Reference Data Model Artifacts

Artifacts — Layer 1

Layer 1 is the physical object: what it is, where it came from, what catalog records describe it.

Key tables

artifacts — One row per cuneiform object. Key fields:

Field	Notes
p_number	Primary key. CDLI universal identifier.
designation	Human-readable label (e.g., "CDLI Seals 001234")
museum_number	Museum accession number
excavation_number	Field excavation ID
period	Normalized period string (e.g., "Ur III (ca. 2112-2004 BC)")
provenience	Find site (e.g., "Nippur (mod. Nuffar)")
genre	Text genre (Administrative, Literary, Legal, etc.)
language	Primary language
object_type	Tablet, prism, cone, seal, etc.
material	Clay, stone, metal, etc.
pipeline_stage	One of: captured, recognized, transcribed, lemmatized, translated
collection	Holding institution

surfaces — Named surfaces of the physical object: obverse, reverse, left/right/top/bottom edges, seal. One tablet can have multiple surfaces.

artifact_identifiers — Alternative identifiers for the same artifact: museum number, excavation number, primary publication. Multiple identifier types per artifact.

composites — Q-number composite texts (abstract literary works).

artifact_composites — Junction table linking P-numbers to Q-numbers.

Where data comes from

Almost all Layer 1 data comes from the CDLI catalog (cdli_cat.csv), which provides 353,283 records with 64 metadata fields per artifact. CDLI is CC0 — public domain.

The CDLI bulk export was last updated in August 2022. Approximately 33,000 additional stub records exist for P-numbers present in the ATF file but absent from the catalog CSV — these have minimal metadata.

Null rates

From a live database audit of the CDLI catalog:

Field	Null rate
museum_number	9.2%
excavation_number	68%
width / height	82%
thickness	91%
period	1.8%
provenience	3.1%
genre	1.4%
language	5.2%

Expect NULLs. Physical measurement data is especially sparse.

Canon tables

Raw source data uses inconsistent strings for period, provenience, language, and genre. Canon tables normalize these to consistent values used throughout the app:

• period_canon — maps period strings to canonical forms and groups (e.g., "Third Millennium")
• provenience_canon — maps raw provenience strings to short canonical names
• language_map — maps ISO-style language codes to display labels
• genre_canon — maps raw genre strings to canonical forms

The filter system operates on canonical values.

The pipeline_stage column

pipeline_stage is a string enum on artifacts tracking the highest stage the artifact has reached: captured, recognized, transcribed, lemmatized, translated. It is set during ingestion and updated as new data arrives.