Reference Data Model

Data Model Overview

Glintstone's schema is organized into five layers. The P-number is the universal join key. Everything connects through it.

The five layers

Layer	Key tables	Current scale
1. Artifacts	artifacts, surfaces, artifact_identifiers, artifact_composites, composites	353,283 artifacts
2. Texts	text_lines, tokens, token_readings, translations	1.4M lines, 3.96M tokens
3. Linguistic Analysis	lemmatizations, morphology, annotation_runs	~309k lemma tokens
4. Lexical Resources	signs, sign_values, sign_variants, lexical_lemmas, lexical_senses, glossary_entries, glossary_forms, glossary_senses	3,367 signs, 61k lemmas, 155k senses
5. Scholarly Context	scholars, publications, publication_authors, artifact_editions, scholarly_annotations, fragment_joins	20,490 scholars, 16,725 publications

The P-number

P227657 is an example P-number. CDLI assigned it. Every system in the field uses P-numbers as the universal artifact identifier. In Glintstone, it is the primary key on the artifacts table and the foreign key in nearly every other table.

Think of the P-number as the DOI for tablets.

Q-numbers for composites

Some tablets are individual copies of a larger literary work — the Epic of Gilgamesh has 73 known exemplar tablets. Q-numbers (e.g., Q000039) identify the abstract composite text. Individual tablets link to composites via artifact_composites. The composites table holds Q-number metadata.

The annotation_run_id

Every row in every annotation table (lemmatizations, morphology, scholarly_annotations, translations) carries annotation_run_id. This is the attribution backbone. Each annotation run identifies the source, method, scholar, and timestamp. Nothing is silently overwritten — new data creates new runs.

Layer pages

• Artifacts — Layer 1: the physical object
• Texts — Layer 2: transliteration and tokens
• Linguistic Analysis — Layer 3: lemmatization and morphology
• Lexical Resources — Layer 4: signs and dictionaries
• Scholarly Context — Layer 5: scholars, publications, editions