 1.5 米的大理石碑"或"所有奥古斯都体铭文"的检索而言, teiHeader 即是正文。其编码, 正是此类查询所赖。

Seven photographs, each linked to its lines七张照片, 各对应所摄之行

<facsimile>
  <graphic url="Photos_0025" decls="#photograph">
    <desc>Lines 1-47 (Department of Antiquities, D 44)</desc></graphic>
  <graphic url="Photos_0026" decls="#photograph">
    <desc>Lines 83-134 (Department of Antiquities, D 46)</desc></graphic>
  <graphic url="Photos_0024" decls="#photograph">
    <desc>Lines 112-144 (Department of Antiquities, D 47)</desc></graphic>
  <graphic url="DSC04081" decls="#photograph"><desc>Face (2008, H.Walda)</desc></graphic>
  <!-- + DSC04082, DSC04083, DSC04084 -->
</facsimile>

Each <graphic> is a typed pointer to an image file. The @url resolves to a real photograph; @decls="#photograph" says "this is a photograph, not a drawing or squeeze" — by pointing back to the <classDecl> in the teiHeader, where xml:id="photograph" is defined. The <desc> tells the reader (and any indexer) which lines that image covers and where it was taken.每一 <graphic> 是指向图像文件的有类型指针。@url 解析至真实照片; @decls="#photograph" 言"此为照片, 非线描亦非拓本"，经由 teiHeader 中 <classDecl> 内 xml:id="photograph" 之定义回连。<desc> 则告知读者(与索引器)该图覆盖第几行、何人何时所摄。

Caveat. The same <facsimile> pattern accepts a <zone> child that pins exact pixel coordinates to specific letters on the image — a step EpiDoc supports but Reynolds did not take here. The framework scales up if you need region-level annotation of the photograph later.编者所诫。同一 <facsimile> 模式可加一 <zone> 子元素, 把照片上特定像素坐标钉至具体字母，此步 EpiDoc 支持, Reynolds 此处未做。日后若需对照片做区域级标注, 框架仍可扩展。

Dating: stone-level and edict-level, both encoded断代:碑级与文级, 皆有编码

<origin>
  <origPlace>Unknown.</origPlace>
  <origDate notBefore="-0007" notAfter="-0006"
            evidence="titulature" n="I-IV">I-IV: 7-6 BCE; </origDate>
  <origDate notBefore="-0004" notAfter="-0004"
            evidence="titulature" n="V-VI">V-VI: 4 BCE</origDate>
</origin>

<!-- inside each textpart edition, the imperial titulature carries a godot.date URI: -->
<date ref="https://godot.date/id/k2sqW4JtApudigweWEWKNb">
  <persName type="emperor" key="augustus">…</persName>
  …δημαρχικῆς ἐξουσίας ἑπτακαιδέκατον…</date>

Two <origDate> elements record the two dating campaigns on the stone — @evidence="titulature" says "we know the year from the emperor's tribunician titulature, not from the stratigraphy or letterforms". The @n attribute spells out which textparts each range covers. Inside the edition, the actual phrase "in his 17th year of tribunician power" is wrapped in <date ref="godot.date/id/…"/> — a globally stable URI for that single regnal year.两个 <origDate> 元素记录碑面上的两期断代，@evidence="titulature" 言"我们由帝王任职年表得年, 而非由地层或字体"。@n 属性载明各范围所覆盖之文段。编辑层内, "任保民官第十七年"一语被裹入 <date ref="godot.date/id/…"/>，此 URI 对该单一在位年, 在全球范围内皆稳定。

Caveat. Stone-level <origDate> uses ISO-8601 BCE syntax (-0007 = 7 BCE; the leading hyphen and four-digit padding are mandatory). godot.date URIs are opaque — never invent one; always look up the exact ID at godot.date.编者所诫。碑级 <origDate> 用 ISO-8601 公元前语法(-0007 即公元前 7 年; 前导短横与四位补零均强制)。godot.date URI 是不透明的，切勿臆造, 务必至 godot.date 查得精确 ID 后使用。

Where does xml:lang go, and why three times?xml:lang 标在哪里, 为何标三次?

<TEI xmlns="…" xml:id="C08600" xml:lang="en">          <!-- ① file metadata is in English -->
  <teiHeader>
    <profileDesc>
      <langUsage>                                   <!-- ② declare every language used -->
        <language ident="ar">Arabic</language>
        <language ident="grc">Ancient Greek</language>
        <language ident="la">Latin</language>
        …
      </langUsage>
    </profileDesc>
  </teiHeader>
  <text><body>
    <div type="edition" xml:lang="grc">            <!-- ③ inscription is in Ancient Greek -->
    <div type="translation" source="johnson1961" xml:lang="en">
    <div type="translation" source="riccobono1941" xml:lang="la">

Three distinct levels: ① the file's metadata language (English), ② the <langUsage> declaration listing every language anything in the file uses (here: ar, en, fr, de, grc, grc-Latn, el, he, it, la), and ③ the language of each editorial block. The same XML file legitimately contains Greek inscribed text, Latin translation, and English commentary — and a renderer can pick the right script and direction for each.三层各异:① 文件元数据所用之语(英语); ② <langUsage> 声明本件涉及之全部语言(此处含阿拉伯文、英文、法文、德文、古希腊文、拉丁化希腊文、现代希腊文、希伯来文、意大利文、拉丁文); ③ 每一编辑块各自之语种。同一 XML 文件合法地兼载希腊文铭文、拉丁文译文、英文注释，渲染器可据此为各部分选定正确字体与方向。

Caveat. Always use ISO 639-3 codes (grc = Ancient Greek, la = Latin, en = English). Codes like grc-Latn add a script subtag — useful for transliterated Greek. The xml:lang attribute cascades: any element without its own value inherits its parent's. So you only declare a switch.编者所诫。务用 ISO 639-3 代码(grc 为古希腊文, la 为拉丁文, en 为英文)。grc-Latn 之类加书写脚本子标签，对拉丁化希腊文颇为有用。xml:lang 具继承性:任何未显式标注之元素, 自动继承其父语言。故只须于换语处声明一次。

Naming the named: emperor, defendant, consul, freedman命名所命:元首、被告、执政官、释奴

<!-- ① the emperor, by controlled key (no lookup needed) -->
<persName type="emperor" key="augustus">
  <w lemma="Αὐτοκράτωρ">Αὐτοκράτωρ</w>
  <name nymRef="Καῖσαρ">Καῖσαρ</name>
  <name nymRef="Σεβαστός">Σεβαστὸς</name></persName>

<!-- ② a defendant attested by LGPN prosopography (line 42) -->
<persName type="attested" key="lgpn:V1-65612">
  <name nymRef="Πόπλιος">Ποπλίωι</name>
  <name nymRef="Σέξστιος">Σεξστίωι</name>
  <name nymRef="Σκευᾶς">Σκεύαι</name></persName>

<!-- ③ a consul named by Roman tria nomina components (line 84) -->
<persName type="attested">
  <name nymRef="Γάϊος">Γάϊος</name>
  <name nymRef="Καλουίσιος">Καλουίσιος</name>
  <name type="cognomen" nymRef="Σαβεῖνος">Σαβεῖνος</name></persName>

<!-- ④ a freedman, identified through his patron (line 43–44) -->
<persName type="attested" key="lgpn:V1-65907">
  <name nymRef="Πόπλιος">Πόπλιον</name>
  <name nymRef="Λακουτάνιος">Λακουτάνιον</name>
  <persName type="attested"><name nymRef="Πόπλιος">Ποπλίου</name></persName>
  <w lemma="ἀπελεύθερος">ἀπελεύθερον</w>
  <name nymRef="Φιλέρως">Φιλέρωτα</name></persName>

One element family — <persName>/<name> — encodes four very different naming acts. @type separates the imperial protocol from a private defendant; @key with an LGPN URI links the defendant to the Lexicon of Greek Personal Names; @type="cognomen" on a child <name> marks the Roman tripartite name (praenomen-gentilicium-cognomen); a nested <persName> for the patron carries the freedman's legal genealogy.同一族元素，<persName>/<name>，承载四种全然不同的命名行为。@type 区分帝王官式与平民被告; @key 配 LGPN URI, 将被告挂接《希腊人名词典》; 子元素 <name> 之 @type="cognomen" 标罗马三段式名(praenomen-gentilicium-cognomen); 内嵌之 <persName> 则装入施主之名, 以承载释奴之法律出身。

Caveat. The Romans on this stone wear Greek inflectional clothing: Σέξστιος Σκεύας = Sextius Scaeva. The @nymRef on each <name> gives the canonical (nominative) form, so a search for "Sextius" works even if the inscription has the dative Σεξστίωι. Do not collapse the layers — Reynolds's decision to mark both the carved inflected form and the canonical nymRef is the whole point.编者所诫。本碑上的罗马人皆着希腊屈折之衣:Σέξστιος Σκεύας 即 Sextius Scaeva。每个 <name> 上的 @nymRef 给出标准(主格)形式, 故搜索 "Sextius" 即便铭文用与格 Σεξστίωι 亦可命中。切勿合并此二层，Reynolds 同时标注所刻屈折形与标准 nymRef, 正是要点。

A place can be six different kinds of thing"地"可以是六种不同之物

<!-- ancient findspot (where stone was carved/erected) -->
<placeName type="ancientFindspot" ref="https://www.slsgazetteer.org/909">Cyrene</placeName>

<!-- monuList — a building inside the findspot -->
<placeName type="monuList" ref="https://www.slsgazetteer.org/1327">Assembly Building</placeName>

<!-- ancientRegion — a province or geographic region -->
<geogName type="ancientRegion" key="Cyrene">Cyrenaica</geogName>

<!-- ethnic name (NOT a place per se — a demonym) -->
<placeName type="ethnic" nymRef="#Ῥωμαῖος" ref="#roma">Ῥωμαίους</placeName>

<!-- modernCountry / modernFindspot — for the present-day location -->
<geogName type="modernCountry" key="LY">Libya</geogName>
<placeName type="modernFindspot" ref="http://sws.geonames.org/82972">Shahat</placeName>

Six semantic roles, all expressed with a placeName/geogName element discriminated by @type. Cyrene as the ancient findspot resolves to the SLS gazetteer; Cyrenaica as the ancient region uses a controlled @key; Romans as an ethnic resolves to an internal #roma anchor in the file's prosopography list; Libya uses ISO 3166-1 alpha-2; Shahat is the modern town built over Cyrene, pinned to GeoNames.六种语义角色, 皆由 placeName/geogName 元素表达, 由 @type 区分。Cyrene 作古代出土地, 解析至 SLS 地名表; Cyrenaica 作古代区域, 用一受控的 @key; Romans 作民族名, 解析至本件人物志表中之内部锚点 #roma; Libya 用 ISO 3166-1 alpha-2; Shahat 是覆于昔兰尼故址之上的今镇, 钉至 GeoNames。

Caveat. Internal anchors (ref="#roma") and external URIs (ref="https://www.slsgazetteer.org/909") both use the same @ref attribute. The hash-only form points to an entry inside this XML file's prosopography; the full URL points outward. Don't confuse them — both are valid, but they resolve in very different places.编者所诫。内部锚点(ref="#roma")与外部 URI(ref="https://www.slsgazetteer.org/909")共用同一个 @ref 属性。仅含井号者指向本 XML 文件内部人物表之条目; 完整 URL 则外引。切勿混淆，二者皆合法, 但解析至全然不同的地方。

When letters mean numbers: ιθ = 19字母作数:ιθ = 19

<!-- line 73 of C.101: "the 19th tribunician year" -->
<lb n="73"/><w lemma="δημαρχικός">δημαρχικῆς</w>
<w lemma="ἐξουσία">ἐξουσίας</w> <space quantity="3" unit="character"/>
<num value="19"><hi rend="supraline">ιθ</hi></num>

Caveat. <num>, <hi>, <g> (special glyph), and <am> (abbreviation mark) overlap in scope but each says something different. num: this is a number, and its value is N. hi: the cutter rendered this part of the line in some special way (supraline, ligature, tall letter). g: this is a non-textual glyph (an interpunct, a Christogram). am: this stretch of letters is a mark of abbreviation (the doubled dd for dominorum). Each is independently filterable.编者所诫。<num>、<hi>、<g>(特殊字形)、<am>(缩写符号)四者范围有重叠, 但各言其意。num:此为数, 值为 N。hi:刻工以特殊方式表现此处(加横线、连写、高字)。g:此为非文字之字形(间隔点、十字章)。am:此段字母为缩写之标记(dominorum 之双写 dd)。四者皆可独立过滤。

Intentional blank space: three encodings有意留白:三种编码

<!-- ① character-extent blank between phrases (line 3 of C.101) -->
<lb n="3"/> <space quantity="12" unit="character"/>
<w lemma="λέγω">λέγει</w> <space quantity="15" unit="character"/>

<!-- ② full-line blank (line 71a of C.101) -->
<lb n="71a"/> <space quantity="1" unit="line"/>

<!-- ③ uncertain extent (after Edict II ends) -->
<space quantity="8" unit="character" precision="low"/>

<!-- ④ unmeasured trailing space -->
<space precision="low"/>

EpiDoc draws a sharp line between <gap> ("the stone is damaged here, text is lost") and <space> ("the cutter intentionally left this much blank"). C.101 uses <space> constantly to mark the careful blanks the cutter placed between phrases, after each λέγει ("proclaims:") to introduce the edict's text proper, and to flag the gap between one edict and the next. @quantity + @unit measure the extent (characters or lines); @precision="low" says "approximately, not measured exactly".EpiDoc 严辨 <gap>("此处石面残, 文已失")与 <space>("刻工于此有意留白")。C.101 频用 <space> 标记刻工于句与句之间精心留出的空隙, 用以引出每篇诏令之 λέγει("乃下令曰:"), 或标示一诏令与下一诏令间的过渡。@quantity 与 @unit 计量长度(以字母或行计); @precision="low" 言"大致如此, 未精确测量"。

Caveat. A blank line is encoded twice: as <lb n="71a"/> (giving the blank line a number — note the "a" suffix to preserve the running line count) AND as <space quantity="1" unit="line"/>. The two together say "the stone has a blank 71st line; it counts toward the line-numbering but carries no text". Omitting either would corrupt the line index.编者所诫。一行空白须双编:既作 <lb n="71a"/>(给空白行一个行号，注后缀 "a", 以保行数连贯), 又作 <space quantity="1" unit="line"/>。二者并陈, 始言"碑上有一空白之第 71 行; 计入行号, 不载文字"。漏其一则行号失次。

Five distinct ways to say "this isn't a clean reading"五种不同方式言"此非洁本"

Caveat. A reader who skips over the @reason attribute loses half the meaning. Lost is a statement about the stone's preservation. Omitted is a statement about the cutter's practice. Unclear is the editor confessing doubt. Surplus is the editor exercising judgment. Erasure is a historical event — someone, sometime, deliberately scraped letters off this stone. Five different kinds of footnote, all machine-readable.编者所诫。忽视 @reason 属性者, 已失其义之半。Lost(失)是关于碑面保存状态的陈述。Omitted(漏)是关于刻工习惯的陈述。Unclear(疑)是编者之自承不定。Surplus(衍)是编者之裁断。Erasure(擦)是一历史事件，曾有人, 于某时, 刻意将此石上数字刮去。五种不同的脚注, 皆机器可读。

<lb>: the most important single element on the stone<lb>:石面之第一要素

<!-- ① a normal line break with line number -->
<lb n="1"/>Αὐτοκράτωρ Καῖσαρ Σεβαστὸς…

<!-- ② mid-word continuation (word broken across line end) -->
ἐπιβαρού<lb n="9" break="no"/>σας
<!-- "ἐπιβαρούσας" runs from end of line 8 onto line 9 without a hyphen -->

<!-- ③ a numbered blank line for line-count continuity -->
<lb n="71a"/> <space quantity="1" unit="line"/>
<lb n="73a"/> <space quantity="1" unit="line"/>

<!-- ④ a textpart boundary that shares a physical line (line 55 starts II and III) -->
<div type="textpart" n="II"> …<lb n="55"/> …</div>
<div type="textpart" n="III"><lb n="55"/> … <!-- same physical line, new edict -->

<lb/> is empty (self-closing) and self-anchoring. Its @n is the line number on the stone — not the position in the file. @break="no" says "this line break occurred mid-word, with no hyphen on the stone". Line numbers can repeat (line 55 carries the end of Edict II and the start of Edict III, because they share a single physical line), and the suffix "a" (71a, 73a) preserves count for fully blank lines.<lb/> 是空元素(自封闭), 自带定位之能。其 @n 是石面之行号，不是文件中的位置。@break="no" 言"此换行发生于词中, 石上未刻连字号"。行号可以重复(第 55 行兼载第 II 诏令之尾与第 III 诏令之首, 因二者同居一行); 后缀 "a"(71a、73a)则为全空之行保持行号连贯。

Caveat. <lb/> is the anchor for the whole apparatus criticus: <app loc="5"> means "compare with line 5". Get the line numbers wrong, and every apparatus entry, every cross-reference, every cited passage breaks. The @n attribute is not just a label — it's a citation handle that the entire scholarly apparatus depends on.编者所诫。<lb/> 是整套校勘记之锚点:<app loc="5"> 即"对照第 5 行"。行号一错, 所有 apparatus 条目、所有交叉引用、所有引文皆崩。@n 不仅是标签，是整套学术装置所赖之引用句柄。

What to verify in your own EpiDoc file你自家的 EpiDoc 文件, 应自检何处

The unifying principle. Every encoding choice in C.101 reflects the same discipline: separate what the cutter wrote from what the editor restored, separate visual rendering from semantic content, and link every name and date to an external authority. Get those three habits right, and the rest of EpiDoc follows. The full vocabulary lives in lexicon.html — keep it open while you work.统贯之道。C.101 中每一处编码选择, 皆反映同一规矩:分刻工所刻与编者所补; 分视觉表现与语义内容; 凡名凡日皆挂权威表。三习既正, 其余皆从。完整词汇见 lexicon.html，工作时常开备查。

Monument-as-corpus vs. corpus-as-monument碑刻即语料 vs. 语料即碑刻

Why both matter. A corpus that only knows how to encode monument-as-corpus (the Cyrene pattern) cannot represent dossiers whose unit of meaning is the wall. A corpus that only encodes corpus-as-monument (the Aphrodisias pattern) loses the textpart hierarchy that lets a reader cite C.101.III as a distinct legal act. EpiDoc supports both — the encoder picks whichever pattern matches the ancient editorial reality, not whichever is technically easier.两者都重要。语料库若只会做碑刻即语料(C.101 那一种), 就没办法呈现以墙为意义单位的册档; 若只会做语料即碑刻(阿芙罗狄西亚斯那一种), textpart 的层级又会消失, 读者也就引不到 C.101.III 这样一个独立的法律文书。EpiDoc 两路都走，编者该选哪一种, 跟着古代的编辑实况走, 不要跟着编码上的方便走。

When 40% of the text is editor当四成文本出自编者之手

The SC de Aphrodisiensibus survives as fragments scattered through the city walls, the theatre, and the Aphrodisias museum. In Reynolds’ edition, 2,671 letters out of c. 6,820 sit inside <supplied> — about 40% of the readable text is, formally, the editor’s restitution, distributed across 584 separate <supplied> elements. Compare C.101 (Cyrene): 52 supplied letters out of c. 5,000 — about 1%. The Aphrodisias file demonstrates the convention’s response to this density: every reconstructed segment carries @reason (lost vs. omitted), every gap carries @reason + @extent + @unit="character", and 95 ambiguous letters wear <unclear>. The format is honest about what the editor can and cannot see.《SC de Aphrodisiensibus》(《关于阿芙罗狄西亚斯人的元老院决议》) 的残片散布在城墙、剧场、阿芙罗狄西亚斯博物馆三处。Reynolds 校本印出来的约 6,820 个字母里, 有 2,671 个套在 <supplied> 之内：也就是说, 文本里约 40% 是编者的补订, 分散在 584 个 <supplied> 标记里。再看 C.101 (昔兰尼) 就不一样了:全文约 5,000 字, 只有 52 个补字, 约占 1%。阿芙罗狄西亚斯这份文件, 正好展示出规约如何回应这种残损密度:所有补字都挂 @reason (lost"失刻"或 omitted"漏刻")、所有阙文都挂 @reason + @extent + @unit="character"、95 处可疑字母全用 <unclear> 包起来。这套写法自陈所见之限, 不掩疑处。

What this enables. A reader writing a critical commentary on a single line can run a filter — //supplied[@reason="lost"] in line N — and see exactly which letters the editor recovered from parallel passages, which from grammar alone, and which the cutter never carved. The SC Aphrodisiensibus becomes two readable texts at once: the stone as found, and the stone as reconstructed. This separation is invisible in a printed edition and unrecoverable from a transcription that omits the brackets.这给了我们什么。想给某一行做精校评议, 一句 XPath 就够了，//supplied[@reason="lost"], 限于第 N 行，哪个字得自平行段落、哪个字凭文法所推、哪个字根本没刻过, 一目了然。《SC Aphrodisiensibus》于是双文同陈:出土的石头、与重构的石头。这层分别, 在印本中看不见, 在不带括号的转录里也回不来。

Line-break vs. word-break · encode both换行与分词 · 二者并记

<lb n="1"/>ΥΠΑΤΟΥ ΓΑΙΟΥ ΚΑΛΟΥΙΣΙΟΥ
<lb n="2"/>ΓΑΙΟΥ ΥΙΟΥ ΚΑΙ ΛΕΥΚΙΟΥ ΜΑΡΚΙΟΥ
<lb type="worddiv" n="3"/>ΛΕΥΚΙΟΥ ΥΙΟΥ ΕΚ ΤΩΝ ΕΙΣΦΕΡΟΜΕΝΩΝ
<lb n="4"/>ΕΙΣ ΤΗΝ ΣΥΝΚΛΗΤΟΝ ΔΟΓΜΑΤΩΝ
<lb type="worddiv" n="5"/>ΘΕΜΑ ΠΡΩΤΟΝ…

The Aphrodisias cutter sometimes ended a line on a word boundary and sometimes split a word across the break. The standard <lb n="N"/> records the break itself; <lb type="worddiv" n="N"/> additionally records that the cutter chose to end at a word boundary — a deliberate layout decision, not a scribal accident. C.101 contains no such markup: the Cyrene cutter broke lines wherever space ran out, with no preserved care for word boundaries. This single attribute distinguishes a workshop that respected the eye of the reader from one that did not, and EpiDoc captures the distinction with one extra token.阿芙罗狄西亚斯的刻工有时把行尾停在词的尾巴上, 有时则让单词跨过换行。<lb n="N"/> 只记下换行本身; <lb type="worddiv" n="N"/> 还多记一点:这一处的换行恰好是词尾，也就是刻工有意做的版面决定, 不是手滑。C.101 里并没有这一标记:昔兰尼的刻工凭空间所限就断行, 顾不上词界。就这一个属性, 已经足以分别一位顾及读者眼睛的刻工和不顾的刻工, EpiDoc 一个 token 就把它记下来了。

What this enables. A palaeographer studying inscriptional layout can run count(//lb[@type="worddiv"]) ÷ count(//lb) across the IAph corpus and produce, in one query, a quantitative metric of word-boundary discipline by workshop and by period — exactly the kind of question that a printed edition with its brackets and dashes makes nearly unanswerable.这给了我们什么。研究铭文版面的古文字学者, 在 IAph 中跑一句 count(//lb[@type="worddiv"]) ÷ count(//lb), 即可得分作坊、分时期的"行尾守词"量化指标，这正是带括号、带连字符的印本几乎答不了的题。

Stones travel before they speak石头先漂泊, 后说话

Compare C.101. The Cyrene stone was found in one spot — re-used in the floor of a room next to the Assembly Building — and recorded by one excavator. Block-level provenance never had to be encoded. The Aphrodisias file shows what <history> can do when an inscription is dispersed: it makes the geography of fragmentation part of the citable record.C.101 是反过来的。昔兰尼那块石头只出土于一处，改用作议事堂旁某室的铺地石，由一位发掘者一并记录, 所以不必再分块编码出土史。阿芙罗狄西亚斯这份文件展示的是:当铭文散布在多处时, <history> 这一节能做什么，它把"碎片的地理学"纳进可引用的记录里。

Bibliography as stemma · 1705 → 2007书目即传抄谱系 · 1705 → 2007

Each badge below shows what the editor was looking at when they edited: a manuscript copy, someone else’s photograph, or the stone itself.下方各色标签, 显示每位编者校订时看的是什么:是别人的手稿、别人的照片, 还是自己亲眼看过的石头。

The wall as web: one document completes another视墙为网:一文以补一文

The <commentary> of iAph080027 opens with a remarkable note:
"The text can be further supplemented from an area of the wall which survives intact, where extracts from related documents had been separately inscribed (8.28), providing the substance of ll. 32–35 and 77–82."iAph080027 的 <commentary> 开头有一段话, 特别值得留意:
"本文可以进一步靠档案墙上另一处仍然完好的区域来补字, 也就是另刻于那边的相关文书的节略 (8.28), 正好可以补出本文第 32—35 行、第 77—82 行的实质内容。"

Why this is the “next level”. C.101 demonstrated EpiDoc inside one file. The archive wall demonstrates EpiDoc between files — a corpus whose unit of meaning is the relation, not the inscription. The single most important skill an Aphrodisias encoder learns that a Cyrene encoder does not is the discipline of citing one’s sister files: @corresp, @source, and a bibliography in which every entry is also a digital sibling.这就是“次阶”的所在。C.101 展示的是 EpiDoc 在一份文件之内能做什么; 档案墙展示的, 是 EpiDoc 在多份文件之间能做什么，一座语料库, 它的意义单位是关系, 不是铭文。阿芙罗狄西亚斯的编者比昔兰尼的编者多学到的, 是引用姊妹文件的一种纪律:@corresp、@source, 还有一份每条目都同时是"数字姊妹"的参考书目。

One column · five files · one argument一柱 · 五文 · 一论

1 2 3 4 5

18.29Octavian (still triumvir) to Stephanos: a private letter freeing the city — “this one city for mine out of all Asia.”屋大维 (尚为三人执政官) 致 Stephanos:私函, 赐城以自由 ——“凡亚细亚一州, 我独取此一城为己有。”39/38 BCE

28.32Augustus to the Samians: a subscript refusing Samos the freedom Aphrodisias enjoys.奥古斯都致 萨摩斯人:批答, 拒赐萨摩斯以阿芙罗狄西亚斯所享之自由。38 BCE

38.33Trajan to the Smyrnaeans: restricts Smyrna from imposing liturgies on Aphrodisians (city extra formulam provinciae).图拉真致 士麦那人:限制士麦那不得强令阿芙罗狄西亚斯人服役 (该城已出行省名册)。98 / 100 AD

48.28Extracts from the SC de Aphrodisiensibus (parent file: iAph080027) — the foundational legal text excerpted onto the wall.节录《SC de Aphrodisiensibus》(母本就是 iAph080027)，根本法律文本的节略, 刻在这墙上。39/38 BCE

58.102Gordian III to Aphrodisias, renewing privileges — the column’s only contemporary inscription, “less carefully cut.”戈尔迪安三世致阿芙罗狄西亚斯, 重申特权，本柱唯一同期之刻文, “刻工不甚精”。239 AD

A late-antique anthology of early documents晚期古典对早期文书之结集

<origDate notBefore="0101" notAfter="0300" evidence="lettering context">
  second to third centuries A.D.    <!-- when the stone was carved -->
</origDate>

<ab>
  <lb n="1"/><date when="-0038" evidence="content">
    αὐτοκράτωρ Καῖσαρ θεοῦ Ἰουλίου υἱὸς Αὔγουστος
    Σαμίοις ὑπὸ τὸ ἀξίωμα ὑπέγραψεν…
  </date>                                  <!-- when the text was composed -->
</ab>

Each Aphrodisias XML file carries two encoded dates: an <origDate> in the <teiHeader> for when the stone was carved, and a separate <date> (or the imperial titulature) in the body for when the document was composed. The gap can be more than two centuries. In four of these five files, the gap is ~150–300 years — the Antonine or Severan archive workshop transcribing Caesarian, Augustan and Trajanic letters onto the parodos wall, almost certainly from a roll-form civic archive that no longer survives.阿芙罗狄西亚斯的每一份 XML 都挂两组年代:<teiHeader> 里的 <origDate> 记刻石的年代, 文本中的 <date>(或借帝王名号) 记原书的年代。两者的差距, 有时超过两百年。这五件里的四件, 差距大约 150 到 300 年，安东尼或塞维鲁时代的档案作坊, 把凯撒、奥古斯都、图拉真几朝的信件, 重新刻在剧场 parodos 墙上;底本想必当时还以纸卷形式存放在城里的档库, 现在已经看不到了。

What this enables. A historian writing on civic archival practice in the Roman East can query //origDate[@notBefore > -100] / / following::date[@when < -0030] across the whole IAph corpus and surface every document whose memorialisation post-dates its composition by more than a generation. The wall stops being “an inscription” and becomes “a piece of evidence for how cities remembered their own pasts.”这给了我们什么。研究罗马东方诸城档案实践的史家, 在整个 IAph 里跑一句 //origDate[@notBefore > -100] / / following::date[@when < -0030], 就能取出所有"追刻晚于原书一世代以上"的文书。墙到了这一步, 已经不再只是"一件铭文", 而成了"诸城追忆自家过往的物证"。

Ancient erasure · when the editor was a contemporary古代抹刻 · 编者亦曾同时之人

<ab>
  <lb n="6"/> τῷ δήμῷ τῷ Πλαρασέων καὶ
  <del>
    <unclear reason="damage">Ἀφροδεισιέων</unclear>
  </del>
  ἐπιτάσσεσθαι
</ab>

In 8.28, every occurrence of “Ἀφροδεισιέων” ("of the Aphrodisians") has been deliberately erased on stone — six erasures across fifteen lines, all of the second half of the double polity-name Plarasa-Aphrodisias, leaving only “Πλαρασέων καὶ — —”. Each erasure is encoded as <del> wrapping <unclear reason="damage">…</unclear>: this passage was once readable, the ancient editor removed it, the modern editor can still make it out under the chisel-marks.在 8.28 里, “Ἀφροδεισιέων”(“阿芙罗狄西亚斯人的”) 每一处出现都被当时人有意凿去，十五行间共六处, 凿去的都是“Plarasa-Aphrodisias”这一双名城邦的后半, 只剩“Πλαρασέων καὶ，——”。每一处都用 <del> 套 <unclear reason="damage">…</unclear> 的结构编码:这一处原本可读, 当时的编者凿去了, 今天的编者还能在凿痕底下辨出来。

Four titulatures, 280 years, one wall四帝号、二百八十年、一面墙

Same wall, 280 years, four protocols — the meter bars are the literal dating tool: each emperor carries a stable URI via <persName ref="…">, so a SPARQL query asking “which Roman emperors does the Aphrodisias archive cite, and in what proportions?” comes back in milliseconds.同一墙, 280 年, 四套帝号，上方仪表条之长度, 即断代之实测器:每一帝王以 <persName ref="…"> 挂稳定 URI, 故 SPARQL 一问 ——“阿芙罗狄西亚斯档案征引哪些罗马帝王, 各占几何?”—— 顷刻可答。

Near the wall, not on it近墙而不在墙上

Where it sits — the parodos’s 21 archive-wall documents (purple) sit separated from iAph 8.100 (terracotta), which the cutter inscribed on the analemma. A dashed green arc marks their thematic kinship.这块石头在哪里：parodos 上 21 件档案墙文书 (紫色) 跟 iAph 8.100 (赭红) 是分开的:后者刻在 analemma 上。那条绿色虚弧, 标的是两者主题上的亲缘。

One letter · four palaeographic specimens一函四题 · 古字学样本

Ligatures连字①

25+25+ 式

ΗΜ ΤΗ ΝΗ ΠΗΝΤ

ΗΜ at ll. 2, 7, 17, 18 · ΤΗ at ll. 4, 13–17, 19 · ΝΗ at ll. 6, 9 + many more.ΗΜ 见 2、7、17、18 行 · ΤΗ 见 4、13—17、19 行 · ΝΗ 见 6、9 行, 余者尚多。

<handNote> ligatures: ΗΜ ll. 2, 7, 17, 18 … </handNote>

Apices长音点②

77 处

α ι υ ω ε

Long-vowel marks on ll. 1, 2, 4, 7, 9, 10, 13 — a pronunciation aid preserved across two millennia.长元音上方的斜线, 见 1、2、4、7、9、10、13 行，刻工留下的读音辅助, 历经两千年。

<handNote> apices in ll. 1, 2, 4, 7, 9, 10, 13 </handNote>

Arabesque花押③

11 处

ἐντολὰς ❦

A decorative flourish at the end of l. 12 — the cutter’s only personal mark on the stone.第 12 行行末一花押，刻工在这块石头上唯一的私人留痕。

<handNote> arabesque at end of l. 12 </handNote>

Erasures凿去④

33 处

Ἀφροδεισιέων Ἀφροδείτῃ

Ἀφροδεισιέων at ll. 5, 19 + Ἀφροδείτῃ at l. 14 chiselled out. Even the goddess’s name struck.Ἀφροδεισιέων(5、19 行) 与 Ἀφροδείτῃ(14 行) 凿去。连女神之名亦遭抹刻。

<del><unclear reason="damage">Ἀφροδεισιέων</unclear></del>

What the two case studies teach两案所教

What you learn from one, you cannot learn from the other. Cyrene shows EpiDoc inside a single file — nested hierarchy, dense lemmatisation, multi-level dating in one root. Aphrodisias shows EpiDoc between files — a corpus whose unit of meaning is the relation, where archive-membership is content not stone, and where the cutter’s ligatures, apex marks, and erasures are independent objects of inquiry. Both are needed; both are EpiDoc.一案所教, 另一案教不出来。昔兰尼展示 EpiDoc 在一份文件之内，层级嵌套、词汇标注密集、断代分层于同一棵根。阿芙罗狄西亚斯展示 EpiDoc 在多份文件之间，一座语料库, 它的意义单位在于关系, 档案归属看的是内容, 不是石头;刻工的连字、长音点和抹刻, 也都能独立考察。两者都要紧, 也都是 EpiDoc。

Questions the encoding makes askable编码所开之问

Three domains, six questions, one infrastructure. Material questions ask about the wall; content questions about its text; method questions about how to query at scale. None of them could be put to a printed edition. Each lives where an encoding feature meets a research curiosity — which is, finally, what EpiDoc is for.三个领域、六个问题、一套基础设施。材料的问题问墙;内容的问题问墙上的文字;方法的问题问规模上的可查询性。这些问题, 一个都没法向印本发问。每一个问题都生于"编码特征"跟"研究好奇心"的相遇，这, 正是 EpiDoc 存在的理由。

What still needs to be built尚待建之事

The takeaway. Encoding alone is not enough. The questions on the previous slide only become answerable at scale once the supporting infrastructure exists. Three of seven tasks are wide open — an entire generation of thesis projects waiting for their first sentences.要点。光有编码不够。上一页那些问题, 要做到规模上能回答, 得先把支撑的基础设施建起来。七项里头, 三项现在全开，整整一代博士选题, 都还在等第一句话写下来。

The <expan> wrapper缩写展开标签

The stone carries the abbreviated word man. The Latin is manibus ("of the shades"). The editor's intervention separates what's on the stone from what's supplied:

石上所刻为缩写 man, 拉丁文完整形作 manibus(“亡灵的”, 与格)。编者的工作, 是把"石头上确有"与"为读者补出"两层清楚分开:

<w><expan><abbr>man</abbr><ex>ibus</ex></expan></w>

The interpunct as a glyph间隔点作为符号

Roman engravers used a raised dot to separate words. It is not a letter. EpiDoc encodes it via <g>:

罗马刻工常用一颗居中略高的圆点来分隔单词。它不是字母, 而是分词符号。EpiDoc 以 <g> 元素编码:

D <g ref="#interpunct">·</g> man

@ref="#interpunct" points to an entry in the project's character declaration. Every interpunct across thousands of inscriptions can then be retrieved with a single query.

@ref="#interpunct" 指向项目字符声明中的一条目。语料库中所有间隔点 (跨数千件铭文) 都可一次性检索。

Visual vs. content: <hi>视觉与内容: <hi>

On the stone, H and I in Zethi are joined into a single carved character. The TEXT is still two letters; the SHAPE is one. EpiDoc records appearance through @rend:

石上, Zethi 中的 H 与 I 被刻工合为一字。文字内容仍是两字母, 形态上是一字。EpiDoc 通过 @rend 属性记录这种形态:

<w>Zet<hi rend="ligature">hi</hi></w>

Line 1 of Zethus also carries <w>D<hi rend="tall">i</hi>s</w> — the i in Dis is taller. Look at the photograph and you'll see why.

Zethus 第 1 行还有 <w>D<hi rend="tall">i</hi>s</w>，Dis 中的 i 字高于左右二字。看照片即明。

<num> has both face and value<num> 兼具字形与数值

<num value="6">VI</num>

The text content VI is what's on the stone. The @value="6" is the computed integer. Two layers: a database can sort, filter, sum by value; a renderer displays the original face.

文本内容 VI 是石上所刻; @value="6" 是计算后的整数。两层分开: 数据库可按数值排序、筛选、求和; 渲染器显示原始字形。

Three layers for one name一人之名, 三层包装

<persName type="attested">
  <name>
    <w>Zet<hi rend="ligature">hi</hi></w>
  </name>
</persName>

Why three layers for what looks like one name?

为什么一个看起来一气呵成的名字, 要分成三层来包装?

• <persName> = the person. Attributes can link to a prosopographical authority (PIR, LGPN, individual project IDs).
• <name> = the name itself. For Roman tria nomina (Q. Pomponius Rufus), this layer hosts three siblings: <name type="praenomen">, <name type="gentilicium">, <name type="cognomen">.
• <w> = the word. The orthographic unit that the renderer treats as a token for highlighting, search indexing, lemmatization.

<persName> = 人物本身; <name> = 名字 (罗马三段名时此层下分 praenomen/gentilicium/cognomen 三个子元素); <w> = 词汇单元 (用于高亮、检索索引、词形归并)。

Everything assembled完整编码

<div type="edition" subtype="primary" xml:space="preserve" xml:lang="la">
  <ab>
    <lb n="1"/><w>D<hi rend="tall">i</hi>s</w> <g ref="#interpunct">·</g>
    <w><expan><abbr>man</abbr><ex>ibus</ex></expan></w>
    <lb n="2"/><g ref="#interpunct">·</g>
    <persName type="attested"><name><w>Zet<hi rend="ligature">hi</hi></w></name></persName>
    <lb n="3"/><w><expan><abbr>vix</abbr><ex>it</ex></expan></w>
    <g ref="#interpunct">·</g>
    <w><expan><abbr>a</abbr><ex>nnis</ex></expan></w>
    <g ref="#interpunct">·</g>
    <num value="6">VI</num>
  </ab>
</div>

What an empty-edition file looks like无刻文文件的样貌

<TEI xmlns="http://www.tei-c.org/ns/1.0" xml:id="IRT0102" xml:lang="en">
  <teiHeader>
    <fileDesc>
      <titleStmt><title>Fragment</title></titleStmt>
      <publicationStmt>
        <publisher>Society for Libyan Studies</publisher>
        <distributor>King's College London</distributor>
        <idno type="filename">IRT0102</idno>
        <availability><p>CC-BY UK 2.0</p></availability>
      </publicationStmt>
      <sourceDesc><msDesc><msIdentifier><msName>IRT0102</msName></msIdentifier></msDesc></sourceDesc>
    </fileDesc>
    <revisionDesc><change when="2021-06-24" who="CMR">Merged</change></revisionDesc>
  </teiHeader>
  <text><body>
    <div type="commentary"><p>See now <ref type="inscription" n="IRT0101">101</ref>, of which this is part.</p></div>
  </body></text>
</TEI>

The seven-layer path七层路径

From the root to the actual sentence "See now 101, of which this is part":

自根元素往下走, 直到那句真正的内容“请参看 101 号, 此残片为其一部”:

TEI → text → body → div[@type="commentary"] → p → ref[@type="inscription"]

Six levels of nesting (the seventh layer is the text content itself). Memorize this path — it works on every EpiDoc file in every project.

六层嵌套 (第七层为文本本身)。这条路径背下来：它适用于每一个项目的每一份 EpiDoc 文件。

Why the file exists at all为什么保留这份文件

IRT0101 is a larger inscription. IRT0102 turned out to be a smaller fragment of the same monument. Once the join was made, the editors had two choices:

IRT0101 本是一篇较大的铭文。IRT0102 后来证实是同一座纪念物上的小残片。"两者本为一体"这一判断作出后, 编者面前有两条路:

• (a) Delete IRT0102 entirely. Tidy, but breaks every existing citation to "IRT0102".
• (b) Keep IRT0102 as a stub redirecting to IRT0101.

They chose (b). This is the same principle as a web HTTP 301 redirect: the old reference still resolves, it just points to the new place.

编者发现 IRT0102 实为 IRT0101 之残片。删除会破坏所有现存引用; 保留为重定向残件则保住既有学术参照。同 HTTP 301 重定向之原理: 旧地址仍可访问, 只是指向新地址。

Reading the EpiDoc shell读懂 EpiDoc 外壳

An EpiDoc file is more than its edition. The teiHeader records:

一份 EpiDoc 文件远不止“释读”这一部分。teiHeader 还要记录:

• Who edited the file and when (<revisionDesc>)
• The licence under which it's shared (<availability>)
• The physical stone — material, dimensions, layout, lettering, condition (<msDesc>)
• The findspot and provenance history (<history>)
• The languages and the encoding conventions used (<encodingDesc>, <profileDesc>)

EpiDoc 文件不仅是刻文本身。teiHeader 中还记载: 编辑者与日期、共享许可、实物石头 (材料/尺寸/布局/字体/状况)、出土地与流传史、所用语言与编码规约。

A tablet from the theatre of Catania卡塔尼亚剧场出土的牌匾

A tablet of Proconnesian marble from the Marmara quarries, originally fixed to the base of a statue of the Genius (protective spirit) of the city of Catania. The stone was found on 18 May 1770, during the excavations of the ancient theatre conducted by the Principe di Biscari, and now lives in the Museo Civico di Catania.一块普罗孔尼苏斯大理石板(出自马尔马拉海的古采石场), 原本钉在卡塔尼亚城守护神(Genius)雕像的底座上。出土于 1770 年 5 月 18 日, 当时比斯卡里王子正主持发掘卡塔尼亚的古剧场。如今藏于卡塔尼亚市立博物馆。

The dedicator was Facundus Porfyrius Mynatidius, a vir clarissimus and consularis — the highest-ranking governor of late-Roman Sicily.奉献者是 Facundus Porfyrius Mynatidius, 身份为vir clarissimus(显贵元老)兼consularis(执政官级总督)，晚期罗马西西里岛的最高行政长官。

VERNANTIBVS
SAECVLIS DDD NNN
GENIO SPLENDIDAE VR-
BIS CATINAE
FACVNDVS PORFYRIVS
MYNATIDIVS V·
CONS · EIVSDEM

What does ddd · nnn mean?ddd · nnn 是什么?

On the stone you see d followed by two more small suspended dd, then n followed by two more nn. The intended Latin is dominorum nostrorum — "of our lords."

石上所刻是一个 d, 后跟两个略小、位置稍高的 dd, 再写一个 n, 后跟两个略小的 nn。所代表的拉丁文为 dominorum nostrorum，“我们诸位君主之”。

But why three d's and three n's? The word dominorum doesn't start with three d's. The repetition is a plural marker. Late-Roman convention: doubling/tripling the initial letter signals plurality of the noun. Three lords = three reigning emperors (a tetrarchic-or-later imperial configuration).

石上所见: d 后接两个小 dd, 然后 n 后接两个小 nn。所欲表达的拉丁文为 dominorum nostrorum“我等主上”。但 dominorum 并不以三个 d 开头。这种重复是复数标记: 晚期罗马惯例中, 首字母的双重或三重表示该名词为复数。三重 d = 三位在位元首 (四帝共治或更后期的多帝并立体制)。

Dating logic: three Ds and three Ns means three Augusti reigning simultaneously. That narrows the inscription to one of six joint-reign periods in the 4th–5th centuries: 337–340, 367–378, 379–383, 388–395, 402–408, or 421. The consensus settles on a 4th-century date.断代由此而出: 三个 D 加三个 N, 表示同时有三位奥古斯都在位。这把本铭的年代收紧到四、五世纪的六个并立期之一:337—340、367—378、379—383、388—395、402—408、或 421。学界共识是公元四世纪某一期。

Abbreviation mark, not abbreviation letters缩写标记, 而非缩写字母

If you encoded ddd naïvely as three letters inside <abbr>:

若把 ddd 老实当作 <abbr> 内的三个字母去编码:

<expan><abbr>ddd</abbr><ex>ominorum</ex></expan>

…you'd be claiming the stone carries three letters of the word dominorum. That's false. The first d belongs to the word; the second and third are plural-marker decoration. EpiDoc's <am> ("abbreviation mark") wraps the decoration:

……便等于宣称石上确刻了 dominorum 一词的三个字母，而这是错的。只有第一个 d 属于该词; 第二、第三个 d 不过是复数标记之装饰。EpiDoc 以 <am>(“缩写标记”)包住这层装饰:

<expan><abbr>d<am>dd</am></abbr><ex>ominorum</ex></expan>

Now the structure says: stone has d; that d is followed by a plural-marker device dd; the editor expands d to dominorum; the plurality is already attested by the <am>, so the expansion doesn't need to add a marker of its own.

现在的结构含义是: 石上有 d; 其后跟着复数标记装置 dd; 编者将 d 展开为 dominorum; 复数已由 <am> 承载, 展开中无需再重复表达。

The cutter's hand: Prag's lettering note observes that "A appears with both broken and straight bar; E appears both as a tall narrow standard E and as a vertical stroke with a single horizontal bar across the middle; V appears both as V and as Y. Interpuncts are mostly absent, except after VC and CONS in the final two lines." Letter sizes go 60–80 mm (line 1), then 40–60 mm (lines 2–4), then 50–70 mm (lines 5–7) — the cutter shrank, then grew the script again as space ran out.关于刻工之手: Prag 在字形说明里观察道:"字母 A 既见折角横, 也见直横; E 既作高瘦的标准 E, 也只作中横一笔的简化形; V 既作 V 也作 Y。间隔点 (interpunct) 多数省略, 只在末两行 VC 与 CONS 之后才出现。" 字高自上而下变化:第 1 行 60—80 毫米, 第 2—4 行 40—60 毫米, 第 5—7 行 50—70 毫米，刻工先缩字, 临到末尾空间将尽, 又把字放大了回去。

Latin grammar drives the encoding拉丁语法决定编码方式

If the stone simply read ddd with three full-sized letters and no convention of plural marking, you'd use the naïve encoding. But ddd in this context is a known formal device with a known meaning. The encoding should reflect what is grammatically true, not what looks superficially like three letters.

若石上仅刻三个等大 d 且无复数标记惯例, 才可用朴素编码。但此处 ddd 是一种已知的格式化装置, 具有明确含义。编码应反映语法事实, 而非表面看起来的“三个字母”。

Editor's commentary on cons · eiusdem (line 7): Jonathan Prag notes that "the precise significance of cons eiusdem is extensively debated and several alternatives have been suggested: the most likely interpretation is consularis eiusdem, signifying consularis of the same city — i.e. Catania, and by extension the province of Sicily." Encoding it as cons(ularis) rather than (e.g.) cons(ul) records the editor's decision, and links the inscription to the late-Roman provincial governance vocabulary.编者关于第七行 cons · eiusdem 的注: Jonathan Prag 指出, cons eiusdem 的精确所指学界争论已久; 较可信的释读是 consularis eiusdem, 即"同一城的执政官级总督"，此处指卡塔尼亚, 延伸而指整个西西里行省。在编码中把它展为 cons(ularis) 而非(例如)cons(ul), 即是把编者的判断写进了 XML, 也把这件铭文挂上了晚期罗马行省治理术语的索引。

v, c, cons三个简单缩写

<expan><abbr>v</abbr><ex>ir</ex></expan>
<expan><abbr>c</abbr><ex>larissimus</ex></expan>
<expan><abbr>cons</abbr><ex>ularis</ex></expan>

Single-letter (v, c) and truncation (cons) abbreviations both use the standard <expan> shape. No <am> needed — these letters genuinely are parts of vir, clarissimus, consularis.

单字母缩写 (v, c) 与截短缩写 (cons) 都使用标准 <expan> 形态, 无需 <am>，这些字母确实是 vir、clarissimus、consularis 的组成部分。

<persName> can span <lb><persName> 可跨 <lb>

The dedicator's name occupies lines 5 and 6: Facundus Porfyrius / Mynatidius. EpiDoc is structural, not visual — one name stays inside one <persName>, even when a line break interrupts:

奉献者之名横跨第 5、6 行:Facundus Porfyrius / Mynatidius。EpiDoc 关注的是结构, 不是版式，一个人的名字依然装在一个 <persName> 之内, 即使中间被换行所打断:

<lb n="5"/><persName type="attested">
  <name>Facundus</name> <name>Porfyrius</name>
  <lb n="6"/><name>Mynatidius</name>
</persName>

The <lb> sits between two siblings of the parent element. Two separate <persName> blocks would mean two different people — which is wrong: this is one man with three names.

<lb> 夹在父元素的两个子元素之间。若分成两个 <persName> 就意味着两个不同的人，实际上这是同一人三段名。

When a word breaks across lines单词跨行

Line 3 ends with ur-, line 4 begins with bis. Together: urbis ("of the city"). The line-break is internal to the word. Mark it:

第 3 行末为 ur-, 第 4 行开首为 bis; 合而读之即 urbis(“城的”, 属格)。换行发生在单词内部。如此标记:

<lb n="3"/>Genio splendidae ur
<lb n="4" break="no"/>bis <placeName>Catinae</placeName>

@break="no" is the EpiDoc equivalent of a hyphen at the end of a printed line. A renderer can show urbis as a linked word (with the original split visible) or as a single token, depending on the user's preference.

@break="no" 相当于印刷品行末连字号。渲染器可根据用户偏好将 urbis 显示为完整单词 (但可见原始分行) 或单一词项。

Exercise 03 assembled完整编码

<div type="edition" subtype="primary" xml:space="preserve" xml:lang="la">
  <ab>
    <lb n="1"/>Vernantibus
    <lb n="2"/>saeculis <expan><abbr>d<am>dd</am></abbr><ex>ominorum</ex></expan>
    <expan><abbr>n<am>nn</am></abbr><ex>ostrorum</ex></expan>
    <lb n="3"/>Genio splendidae ur
    <lb n="4" break="no"/>bis <placeName>Catinae</placeName>
    <lb n="5"/><persName type="attested"><name>Facundus</name> <name>Porfyrius</name>
    <lb n="6"/><name>Mynatidius</name></persName>
    <expan><abbr>v</abbr><ex>ir</ex></expan>
    <expan><abbr>c</abbr><ex>larissimus</ex></expan>
    <g ref="#interpunct">·</g>
    <lb n="7"/><expan><abbr>cons</abbr><ex>ularis</ex></expan>
    <g ref="#interpunct">·</g> eiusdem
  </ab>
</div>

One script, two languages一种字母, 两种语言

The Punic-speaking populations of Roman Tripolitania kept writing their Semitic language using Latin letters long after Phoenician script went out of use. Analogously to modern speakers of Cantonese sometimes writing in romanization. The technical label is Latino-Punic, ISO code xpu-Latn.

罗马时代的的黎波里塔尼亚, 操布匿语(闪族语)的居民, 即便腓尼基字母早已弃用, 仍长期以拉丁字母书写自己的母语。这种情形, 一如今日有些粤语使用者以罗马字记下口语。学界称之为“拉丁字母布匿文”(Latino-Punic), ISO 代码 xpu-Latn。

<div type="edition" xml:lang="xpu-Latn" xml:space="preserve">

The xml:lang attribute uses ISO 639's hyphen-suffix to indicate a non-standard script. Same convention as zh-Hant for Traditional Chinese or sr-Latn for Latin-alphabet Serbian.

罗马治下的的黎波里塔尼亚, 布匿语使用者长期用拉丁字母书写其闪族语言, 远晚于腓尼基字母的废止。这种现象学术上称拉丁字母布匿文 (ISO 代码 xpu-Latn)。同样的写法在 zh-Hant (繁体中文)、sr-Latn (拉丁字母塞尔维亚语) 中亦可见。

The standard funerary formula墓志套语

Roman tombs nearly always open with Dis Manibus ("to the divine shades") or Dis Manibus Sacrum ("sacred to the divine shades"). The Manes are the gods of the underworld — addressees of the dedication.

罗马墓志几乎都以 Dis Manibus(“献给亡灵”)或 Dis Manibus Sacrum(“献给亡灵, 神圣不可侵”)开篇。Manes 是冥府之神，此处即是奉献的对象。

<persName type="divine" ref="#manes">
  <w lemma="deus"><expan><abbr>D</abbr><ex>is</ex></expan></w>
  <w lemma="manes"><expan><abbr>m</abbr><ex>anibus</ex></expan></w>
</persName>
<w lemma="sacer"><expan><abbr>s</abbr><ex>acrum</ex></expan></w>

type="divine" classifies the Manes as gods. @ref="#manes" links to the project's authority list — every D M across the corpus resolves to the same divine entity. Sacrum stays outside the <persName>: it's an adjective qualifying the dedication, not part of the gods' name.

type="divine" 将 Manes 归为神祇类。@ref="#manes" 连接到项目权威表，语料库中所有 D M 都解析到同一神祇实体。Sacrum 留在 <persName> 之外: 它是形容词, 修饰整个奉献行为, 不是神名的一部分。

ISO 8601 duration on <date>机器可读的寿命数据

The stretch auo sanu n LXXX records the deceased's age — 80 years. Wrap as:

石上 auo sanu n LXXX 一段, 记录的是亡者的寿数，八十岁。如此包装:

那一段 auo sanu n LXXX 记录的是亡者的寿数，80 岁。如此包装:

<date type="life-span" dur="P80Y">
  <w lemma="auo">auo</w>
  <lb n="5"/><w lemma="sanuth">sanu</w>
  <lb n="6"/><w lemma="numerus"><expan><abbr>n</abbr><ex>umero</ex></expan></w>
  <num value="80">LXXX</num>
</date>

@dur="P80Y" is ISO 8601 duration format: "Period of 80 Years." Now «find every tombstone of someone who died older than 70» becomes one SPARQL query.

@dur="P80Y" 是 ISO 8601 时段格式: “80 年之期”。“找出所有 70 岁以上的墓主”即变为一条 SPARQL 查询。

Saying how much is missing陈述缺失多少

The bottom-left of Nabor's stone is damaged. The editor estimates 2, 2, and 4 lost characters at the start of lines 8, 9, 10:

Nabor 之石的左下角已残。编者据残痕估计:第 8、9、10 行开首分别缺失约 2、2、4 个字母:

<lb n="8"/><gap reason="lost" quantity="2" unit="character"/><orig>milim e</orig>
<lb n="9"/><gap reason="lost" quantity="2" unit="character"/><orig>duo</orig><space quantity="1" unit="character"/>
<lb n="10"/><gap reason="lost" quantity="4" unit="character"/><orig>s</orig>

The triple @reason / @quantity / @unit tells software what kind of absence (lost vs. illegible vs. omitted), how much (a count), and of what (characters, lines, words). This is the standard EpiDoc lacuna shape.

三元组 @reason / @quantity / @unit 告诉软件: 是什么类型的缺失 (失落、不可辨、省略)、缺失多少 (计数)、缺失什么 (字符、行、词)。这是 EpiDoc 标记残缺的标准形态。

Editor's commentary (Reynolds & Ward-Perkins, IRT 1952): "Lines 8–10: the left-hand margin of the stone is missing but it is not clear whether any letters have been lost." This is the canonical reason to use <gap reason="lost" extent="unknown"/> rather than a quantified gap — the editor honestly doesn't know.编者注 (Reynolds 与 Ward-Perkins, IRT 1952): "第 8—10 行:石头的左缘已残, 但是否确有字母丢失, 尚不明朗。" 这正是该用 <gap reason="lost" extent="unknown"/> 而非带具体数量的阙文的标准理由，编者老实承认自己不知道丢了几个字。

<orig> as a semantic placeholder<orig> 作为语义占位

The lower lines of Nabor's stone contain Punic words (nsath fo, milim e, duo, trailing s) that scholars cannot securely translate. Mark them as original-as-received:

Nabor 之石下半部分有几组学者尚不能确译的布匿词汇(nsath fo、milim e、duo, 以及末尾的 s)。将其标记为“原貌, 一仍其旧”:

<orig>nsath <unclear>fo</unclear></orig>

<orig> says: "I'm keeping this exactly as the stone has it, without normalising or interpreting." Its sibling <reg> would supply a regularised reading. Using <orig> alone is honest about ignorance — and leaves the door open for a future Punic specialist to add <reg> later.

<orig> 的含义是: “原样保留, 不正规化, 不解释”。其同伴标签 <reg> 提供规范化读法。仅用 <orig> 即为诚实地承认无知, 并为将来精通布匿语的学者预留补充 <reg> 的空间。

The little panel on the side侧面板

Nabor's stone has a second small inscribed panel (b) that carries only ny / fo. Encode each panel as a <div type="textpart">:

Nabor 之石另有一小面板(称为 b 面板), 仅刻 ny / fo 数字。两个面板各以一个 <div type="textpart"> 编码:

<div type="edition" xml:lang="xpu-Latn" xml:space="preserve">
  <div subtype="section" n="a" type="textpart"><ab>…ten lines…</ab></div>
  <div subtype="section" n="b" type="textpart">
    <ab><lb n="1"/><orig>ny</orig> <lb n="2"/><orig>fo</orig></ab>
  </div>
</div>

The @n="a" / @n="b" labels let apparatus entries refer to specific panels: <app loc="b.1">. We'll meet this pattern again in Exercise 07.

@n="a" / @n="b" 标签让校勘记可针对具体面板引用 (如 <app loc="b.1">)。此模式在练习 07 中将再次出现。

The main text — what IRT calls panel a — fills the moulded frame from top to bottom. But look at line 7: where the cutter ran out of words, someone later squeezed in a second small inscription, panel b, in smaller letters in the vacant space.主文(IRT 称之为“a” 面板)在凹陷的框线内自上而下占满整个版面。但看第七行，主文写不下了, 留下一小块空白, 后来又有人用更小的字, 在那块空白里挤进了第二段铭文, 即“b” 面板。

Panel b reads: NY / FO — two lines, two letters each. The editors' verdict: "not translatable." J. M. Reynolds left both English translations bracketing this fact: "b. (not translatable.)"“b”面板所刻: NY / FO，两行, 每行两个字母。J. M. Reynolds 的翻译至此搁笔, 只写了一句 "b. (not translatable.)"，不可译。

In the XML this is encoded as a second <div type="textpart" subtype="addition">, sibling to the main text. The encoding records the social fact: a stone's text can grow after the first carving.在 XML 中, 这一段被编为第二个 <div type="textpart" subtype="addition">, 与主文并列。编码所记录的, 不仅是文本, 还是一桩社会事实:一块石头上的字, 在首刻之后, 还会继续生长。

What the stone shows石上所刻

The leading L on lines 1 and 6 is the year-symbol. The dashes mark missing parts of the date and the patronymic.
第 1 和第 6 行前的 L 是“年”字符号; 破折号表示日期与父名中残缺的部分。

The panel measures roughly 24.5 × 33 cm, outlined in low relief, with letters about 4 cm high. L appears as the symbol for ἔτους ("year") and ἐτῶν ("aged"); the sigma is squared. Surface erosion has eaten most of the year and month name.面板约 24.5 × 33 厘米, 四周有浅浮雕边线, 字高约 4 厘米。L 作为符号同时代表 ἔτους(年)与 ἐτῶν(岁); sigma 作方形。石面已严重风化, 大半的年序与月份名都被风蚀吃掉。

Interpretive reading (Reynolds 1983):
(ἔτους) [.] Με[χ-]
ὶ[ρ---]
Νείκαις
Ἀνδ̣[ροσ-]
θένευς
(ἐτῶν) ιε΄
"Year ?, Mecheir ?, Neikai(o)s [scil. son] of Androsthenes, aged 15."Reynolds 1983 之释文:
(ἔτους) [.] Με[χ-]
ὶ[ρ---]
Νείκαις
Ἀνδ̣[ροσ-]
θένευς
(ἐτῶν) ιε΄
"第?年, Μεχίρ 月某日。Androsthenes 之子 Neikai(o)s, 享年十五。"

One little glyph, six layers of XML一个小符号, 六层 XML

The L-shaped tick at the start of the date is shorthand for ἔτους ("year, gen."). It's an abbreviation, but instead of a letter, the abbreviation IS a glyph. The canonical encoding is six layers deep:

日期开头那个 L 形的小符号, 即代表 ἔτους(“年”, 属格)。它确是缩写, 只是缩去的不是字母，缩写本身就是一个图形符号。标准编码共有六层:

<w lemma="ἔτος">
  <expan>
    <abbr>
      <am>
        <g ref="symbols.xml#year"/>
      </am>
    </abbr>
    <ex>ἔτους</ex>
  </expan>
</w>

Read from inside out: glyph → abbreviation mark → abbreviation → expansion → expansion-device → word. Each layer adds a different kind of information.

由内而外读: 符号 → 缩写标记 → 缩写形 → 编者展开 → 整个缩写装置 → 词项。每一层都贡献不同类型的信息。

Found in context, encoded in isolation: on the rock wall, this L is one of several markers on a single façade that holds nine inscriptions (T.451–459). The encoding strips it from its tomb and gives it six nested elements so that across IRCyr2020's ~2,300 inscriptions, every "year-symbol followed by numeral" can be searched, sorted, and dated as one consistent kind of object.同壁数铭, 各自成档: 在原石上, 这个 L 不过是同一面崖壁(T.451—459 共九块面板)中的一个小记号。EpiDoc 把它从墓壁剥离, 套上六层嵌套, 使 IRCyr2020 全集约 2,300 件铭文中所有"L 符 + 数字"的组合, 都能作为同一类对象被检索、排序、断代。

Each layer earns its place每层皆有所司

• <g> — the glyph itself, defined in the corpus's character list.
• <am> — "this glyph IS the abbreviation mark for the word." Without <am>, the glyph would look like decoration.
• <abbr> — what appears on the stone, in lieu of the spelled-out word.
• <ex> — the editor's expansion, the actual Greek word.
• <expan> — the device as a whole, what enables the renderer to switch between «show the glyph» and «show ἔτους».
• <w> — the word as a lexical unit, with its dictionary form in @lemma.

每一层都对应一种独立的信息: 实物符号 / 标记功能 / 石面所见 / 编者展开 / 整体装置 / 词条归并。缺一不可。

Restoring Μεχίρ补足 Μεχίρ

The month name Μεχίρ (Mecheir, ≈ Jan–Feb) is split across lines 1 and 2 with two letters missing:

月份名 Μεχίρ(Mecheir, 约公历一二月间)跨第 1、2 行, 中间缺失两个字母:

<rs type="month" key="mechir">
  <w lemma="Μεχίρ">
    Με<supplied reason="lost">χ</supplied>
    <lb n="2" break="no"/>
    ὶ<supplied reason="lost">ρ</supplied>
  </w>
</rs>

Two <supplied reason="lost"> elements restore the missing letters. The <lb n="2" break="no"/> says the word continues across the line break. The whole sequence is wrapped in <rs type="month" key="mechir"> — a referring string identifying this stretch as the Egyptian month Mecheir.

两处 <supplied reason="lost"> 补出缺失字母; <lb n="2" break="no"/> 标示单词跨行延续; 整段包于 <rs type="month" key="mechir"> 中, 指明此为埃及历 Mecheir 月。

Why "Mecheir" (Μεχίρ)? An Egyptian month — the sixth in the calendar reformed under Augustus — roughly 26 January to 24 February. Reynolds saw the first two letters on the stone and supplied the rest. Mecheir dating in Cyrenaica is a thin but real strand of the calendar continuum between Egypt and the Greek-speaking world.为何作 "Mecheir" (Μεχίρ)? 此乃埃及历的第六月; 奥古斯都改历之后, 大致对应公历 1 月 26 日—2 月 24 日。Reynolds 在石上仅辨出 Mecheir 的头两个字母, 余皆由她补足。在昔兰尼加, Mecheir 系年的铭文虽稀, 却确是埃及与希腊语世界之间历法接续的一脉。

When you can't count, say so数不清时, 明言之

After the month-name, the day-number is gone. We don't know whether it was 7, 17, or 27. Use extent="unknown", not a fake quantity:

月份名之后, 日期数字已佚。我们无从分辨是初七、十七还是二十七。此时应用 extent="unknown", 而不是凭空填一个 quantity:

<gap reason="lost" extent="unknown" unit="character"/>

This is editorial honesty made queryable. A statistical study can later ask «what fraction of dates in this corpus have lost day-numbers?» and get a real answer.

这是把编者诚实变为可查询的属性。统计研究可问“语料库中有多大比例的日期缺失日数?”并获得真实答案。

godot.date — every date has a URLgodot.date，每个日期都有 URL

The whole date construction sits inside a <date> element whose @ref points to godot.date, an authority project that gives each ancient date a stable identifier:

整段日期被包入一个 <date> 元素, 其 @ref 指向 godot.date，一个为每一古代年代分配稳定标识符的权威项目:

<date ref="https://godot.date/id/BcwJL846woVx95NLsfd9PK">
  <!-- L (year-symbol) gap-1 Mεχίρ gap-unknown -->
</date>

Even though this date is partly lost (we don't know the regnal year or the day), the editor's best estimate is tied to a canonical godot.date entry. Linked data in practice: two independent scholarly projects (the inscription corpus and the date authority) talk through stable identifiers.

尽管此处日期部分残缺 (王年和日期皆缺), 编者的最佳估计仍挂在 godot.date 的规范条目上。这就是实践中的链接数据: 铭文语料库与日期权威, 通过稳定标识符对话。

This inscription's actual godot.date URL: godot.date/id/BcwJL846woVx95NLsfd9PK ↗. Click and you see: regnal year unknown, month Mecheir ("month 6 of the Cyrenaican / Alexandrian calendar"), every date attestation Reynolds 1983 records, and links to other inscriptions Reynolds dated the same way.本铭对应的 godot.date 链接: godot.date/id/BcwJL846woVx95NLsfd9PK ↗。点开即可见:在位年序不详、月份为 Mecheir(昔兰尼加—亚历山大历第六月)、Reynolds 1983 所记的所有相关日期标注, 以及由她同样断代的其他铭文之链接。

Nested <persName>嵌套 <persName>

Greek family identity is a chain of genitives: "Neikais of Androsthenes of Lysippos of …". EpiDoc encodes this with nested <persName>:

希腊式的家世表述是一条属格的链:“Neikais, Androsthenes 之子, Androsthenes 又为 Lysippos 之子……”EpiDoc 用嵌套的 <persName> 编码这种关系:

<persName type="attested" key="lgpn:V1-60818">
  <name nymRef="Νίκαιος">Νείκαις</name>
  <persName type="attested" key="lgpn:V1-60027">
    <name nymRef="Ἀνδροσθένης">Ἀν<unclear>δ</unclear>
      <supplied reason="lost">ροσ</supplied>
      <lb n="5" break="no"/>θένευς</name>
  </persName>
</persName>

Outer <persName> = the deceased. Inner = the father. A query for «all sons of Androsthenes in the Cyrenaican corpus» can find them even when the father appears only as a genitive modifier.

外层 <persName> = 死者; 内层 = 父亲。“在昔兰尼加语料库中找出所有 Androsthenes 的儿子”即可查到, 哪怕父亲名仅以属格修饰语形式出现。

A genitive of paternity: on a Greek funerary stone, X (son/daughter) of Y is the standard formula. Here Neikais' father is Androsthenes; the stone shows only "Ἀνδ̣[ροσ-]θένευς" — genitive case, with the medial consonant supplied. Nesting Androsthenes' inner <persName> inside the outer one for Neikais means a SPARQL query can find "all children of Androsthenes in Cyrenaica" without ever parsing prose.属格表父名: 希腊墓志的标准套语是“某人, 某人之子(女)”。此处 Neikais 之父名为 Androsthenes, 石上仅刻 "Ἀνδ̣[ροσ-]θένευς"，取属格, 中间辅音由编者补出。在 XML 中, Androsthenes 的 <persName> 嵌套于 Neikais 的 <persName> 之内; 这使日后用 SPARQL 查询"昔兰尼加所有 Androsthenes 之子女"时, 不必再解析散文文本, 一查即得。

ιε = 15ιε = 15

Greek used its letters as numerals: α = 1, β = 2, …, ι = 10, ε = 5, so ιε = 15. Wrap as <num value="15">ιε</num> inside a life-span date:

希腊文以字母兼作数字:α = 1、β = 2、……、ι = 10、ε = 5, 故 ιε = 15。在记寿数的日期之内, 包装为 <num value="15">ιε</num>:

<lb n="6"/><date type="life-span" dur="P15Y">
  <w lemma="ἔτος"><expan><abbr><am><g ref="symbols.xml#year"/></am></abbr><ex>ἐτῶν</ex></expan></w>
  <num value="15">ιε</num>
</date>

The L-symbol now expands to the genitive plural ἐτῶν (not ἔτους) because it goes with "years" in the plural after the numeral.

同一个 L 符号, 此处展开为属格复数 ἐτῶν(而非单数属格 ἔτους)，因为接在数字之后, 与"年"取复数。

“年”字符号此处展开为属格复数 ἐτῶν (不是 ἔτους), 因数字之后接复数。

An elegiac couplet for Palmatos献给 Palmatos 的哀歌对句

Translation (Roueché). «The Carians, remembering many benefits, and greatly admiring the rightly just Palmatos, [set up this statue].»

译文 (Roueché): “卡里亚人感念其无数恩泽, 深慕公允无私之 Palmatos, [立此像]。”

Survives only in the Planudean Anthology (Anth. Pal. 16.35). The Aphrodisias editors believe the stone was originally inscribed at Aphrodisias and later copied by a Byzantine scholar.

<lg> and <l>诗节组与诗行

Prose stretches go inside <ab>. Verse stretches go inside <lg> ("line group"), with individual lines as <l>:

散文段落装入 <ab>; 韵文段落装入 <lg>(line group, 行组), 每一行单独作 <l>:

<lg met="alegaic?">
  <l n="1" met="hexameter">…</l>
  <l n="2" met="pentameter?">…</l>
</lg>

The @met attribute records the metrical pattern. A question-mark in the value ("alegaic?") signals editorial uncertainty — kept literally in the value, not as a separate @cert attribute, by convention in the Aphrodisias corpus.

散文段落用 <ab>; 诗歌段落用 <lg> + 内含的 <l>。@met 记录格律; 值中的问号 ("alegaic?") 标示编者不确定，阿弗洛狄西亚语料库的惯例。

«The Carians» — three layers“卡里亚人”，三层包装

<rs type="subject" key="koinon">
  <w lemma="Κᾶρ">
    <placeName type="ethnic" full="yes" reg="Κάρ">Κᾶρες</placeName>
  </w>
</rs>

Three layers, three facts:

三层包装, 各陈一桩事实:

• <rs type="subject" key="koinon"> — Functionally, this is the dedicator: the koinon, the corporate federation of Carian cities.
• <w lemma="Κᾶρ"> — Lexically, it's a word.
• <placeName type="ethnic"> — Grammatically, it's a noun derived from a place-name. «Ethnic» is a technical term in epigraphy for inhabitant-nouns: Athenians, Romans, Spartans, Carians.

三层、三种事实: 功能上是奉献者 (Carian koinon, 卡里亚共同体); 词汇上是一个词; 语法上是由地名衍生的“民族名”。“Ethnic”是铭文学专业术语, 指由地名衍生的居民称谓 (雅典人、罗马人、斯巴达人、卡里亚人)。

@reg recovers the nominative@reg 恢复主格

<persName key="Palmatus" type="aphrodisian" full="yes">
  <name reg="Παλμᾶτος">Παλμᾶτον</name>
</persName>

The inscribed form Παλμᾶτον is the accusative ("admiring Palmatos"). The @reg attribute supplies the canonical nominative Παλμᾶτος, so a database query for the name finds this attestation despite the case ending.

石上所刻 Παλμᾶτον 是宾格形式(“敬慕 Palmatos”)。@reg 属性补出该名的规范主格 Παλμᾶτος; 这样, 数据库无论以哪种格变化检索, 都能从此铭文中找到这条记录。

Compare with Exercise 05's @nymRef on <name> and key="lgpn:..." on <persName>: different attributes for different authority projects, all serving the same canonicalisation purpose.

石上 Παλμᾶτον 为宾格。@reg 给出规范主格 Παλμᾶτος, 使数据库查询不受格变化影响。比较练习 05 的 @nymRef 与 key="lgpn:...": 不同权威项目用不同属性, 但目的相同，规范化。

One lemma per word每词一个词典原形

Every Greek word in this couplet has a @lemma on its <w> wrapper:

这组对句中的每一个希腊词, 其 <w> 包装层上都带 @lemma 属性:

<w lemma="μνήμων">μνήμονες</w>
<w lemma="πολύς">πολλέων</w>
<w lemma="εὐεργεσία">εὐεργεσιάων</w>
<w lemma="ἰθυδίκης">ἰθυδίκην</w>
<w lemma="τόσος">τόσσον</w>
<w lemma="ἄγαμαι">ἀγασσάμενοι</w>

Without lemmatization, a search for «benefactions» (εὐεργεσία) would miss this inscription because the inflected form on the stone is the genitive plural εὐεργεσιάων. With @lemma, every form of the word is findable through the dictionary entry.

没有词形归并, 搜索“恩泽”(εὐεργεσία) 会漏掉此铭文，石上形式为属格复数 εὐεργεσιάων。有了 @lemma, 任何形式都能通过词典原形找到。

The choice of meter is a statement格律之选, 即一种宣言

By the late 5th century CE, prose honorifics were the norm. Choosing a hexameter+pentameter elegiac couplet for Palmatos signals the late-antique cultural ethos of paideia: the honorand is being celebrated as a man of education, and the inscription's medium is itself a tribute to the literary culture he and his honorers shared.

至公元五世纪晚期, 颂辞已多以散文写成。为 Palmatos 题写一组六音步配五音步的哀歌对句, 是在示意一种古代晚期的paideia(古典教养)风尚:受颂者被作为一位“学养之士”来表彰, 而铭文本身的形式, 又正是对他与其颂者共同分享的文学文化的一份致敬。

The vocabulary is consciously Homeric — μνήμων, εὐεργεσία, ἰθυδίκης — placing Palmatos in the line of heroic exemplars. EpiDoc records this through @met + @lemma; the historical reading is the encoder's commentary.

公元 5 世纪晚期, 散文荣誉铭已是常态。选择六步格+五步格的哀歌对句体彰显古代晚期 paideia (博雅教育) 文化精神: 颂主作为有教养之人受表彰, 铭文形式本身即向其与颂主所共享的文学文化致敬。词汇有意采用荷马式 (μνήμων、εὐεργεσία、ἰθυδίκης), 把 Palmatos 置于英雄序列。EpiDoc 通过 @met + @lemma 记录其形式; 历史解读则由编者注释呈现。

One edition, two textparts一份刻文, 两个语段

<div type="edition" subtype="primary" xml:space="preserve">
  <div type="textpart" subtype="section" n="a" xml:lang="grc">
    <ab><lb n="1"/><w>ἀ</w> <w>ὠ</w></ab>
  </div>
  <div type="textpart" subtype="section" n="b" xml:lang="la">
    <ab>…Latin lines…</ab>
  </div>
</div>

The outer <div type="edition"> intentionally lacks @xml:lang — language is set on each <textpart>. A bilingual renderer can apply different fonts and bidirectional behaviour per panel.

外层 <div type="edition"> 不设 @xml:lang; 语言在每个 <textpart> 上局部声明。双语渲染器可对不同面板应用不同字体和文本方向。

<textLang> with sociolinguistic analysis语言层面的社会语言学标注

<textLang mainLang="la" otherLangs="grc"
  ana="#bilingualism.bilingual-phenomena.code-switching.tag-switching">
  Latin and Ancient Greek
</textLang>

@ana (analysis) points to a hierarchical taxonomy of bilingualism phenomena. "Tag-switching" is a specific contemporary-linguistics term: inserting a short token from one language while the main text stays in another. The Greek Α Ω is a religious tag inside a Latin biographical inscription. EpiDoc captures this as sociolinguistic data.

@ana 指向一个层级化的双语现象分类。“Tag-switching” (标签转换) 是当代语言学专业术语, 指主体保持一种语言, 仅插入另一语言的短小标签。希腊文 Α Ω 是嵌入在拉丁文传记中的宗教标签。EpiDoc 把它当作社会语言学数据来捕捉。

Why this stone matters sociologically: Selinus in late antiquity sat at a confessional boundary. The Museo Salinas placard cites Rossius (ap. Salinas, then Mommsen, then Bivona) on a hypothesis worth knowing: "The presence of a deacon in a locality that was not an episcopal see and that did not have a prominent role in the Roman period remains problematic. The name Ausanius (Auxanius outside Africa) has been related to Ausana, an episcopal see in proconsular Africa. This has led to the hypothesis that a group of African Christians might have settled in Selinunte after fleeing the persecutions of Arian Vandals." Encoding the bilingualism therefore touches a real historical question.这块石头的社会史价值: 古代晚期的塞利努斯正位于一处宗派分界线上。萨利纳斯博物馆展签转述 Rossius(后经 Salinas、Mommsen、Bivona 沿袭)的一个值得记住的假说:“这位执事(deacon)出现于一处既非主教座、又在罗马时期并不显赫的小地, 颇为蹊跷。Ausanius 这个名字(在非洲以外多作 Auxanius)曾被联系到非洲行省的主教座 Ausana。由此一说便有此推测:一批北非基督徒因逃避阿里乌斯派汪达尔人的迫害, 在塞利努斯定居下来。” 因此, 对这件铭文的双语编码所触及的, 是一桩真实的历史问题。

Α Ω is not an abbreviationΑ Ω 不是缩写

The Greek letters Α and Ω on the upper panel might look like they abbreviate Greek words. They don't. They are a theological symbol: «I am the Alpha and the Omega, the beginning and the end» (Rev 1:8). Encoding them with <expan> would falsely imply they're shorthand for spelled-out words.

上部面板上的两个希腊字母 Α 与 Ω, 乍看像是某词的缩写, 其实并不是。它们是一个神学符号:“我是阿尔法, 我是奥米伽, 是始, 是终”(《启示录》 1:8)。若用 <expan> 来编码它们, 便等于误称它们是某个被省略之词的缩写形。

<ab><lb n="1"/><w>ἀ</w> <w>ὠ</w></ab>

The religious meaning lives in <div type="commentary">. The encoding records only what the stone shows: two letters wrapped as words.

希腊字母 Α 与 Ω 看似缩写, 实非。它们是神学符号: “我是阿尔法, 我是俄梅伽, 是首先的, 是末后的”(《启示录》1:8)。若用 <expan> 会错误地暗示它们缩写了某词。神学意义放在 <div type="commentary"> 里; 编码只记录石上所见: 两个被包装为“词”的字母。

The Α / Ω formula in late-antique Christian epigraphy: drawn from Revelation 1:8 ("I am the Alpha and the Omega"), it stood for Christ as beginning and end. Carved as two solitary Greek letters around a Christogram, A and Ω are quotations, not abbreviations — there is no Latin word being shortened. Encoding them inside a <foreign> wrapper rather than <expan> records this distinction, which matters when a corpus query later asks "how many Sicilian inscriptions cite Revelation?"晚期基督教铭文中 Α / Ω 公式的来历: 出自《启示录》1:8 “我是阿尔法, 我是奥米伽”, 借以宣告基督为始与终。两个孤立的希腊字母围绕一个基督单字符号被刻出, 它们是引文, 不是缩写，没有任何拉丁文词被它们省略。在 XML 中, 它们装入 <foreign> 而非 <expan>，这一区分日后查询"西西里铭文中有多少件引用《启示录》"时就派上用场了。

<roleName> for the deacon用 <roleName> 标记执事

<roleName type="religious" subtype="diaconus">
  <w>diaconus</w>
</roleName>

<roleName> wraps functions, ranks, offices — anything someone does, as opposed to what they are named. @type distinguishes religious roles from civic (magistracy), military, or honorific. @subtype pinpoints the specific office.

A query like «every early Christian deacon attested in Sicily» becomes one SPARQL filter.

<roleName> 包装职能、品级、官职，一个人做什么而非叫什么。@type 区分宗教/市政/军事/荣誉; @subtype 精确到具体职位。“西西里早期所有有铭文证据的执事”即可一查而得。

"Diaconus" — a clerical ladder rung: in the 4th-5th-century Latin Church, a diaconus was the third rank, below bishop and presbyter. Encoding "Diaconus" as <roleName type="religious"> rather than as part of the name lets a prosopographical query distinguish deacons from priests from bishops, and lets the same XML feed a future map of clerical offices across late-Roman Sicily.“Diaconus” — 教阶中的一级: 在四、五世纪的拉丁教会中, diaconus(执事)是第三级, 居主教与司铎之下。在 XML 中把 "Diaconus" 装入 <roleName type="religious"> 而非视作名字的一部分, 便能让人物志检索得以区分执事、司铎、主教; 同一份 XML 日后也可用来绘制晚期罗马西西里的教会职务地图。

The middle is supplied, the ends are on the stone头尾在石上, 中间由编者补

Line 7 has Ian s: the Ian is on the stone, then a gap (the middle of Ianuarias is left out), then s is on the stone again. EpiDoc handles this with multiple <abbr> and <ex> children inside one <expan>:

第 7 行有 Ian s:石上刻 Ian, 中间空一段(Ianuarias 的中段未刻), 然后又刻 s。EpiDoc 把这种结构编为一个 <expan> 之内多个 <abbr> 与 <ex> 交替排列的样子:

<w><expan><abbr>Ian</abbr><ex>uaria</ex><abbr>s</abbr></expan></w>

The reader sees Ian(uaria)s, with parentheses only around the supplied middle. The structure says: stone, editor, stone — three children, alternating.

读者看到 Ian(uaria)s, 括号只包中间补出部分。结构含义: 石上、编者、石上，三个子元素交替排列。

What survives, what's restored残存与补字

 1. [ — ] τὸν κόσμον ἅπαντα Ε+
 2. [ — ]εως καὶ Βασιλίδος
 3. [ἀμέμ]πτως καὶ ἀκαταγνώ-
 4. [στως ἔζησ]αν μοι ἔτη  ιδ̄  ἐν
 5. [—  ζήσ]ασα ἔτη  κη̄  ὁρκί-
 6. [ζω τὸν Παντο]κράτορα καὶ
 7. [— τὸν μέλλ]οντα αἰῶνα
 8. [— μη]δὲν ἀνῦξαι
 9. [— ἐκεί]νων κατ[ά-
10. [θεσιν —]
  

Translation: «…all the world…of (name) and of Basilis. They lived perfectly and blamelessly with me fourteen years…lived twenty-eight years. I beseech in the name of the Almighty…for all eternity…let no one open [the tomb]…»

译:“……整个世界……(某人)与 Basilis 的。他们与我同居十四年, 全然无可指摘……享年二十八岁。我以全能者之名恳请……直至永远……勿使人启 [此墓]……”

Ages: ιδ̄ (14), κη̄ (28)年龄: 14 与 28

<num value="14"><hi rend="supraline">ιδ</hi></num>
<num value="28"><hi rend="supraline">κη</hi></num>

The supraline (overline) is the standard Greek scribal convention to flag a letter functioning as a numeral. ι alone could be the word «you (plural)» or the numeral 10 — the overline disambiguates. EpiDoc records both layers: the visual mark with <hi rend="supraline">, the numeric meaning with @value.

字母上方的上横线(supraline)是希腊抄写者的标准惯例, 用以提示“此字此处作数字”。ι 单独出现, 可能是“你们”一词, 也可能是数字 10，上横线即用来消除这种歧义。EpiDoc 同时记录两层:视觉符号以 <hi rend="supraline"> 标出, 数字含义则由 @value 携带。

Notice the deliberate <space quantity="1" unit="character"/> surrounding each numeral on the stone — the engraver gave them breathing room as visual markers.

上划线是希腊文标示“此字母作数字用”的标准方式。ι 单独出现可能是“你们”或数字 10, 上划线消除歧义。EpiDoc 同时记录视觉标记 (<hi rend="supraline">) 与数值含义 (@value)。注意数字两侧的 <space>，刻工有意留出空白以作视觉强调。

Word-internal continuations单词跨行延续

Lines 3–4 carry the formula ἀμέμπτως καὶ ἀκαταγνώστως «blamelessly and irreproachably.» Survives only as …πτως καὶ ἀκαταγνώ-/στως. The encoding restores the lost letters and marks the line-internal break:

第 3、4 行有一句套语 ἀμέμπτως καὶ ἀκαταγνώστως(“无可指摘, 无可责备”)。残存形式只剩 …πτως καὶ ἀκαταγνώ-/στως。编码把佚失字母补回, 并标记词内的换行:

<lb n="3"/><w><supplied reason="lost">ἀμέμ</supplied>
  <unclear>π</unclear>τως</w> <w>καὶ</w>
  <w>ἀκαταγνώ
<lb n="4" break="no"/><supplied reason="lost">στως</supplied></w>

The opening of ἀμέμπτως is supplied, the middle π is unclear, the end στως across line 4 is also supplied. <lb break="no"/> ties the two halves into one word.

ἀμέμπτως 词首由编者补出, 中间 π 标为存疑, 跨行的词尾 στως 亦由编者补出。<lb break="no"/> 把两半连缀为一词。

When you're guessing, say so推测时, 明言之

<lb n="9"/><gap reason="lost" extent="unknown" unit="character"/>
<w><supplied reason="lost" cert="low">ἐκεί</supplied>νων</w>
<w>κατ<supplied reason="lost">ά</supplied>
<lb n="10" break="no"/><supplied reason="lost" cert="low">θεσιν</supplied></w>

Lines 9–10 restore [ἐκεί]νων κατ[ά-/θεσιν]. Other scholars proposed alternatives (Ferrua suggested κατετέθη «was laid down»; Manganaro suggested κατάγειον μνημείον). The chosen reading goes in the edition with cert="low"; competing readings go in the apparatus.

第 9、10 行所补的 [ἐκεί]νων κατ[ά-/θεσιν] 并非唯一方案。Ferrua 提议 κατετέθη(“被葬下”), Manganaro 则提议 κατάγειον μνημείον(“地下纪念室”)。本电子版编者选取其一入释读栏, 并标 cert="low"; 其余诸说则置于校勘栏。

Where the scholarly conversation lives学术对话之所

<div type="apparatus" resp="#martaf">
  <listApp>
    <app loc="line 1"><note>Manganaro: ---[τ]ὸ<unclear>ν</unclear> κόσμον ἐ<unclear>ν</unclear>. After ε, Korhonen reads…</note></app>
    <app loc="line 2"><note>Libertini, Ferrua, Korhonen: [βασιλ]έως καὶ βασιλίδος; Manganaro: [τύμβῳ (name)]εως καὶ βασιλίδος</note></app>
    <app loc="lines 4-5"><note>Libertini: ἐν | [ζήσ]ασα; Ferrua: ἐν | [γάμῳ ἅπ]ασα; Manganaro: ἕν, (?) | [ἑτέρα πληρώσ]ασα</note></app>
  </listApp>
</div>

Modern editions rarely show only one reading. The chosen reading goes in <div type="edition">; competing readings are recorded in <app> entries with @loc pointing to specific lines. The encoder's job is to make both layers machine-readable.

现代版本几乎不会只展示一种读法。所选读法置于 <div type="edition">; 其他读法记录在 <app> 中, 用 @loc 指向具体行。编码者的任务是让两层都机器可读。

The «μηδὲν ἀνῦξαι» formula“不要打开 (墓)”之套语

Lines 6–8 invoke the Pantokrator (Almighty) and threaten anyone who opens the tomb: «I beseech the Almighty…that no one open [it].» This is a well-known late-antique Christian feature, studied notably by Denis Feissel, BCH 104 (1980): 464–470.

Three reasons curses appeared in Christian epitaphs:

基督教墓志何以频频出现这种诅咒之辞, 大致有三层缘由:

• Theological. The body matters for the bodily resurrection; disturbing it carries eschatological consequences.
• Practical. Tomb reuse and grave-good theft were widespread; legal sanctions were weak.
• Cultural. The curse genre was inherited from pagan funerary practice; Christianity adapted rather than abolished it.

基督教墓志诅咒侵墓者, 古代晚期常见现象 (经典研究: Denis Feissel, 《BCH》104 (1980): 464–470)。三个动机: 神学 (身体复活之重要性)、实用 (盗墓常见而法律弱势)、文化 (沿用异教葬俗的诅咒传统)。

What is a Roman career inscription?罗马“履历铭”之意涵

A Roman senator's career was a ladder of public offices held in a more-or-less fixed sequence — quaestor, tribunus plebis, praetor, consul, plus optional priesthoods, military commands, and provincial governorships. An honorific inscription listing these offices in order was a cursus honorum, «course of honours.»

The inscription thus functions as a CV in stone. It is the principal source for Roman prosopography: who held what office when, and in what sequence.

罗马元老的仕途是一系列公职的阶梯, 大致按固定顺序晋升: 财务官 → 平民保民官 → 法务官 → 执政官, 另可兼任祭司、军事指挥与行省总督等。荣誉铭中按序排列这些官职即为“荣誉之路”。铭文实际上充当石上简历, 是罗马人物志研究的主要史料，谁、何时、按怎样的顺序担任何职。

Roman tria nomina罗马三段名

<persName type="attested" key="pir-p-0749">
  <name type="praenomen" nymRef="Quintus"><expan><abbr>Q</abbr><ex>uintus</ex></expan></name>
  <name type="gentilicium" nymRef="Pomponius">Pomponius</name>
  <name type="cognomen" nymRef="Rufus">Rufus</name>
</persName>

• praenomen — personal first name (Quintus, often abbreviated as Q.)
• gentilicium — clan name (Pomponius — the gens)
• cognomen — family-branch name (Rufus — the cognomen)

• praenomen(个人名)，即位于最前的第一个名字(此处为 Quintus, 常缩为 Q.)。
• gentilicium(氏族名)，表家族支系所属(Pomponius — gens 之名)。
• cognomen(家支别名)，标明此人所属的家族分支(Rufus 即此人的 cognomen)。

@key="pir-p-0749" links to Prosopographia Imperii Romani P 0749 — the canonical prosopographical reference for senators of the imperial period.

praenomen 为个人之名 (Quintus, 常缩写为 Q.); gentilicium 为氏族名 (Pomponius); cognomen 为家族分支名 (Rufus)。@key="pir-p-0749" 连接到《罗马帝国人物志》P 0749 条目，帝国时期元老的规范参考。

This man, in prose: Quintus Pomponius Rufus — Reynolds' translation reads "Quintus Pomponius Rufus, consul, priest, member of the priestly college for cult of the Flavians, curator of public works, imperial legate with propraetorian powers in the provinces of Moesia, Dalmatia and Hispania, commander of Legion Five, prefect of the coasts of nearer Spain, and Gallia Narbonensis in the war which Emperor Galba fought for the [Republic], proconsul of the province of Africa, through the agency of Lucius Asinius Rufus, propraetorian legate [---]." Cross-referenced in the standard prosopography as PIR P 0749. The XML @ref on the outer <persName> points there, and a SPARQL query can pull every inscription mentioning this senator across the EpiDoc-encoded Mediterranean.这个人, 用散文来说: Quintus Pomponius Rufus，Reynolds 译为 “Quintus Pomponius Rufus, 执政官, 司祭, 弗拉维家族祭祀团成员, 公共工程总监, 莫西亚、达尔马提亚、西班牙诸行省具副执政官权之元首特使, 第五军团团长, 加尔巴元首为[共和国]而战时之近西班牙与高卢纳尔波内西斯沿海防务长官, 非洲行省代行总督, 经其副执政官特使 Lucius Asinius Rufus 之手 [---]。” 在标准人物志中, 他的条目编号为 PIR P 0749。外层 <persName> 的 @ref 即指此, 一条 SPARQL 查询便可把整个 EpiDoc 编码的地中海世界中所有提到他的铭文一网打尽。

cos for consulcos 缩写 consul

The Latin word consul was conventionally abbreviated as co·s (with an interpunct between, but readable as cos). The encoder treats this as a discontinuous abbreviation, just like Ian(uaria)s in Exercise 07:

拉丁文 consul 一词惯常缩写为 co·s(中间夹一间隔点, 但实际读作 cos)。编码视其为不连续缩写, 与练习 07 中的 Ian(uaria)s 同理:

<expan><abbr>co</abbr><ex>n</ex><abbr>s</abbr><ex>ul</ex></expan>

Reader sees co(n)s(ul): co on stone, n supplied, s on stone, ul supplied. The same pattern appears again for procos = proconsul.

拉丁词 consul 习惯缩写为 co·s。编码作间断式: co 石上, n 编者补, s 石上, ul 编者补。procos = proconsul 同此处理。

The Fifth Legion as <orgName>第五军团作为机构

<orgName type="military" ref="#legio5">
  <w lemma="legio">
    <expan><abbr>leg</abbr><ex>ionis</ex></expan>
  </w>
  <num value="5"> V</num>
</orgName>

<orgName type="military"> wraps the legion as an organisation. Inside, the word legionis is expanded, and the numeral V is given its @value="5". The @ref="#legio5" ties this attestation to a corpus-wide entity, so a query for «every inscription naming the Fifth Legion» retrieves all of them at once.

<orgName type="military"> 把军团作为机构包装。内部包含 legionis 的展开与数字 V。@ref="#legio5" 把此处归并到语料库范围的同一实体, 便于“所有提及第五军团的铭文”一次性检索。

Why "leg(atus) leg(ionis) V" matters: a senator commanded a single specific legion before he became proconsul. In this inscription Pomponius Rufus served as legatus (commander) of Legio V. Six legions in the Roman army carried the numeral V over time (Alaudae, Macedonica, Gallica, Urbana, Apollinaris, Iovia), and which one is meant matters historically — but the inscription is silent on the surname. Encoding "Legion Five" as <orgName ref="trismegistos:orgs/..."/> lets the corpus stay honest about the ambiguity while still attaching the unit to a queryable record.“leg(atus) leg(ionis) V”何以紧要: 元老成为行省总督前, 必先指挥过某一支具体的军团。本铭中, Pomponius Rufus 为第五军团之 legatus(团长)。罗马军中曾先后有六支军团带编号“V”(Alaudae、Macedonica、Gallica、Urbana、Apollinaris、Iovia), 究系何者, 史家意见分歧, 而铭文本身不具姓号。把“第五军团”编码为 <orgName ref="trismegistos:orgs/..."/>, 既诚实保留了这种含混, 又让这支军队能挂上一条可检索的记录。

A name a senator usually erased通常被元老抹去的名字

<persName type="emperor" ref="#galba">
  <name type="cognomen" nymRef="Galba">G<supplied reason="lost">a</supplied>lba</name>
</persName>
<w lemma="pro">pro</w>
<w lemma="republica"><supplied reason="lost">re<expan><abbr>p</abbr><ex>ublica</ex></expan></supplied></w>
<w lemma="gero">gessit</w>

«…in the war which Emperor Galba waged for the [Republic].» Galba was the loser of the 68–69 CE civil war; most senators who had served him later erased the reference. Pomponius Rufus kept it — fifty years later, under Trajan. This is one of the only epigraphic attestations of senatorial service under Galba.

“……加尔巴元首为[共和国]而战之役。”加尔巴是公元 68—69 年内战的败者;凡曾在其麾下任职的元老, 多在后来悄悄抹去这段经历。Pomponius Rufus 偏偏没抹，五十年后, 在图拉真朝, 仍把这段履历刻在石上。这便成了关于加尔巴朝任职的极少数铭文见证之一。

Note the nested supplements + expansion on [re(p(ublica))]: the editor supplies re-, then inside the supplement an <expan> records that the rest is an abbreviation. Three-deep nesting; perfectly legal EpiDoc.

注意 [re(p(ublica))] 处嵌套的补字加展开: 编者补出 re-, 补字内部还含一个 <expan> 标记其余为缩写。三层嵌套, 完全合法的 EpiDoc。

The corpus as prosopographical database语料库作为人物志数据库

When every office is encoded as <w lemma="…"> + <expan> + optional <placeName>/<orgName>, the corpus answers:

一旦把每一职务都编码为 <w lemma="…"> + <expan>, 必要时再加 <placeName> 或 <orgName>, 整个语料库便能直接回答如下问题:

• Which senators held the consulship before becoming proconsul of Africa?
• Which provinces were governed by men who had previously commanded the Fifth Legion?
• Was the post of curator operum publicorum held earlier or later in a typical senatorial career?

• 有哪些元老在出任非洲行省总督之前, 已先任执政官?
• 有哪些行省的总督, 曾先指挥过第五军团?
• “公共工程总监”(curator operum publicorum)一职, 在元老的标准履历中通常位于早段还是晚段?

These are real historical questions. Without structural encoding, each requires hand-counting through thousands of inscriptions. With it, a SPARQL query returns the answer in seconds.

每一官职若编为 <w lemma="…"> + <expan> + 选附 <placeName>/<orgName>, 语料库即能回答: 阿非利加总督之前先任执政官者有谁? 由前第五军团长官治理之行省有哪些? 公共工程总监一职通常在仕途较早还是较晚阶段?，这些是真实的历史问题。无结构化编码, 每问皆需手工查阅; 有结构化编码, 一条 SPARQL 即得。

The second man in the inscription: L. Asinius Rufus, propraetorian legate. Q. Pomponius Rufus governed Africa per — "through the agency of" — Lucius Asinius Rufus, his deputy. Asinius is likely PIR A 1248 (or 1250). So one inscription names two senators, in a chain of delegation. Encoded properly, this single block of stone produces: 2 persons, 2 PIR cross-refs, ~9 roles, ~5 provinces, 1 legion, 1 historical war (Galba's), 1 proconsulate-of-Africa, 1 building-event for an arch in Lepcis Magna — about fifteen prosopographical facts. That is the corpus-as-database moment.铭文中第二个出现的人:L. Asinius Rufus, 副执政官特使。 Q. Pomponius Rufus 治理非洲, 是 per，“经其手”，Lucius Asinius Rufus 之代行。Asinius 一人, 大概对应 PIR A 1248(或 1250)。也就是说, 此铭名两位元老, 一条委派之链。若编码到位, 一块石头便可产出:二人、二条 PIR 互引、约九种官职、约五处行省、一支军团、一场历史性战争(加尔巴之战)、一任非洲行省代行总督、一座大莱普提斯之拱门，约十五条人物志事实。这就是"语料库即数据库"的瞬间。

Four lines for Pytheas四行献予 Pytheas

«Not even after death have you lost your fine reputation in the whole earth, but still all the splendid [achievements] of your soul remain — both those which you inherited, and those which you learnt, according to your nature, most excellent in intellect. So now, Pytheas, you have also gone to the Island of the Blest.»

“纵使身死, 你卓著之名遍传寰宇, 未曾湮灭; 你灵魂的一切光辉犹存，那些你天生秉受的, 那些你顺其本性所学得的; 你心智之卓越无人能及。Pytheas 啊, 于是你也步入了极乐之岛。”

Two couplets stacked: H/P/H/P两组对句叠合: 六/五/六/五

<lg met="elegaic">
  <l n="1" met="hexameter">…</l>
  <l n="2" met="pentameter">…</l>
  <l n="3" met="hexameter">…</l>
  <l n="4" met="pentameter">…</l>
</lg>

One outer <lg> wraps the whole quatrain. No nested <lg> for each couplet — the alternation of @met values implicitly marks the couplet structure.

一个外层 <lg> 包整个四行。不必为每组对句单独嵌套，@met 值的交替已隐含对句结构。

@reg on both levels两层皆带 @reg

<persName reg="Πύθεας" key="Pytheas" type="aphrodisian" full="yes">
  <name reg="Πυθέας">Πυθέα</name>
</persName>

The inscribed form Πυθέα is the vocative («Pytheas!»). Two regularisations:

石上所刻 Πυθέα 是呼格形式(“Pytheas 啊!”)。两层正字化:

• <persName reg="…"> — the canonical «person tag» as the corpus database represents this individual. Sometimes uses an idiosyncratic transliteration.
• <name reg="…"> — the canonical «word tag», the dictionary nominative form of the Greek noun.

They can differ in spelling conventions; both serve canonicalisation but at different levels of abstraction.

二者在拼写约定上可能略有差异; 都是规范化, 但抽象层级不同。

两个 @reg 含义不同: <persName reg="…"> 是语料库中规范化的“人物标签”; <name reg="…"> 是规范化的“词汇标签”(即希腊文名词的主格)。两者拼写可能略有差异; 都做规范化, 但抽象层级不同。

What NOT to encode不必编码的部分

• No <apparatus> — the text comes through manuscript transmission (the Greek Anthology), not from a surviving stone. No carving variants to record.
• No <gap> or <supplied> — the Byzantine scribe transmitted the text complete. Nothing to restore.
• Minimal <facsimile> — no photograph; no diplomatic transcription.

• 无 <apparatus>：此文是通过抄本(《希腊文选》)传下的, 并非出自现存的石头, 故无刻文异读可记。
• 无 <gap> 与 <supplied>：拜占庭抄写者把全篇完整传下, 无所遗失, 故无须补字。
• 极简 <facsimile>：没有照片, 也没有逐字转录。

EpiDoc encodes what is editorially relevant, not what could in principle be encoded. If you find yourself adding tags that don't say something true about the text or its transmission, delete them.

EpiDoc 只编码对编辑而言相关的, 不编码所有可编码的。若加的标签未陈述文本或流传的真相, 请删去之。

The cultural meaning of the choices编码选择背后的文化意义

• κλέος ἐσθλόν (line 1) — straight Homeric vocabulary. The «glory that survives death.» Pytheas placed alongside Achilles.
• μακάρων νῆσον (line 4) — the Homeric/Hesiodic «Isle of the Blest» reserved for heroes (Hesiod, Works and Days 168–173). A Christian writer would normally use Christian eschatology; choosing the Homeric afterlife is a cultural statement.
• μῆτιν πανάριστε (line 3) — μῆτις is Odysseus's defining virtue; πανάριστος is a Homeric superlative. Pytheas placed in the line of Odysseus.

• κλέος ἐσθλόν(第 1 行)，道地的荷马用语, 即“死后不朽的荣名”。把 Pytheas 置于阿喀琉斯之列。
• μακάρων νῆσον(第 4 行)，荷马与赫西俄德笔下专为英雄而设的“至福者之岛”(参《工作与时日》168—173)。基督徒作者通常用基督教终末论, 此处偏取荷马式来世, 是一种有意为之的文化姿态。
• μῆτιν πανάριστε(第 3 行)，μῆτις(“机谋”)是奥德修斯的标志性德性; πανάριστος 又是荷马式的最高级。Pytheas 由此置于奥德修斯之列。

EpiDoc encodes the form via @met + @lemma; the cultural reading is the encoder's <commentary>. Both layers are part of the edition.

EpiDoc 通过 @met + @lemma 记录形式; 文化解读则在编者的 <commentary> 中呈现。两层都是版本的一部分。

What EpiDoc actually doesEpiDoc 到底在做什么

Read → Encode → Review读 → 编 → 互评

What you can do next毕业之后