Glossary

Technical terms of manuscript studies, textual criticism and project architecture

This glossary gathers the specialized terms used throughout Manuscriptum Atlas. Organized by category, it clarifies concepts from New Testament textual criticism, palaeographic dating, historical sources and the LLM architecture that underpins the project.

Manuscripts

Papyrus: Manuscript written on papyrus (an Egyptian plant). These are the earliest witnesses to the NT (2nd–4th centuries). Designated by the prefix P followed by a number (P52, P66, P75). Often fragmentary, preserved by Egypt's dry climate.
Parchment: Treated animal skin (goat, sheep, calf) used as a writing surface. More durable than papyrus, it progressively replaced it from the 4th century onward. The material of the great uncial codices.
Uncial (Majuscule): Manuscript written in separate capital letters (3rd–9th centuries). Designated by number 01–0323 or by letters (Sinaiticus = א/01, Vaticanus = B/03, Alexandrinus = A/02).
Minuscule (Cursive): Manuscript in cursive lowercase letters, with connected words (9th–15th centuries). Designated by number only (1–2900+). The vast majority of all surviving manuscripts.
Lectionary: Liturgical manuscript containing biblical pericopes (passages) arranged for reading according to the ecclesiastical calendar. Designated by the prefix l (l1–l2500+).
Fragment: Incomplete manuscript, preserving only a few verses or leaves. Most early papyri are fragmentary; P52 contains only John 18:31–33 and 37–38.
Gregory-Aland (GA): Standard cataloguing system for Greek NT manuscripts created by Caspar René Gregory and updated by Kurt Aland. Assigns a unique identifier to each manuscript by category (papyri, uncials, minuscules, lectionaries).
NTVMR: New Testament Virtual Manuscript Room — the digital platform of the INTF (Institut für Neutestamentliche Textforschung, Münster) gathering TEI/XML transcriptions, images and manuscript metadata. Primary data source of Manuscriptum Atlas.
INTF: Institut für Neutestamentliche Textforschung at the University of Münster. The world's leading academic center for NT textual criticism; it maintains the NTVMR and the Gregory-Aland catalogue.

Textual coverage

Binary coverage: Coverage model used in the project: a verse is considered covered if it appears in at least one surviving manuscript. It measures neither textual variants nor the completeness of the verse.
Cumulative coverage: Coverage of century N that includes all manuscripts from centuries 1 through N. It answers the question: 'by the 4th century, how much of the NT was already attested?'
Fragmentation Index: Academic metric measuring how fragmentary the attestation of a book is. Values close to 1 indicate that few manuscripts cover each verse (high fragmentation); values close to 0 indicate broad attestation.
Stabilization century: The century in which cumulative coverage reaches ≥ 90% of the verses of a book or of the entire NT. It marks the point at which the text is sufficiently attested for reconstruction.
Century Growth Rate: Rate of coverage growth between two consecutive centuries. A high growth rate indicates strong expansion (e.g., the 3rd century for the NT, with the arrival of the papyri).

Dating

Palaeography: The science of analyzing ancient writing styles. The basis for dating pre-modern manuscripts, with a typical margin of error of 50–100 years.
Terminus ante quem: Latin: 'limit before which'. The latest possible date for the production of a manuscript (it cannot be later). Combined with the terminus post quem, it defines the plausible range.
Terminus post quem: Latin: 'limit after which'. The earliest possible date for the production of a manuscript (it cannot be earlier).
yearMin / yearMax / yearBest: Year-based dating fields in the project. yearMin/yearMax = plausible range. yearBest = consensus year when explicitly attested by an academic source (never computed as an average).
datingConfidence: Confidence level of the dating: HIGH (NTVMR or primary academic source), MEDIUM (manually curated seed data), LOW (LLM-generated estimate awaiting review). Promotion only through human review.
Floruit: Latin: 'flourished'. The period in which a church father or ancient author was at the height of his intellectual activity. Used when exact birth/death dates are uncertain.

Textual criticism

Textual variant: A difference in reading between manuscripts in the same passage. Manuscriptum Atlas currently does not measure variants, only coverage (whether or not the verse exists).
Lectio difficilior: Principle of textual criticism: among variants, the more difficult reading tends to be the original — scribes tend to simplify, not complicate, the text.
Textual family: Grouping of manuscripts with similar readings, reflecting a common origin or transmission. The main NT families are: Alexandrian (P75, Sinaiticus, Vaticanus), Byzantine (most minuscules) and Western (Bezae, Vetus Latina).
Pericope: A self-contained biblical passage with thematic unity. The 'Pericope of the Adulterous Woman' (John 7:53–8:11) is famous for being absent from the early manuscripts.

Interlinear Bible

Lemma: The canonical (dictionary) form of a word. For Greek/Hebrew words inflected in the text, the lemma is the form used for lexicon lookup.
Transliteration: Representation of Greek/Hebrew words in the Latin alphabet for ease of reading (e.g., λόγος → 'logos').
Gloss: A brief, literal translation (a single word or short expression) of each original word in the interlinear line. Distinct from a fluent translation — it preserves the word order of the source text.
Strong's Concordance: Numbering system created by James Strong (1890) that assigns a unique number to each Hebrew (G1–G8674) or Greek (G1–G5624) word of the NT. The standard for lexicon lookup.
Lexicon: Dictionary of ancient words with lemma, transliteration, pronunciation, morphological classification and definition. The project integrates Greek (~5,600 entries) and Hebrew (~8,700 entries) lexicons.
Morphology: Grammatical analysis of a word: part of speech, case, number, gender, tense, mood, voice. Crucial for accurate translation of biblical Greek and Hebrew.

Historical sources

Philip Schaff: Historian (1819–1893). His 'History of the Christian Church' is a PRIMARY source (weight 1.0) for the dating and description of ecclesiastical councils.
Karl Joseph von Hefele: Catholic historian (1809–1893). 'History of the Councils of the Church' is an essential academic reference on councils. PRIMARY source (weight 1.0).
Catholic Encyclopedia: Classic Catholic encyclopedia (1907–1912). ACADEMIC source (weight 0.8). Used as cross-verification for councils.
Fordham Medieval Sourcebook: Digital collection from Fordham University gathering primary medieval texts. PRIMARY source (weight 1.0) for conciliar documents.
Wikidata: Wikimedia's structured knowledge base. STRUCTURED source (weight 0.7). Useful for linking councils to Q-IDs and enriching them with geographic coordinates.
Source Consensus Engine: Project engine that receives claims from multiple sources (year, location, participants) and computes a consensus weighted by each source's weight. Persistent conflicts go to LLM-based resolution.
Source weight: Reliability multiplier applied to each claim: PRIMARY (1.0), ACADEMIC (0.8), STRUCTURED (0.7), AGGREGATOR (0.5). Determines the source's influence on the final consensus.

LLM system

Tier (LOW / MEDIUM / HIGH): Complexity classification of the LLM task. LOW = short enrichment (Haiku); MEDIUM = translations/summarizations (Sonnet); HIGH = academic reasoning, alignment, dating (Opus).
LlmPromptQueue: PostgreSQL table (`llm_prompt_queue`) that stores pending prompts for asynchronous processing. Enables zero cost by using Claude Code (subscription) instead of the paid API.
Prepare / Apply: Two-phase pattern for LLM tasks. _prepare = enqueues prompts. /run-llm processes them. _apply = applies responses to the source database. Guarantees idempotency and replay.
callback_context: JSON field in the queue carrying the metadata needed to reapply the response (target entity IDs, translation parameters, etc.). Read by LlmResponseProcessor during _apply.
Topic llm-results-ready: Kafka topic that receives the name of the completed phase after each processed batch. A backend consumer (`LlmResultsConsumer`) triggers LlmResponseProcessor automatically.
Claude Code: Claude's CLI/IDE client (Anthropic). Acts as the queue's processing engine via the /run-llm skill — replaces paid API calls, eliminating ingestion costs.