Reference Data Model Texts

Texts — Layer 2

Layer 2 is the textual content: transliterations broken into lines and tokens, and translations.

Key tables

text_lines — One row per line of ATF transliteration. Fields include:

• p_number — the artifact
• surface_type — obverse, reverse, edge, etc.
• line_no — ATF line label (e.g., "1.", "r.1.", "e.1.")
• content — raw ATF text for the line
• line_order — integer ordering for display

tokens — Individual words (and determinatives) within a line. One row per word-position. Fields include:

• line_id — foreign key to text_lines
• word_no — position within the line
• form — the written form as it appears in ATF (e.g., "lu₂-gal")
• is_determinative — true for silent classifier signs ({d}, {f}, {ki})
• flags — damage markers from ATF (!, ?, #, *)

token_readings — The reading assigned to a token — the sign sequence used to produce the written form. Separate from lemmatization.

translations — Line-by-line translations grouped by language code. Fields:

• p_number
• line_no
• language — ISO 639 code (en, de, it, fr, es, dk, ca, fa, ts)
• text — the translation text for that line

ATF line structure

ATF (ASCII Transliteration Format) is the standard digital notation for cuneiform. Key conventions:

&P227657 = KTT 188
#atf: lang sux
@obverse
1. ninda
2. kasz
#tr.en: bread
#tr.en: beer
@reverse
(blank)

• &P227657 — tablet ID header
• #atf: lang sux — language declaration (sux = Sumerian)
• @obverse / @reverse — surface markers
• 1. — line number
• #tr.en: — inline English translation
• [...] — broken or missing text
• {d} — determinative (divine name follows)

CDL tree decomposition

ORACC stores annotated texts as CDL (Chunk-Delimiter-Lemma) trees. Each text decomposes into nested nodes: chunk > sentence > lemma > grapheme. The ingestion pipeline walks this tree and maps it into tokens and lemmatizations (Layer 3), keyed by (p_number, line_no, word_no).

Translation language codes

Translations in Glintstone use these language codes:

Code	Language
en	English
de	German
it	Italian
fr	French
es	Spanish
dk	Danish
ca	Catalan
fa	Farsi
ts	Transliteration supplement

43,777 artifacts have at least one translation (~12% of the catalog). Coverage is not uniform — literary and well-studied texts have translations; administrative tablets typically do not.