| { |
| "$schema": "https://json-schema.org/draft/2020-12/schema", |
| "$id": "https://huggingface.co/datasets/bcv-commons/hebrew-clause-layout/schema/clause_segment.schema.json", |
| "title": "Clause segment", |
| "description": "One row of clauses/<BOOK>/<BOOK>_<CCC>.jsonl and data/clauses-<BOOK>.parquet: a maximal run of whole words, inside one verse, that belongs to one MACULA clause. Fields marked (new) are not in the shoresh /scaffold API today.", |
| "type": "object", |
| "additionalProperties": false, |
| "required": [ |
| "book", |
| "chapter", |
| "verse", |
| "seg", |
| "clause", |
| "parent", |
| "clause_key", |
| "parent_key", |
| "sentence", |
| "depth", |
| "kind", |
| "language", |
| "first", |
| "last", |
| "words", |
| "text", |
| "ketiv_qere" |
| ], |
| "properties": { |
| "book": { |
| "description": "USFM book code.", |
| "type": "string", |
| "pattern": "^[1-3A-Z][A-Z0-9]{2}$", |
| "examples": [ |
| "GEN", |
| "RUT", |
| "1SA" |
| ] |
| }, |
| "chapter": { |
| "type": "integer", |
| "minimum": 1 |
| }, |
| "verse": { |
| "description": "Verse in the Hebrew numbering of the source text (Psalm superscriptions are verse 1).", |
| "type": "integer", |
| "minimum": 1 |
| }, |
| "seg": { |
| "description": "Position of the segment within the chapter, from 1, in text order.", |
| "type": "integer", |
| "minimum": 1 |
| }, |
| "verse_ref_en": { |
| "description": "(new) The same verse in English (BSB/KJV) numbering, as 'BOOK C:V', from the bcv-commons/bibles verse map org-to-eng (STEPBible TVTMS). Null if the verse has no English counterpart. A Psalm superscription is 'BOOK C:title'.", |
| "type": [ |
| "string", |
| "null" |
| ], |
| "pattern": "^[1-3A-Z][A-Z0-9]{2} [0-9]+:([0-9]+|title)$", |
| "examples": [ |
| "GEN 31:55", |
| "PSA 3:title" |
| ] |
| }, |
| "clause": { |
| "description": "Build row id of the clause (all segments of one clause share it). Not stable across rebuilds: key on clause_key.", |
| "type": "integer" |
| }, |
| "parent": { |
| "description": "Build row id of the clause this clause is embedded in, or null. Not stable across rebuilds: key on parent_key.", |
| "type": [ |
| "integer", |
| "null" |
| ] |
| }, |
| "clause_key": { |
| "description": "Stable id of the clause: the first word key of its first segment.", |
| "type": "string", |
| "pattern": "^[0-9]{12}$" |
| }, |
| "parent_key": { |
| "description": "clause_key of the nearest enclosing clause that has words of its own, or null. Coordination groups without words of their own (MACULA ClCl…) have no clause_key and are skipped.", |
| "type": [ |
| "string", |
| "null" |
| ], |
| "pattern": "^[0-9]{12}$" |
| }, |
| "sentence": { |
| "description": "Build row id of the MACULA sentence (= the verse). Not stable across rebuilds.", |
| "type": "integer" |
| }, |
| "depth": { |
| "description": "Number of clause ancestors via 'parent' (0 = not embedded). Shown as the number after the colon in layout/ files.", |
| "type": "integer", |
| "minimum": 0 |
| }, |
| "discourse_parent": { |
| "description": "clause_key of the clause this clause depends on in the text-level hierarchy, also across verses. Every clause has one, chain links included (a wayyiqtol on the clause before it); null only for the first clause of a line (a book, a chapter, a new Psalm, and where build.json says the speech stack resets). Not in MACULA; derived in the build (see README).", |
| "type": [ |
| "string", |
| "null" |
| ], |
| "pattern": "^[0-9]{12}$" |
| }, |
| "text_depth": { |
| "description": "Flat depth: the number of links up the discourse_parent chain whose relation is not in build.json relations.flat (0 = main line). The number shown in layout/ files; null until the derivation exists.", |
| "type": [ |
| "integer", |
| "null" |
| ], |
| "minimum": 0 |
| }, |
| "text_depth_law": { |
| "description": "(new, optional) Flat depth with conditionals laid out as in BHSA: where a protasis comes before its apodosis, the protasis takes the apodosis's place in the tree, and the apodosis stands level beside it (its link counts as flat). The protasis is indented one level from the clause the apodosis hung on. A protasis whose apodosis continues the apodosis of an earlier protasis stands level with that protasis (laws in a row). Otherwise as text_depth. Null or absent when not derived. Defined in clause_layout.toml [depth.law].", |
| "type": [ |
| "integer", |
| "null" |
| ], |
| "minimum": 0 |
| }, |
| "tree_depth": { |
| "description": "(new) Tree depth: the number of links up the discourse_parent chain, every link counted (0 = first clause of a line). The BHSA-like view; text_depth is the flat view.", |
| "type": [ |
| "integer", |
| "null" |
| ], |
| "minimum": 0 |
| }, |
| "kind": { |
| "description": "MACULA clause kind, e.g. VC (verbal clause), NC (nominal clause).", |
| "type": "string" |
| }, |
| "rule": { |
| "description": "MACULA constituent rule of the clause, e.g. 'V-S-PP'.", |
| "type": [ |
| "string", |
| "null" |
| ] |
| }, |
| "language": { |
| "description": "hbo = Hebrew, arc = Aramaic.", |
| "enum": [ |
| "hbo", |
| "arc" |
| ] |
| }, |
| "first": { |
| "description": "MACULA key of the first word token of the segment (BBCCCVVVWWW plus part digit, as in the scaffold API).", |
| "type": "string", |
| "pattern": "^[0-9]{12}$" |
| }, |
| "last": { |
| "description": "MACULA key of the last token of the segment.", |
| "type": "string", |
| "pattern": "^[0-9]{12}$" |
| }, |
| "words": { |
| "description": "MACULA keys of all tokens of the segment, in text order (prefixes included).", |
| "type": "array", |
| "items": { |
| "type": "string", |
| "pattern": "^[0-9]{12}$" |
| }, |
| "minItems": 1 |
| }, |
| "text": { |
| "description": "Surface text of the segment (UHB, with vowels and accents; qere reading where there is a ketiv/qere). Words separated by single spaces; maqaf kept; paseq written ' ׀ '; no trailing space.", |
| "type": "string", |
| "minLength": 1 |
| }, |
| "ketiv_qere": { |
| "description": "(new) Words of this segment that have a ketiv/qere. Empty list if none.", |
| "type": "array", |
| "items": { |
| "type": "object", |
| "additionalProperties": false, |
| "required": [ |
| "key", |
| "ketiv", |
| "qere" |
| ], |
| "properties": { |
| "key": { |
| "type": "string", |
| "pattern": "^[0-9]{12}$" |
| }, |
| "ketiv": { |
| "description": "Written form (consonants, as in the manuscript). Null for a qere without ketiv (e.g. Ruth 3:5).", |
| "type": [ |
| "string", |
| "null" |
| ] |
| }, |
| "qere": { |
| "description": "Read form, vocalised. Null for a ketiv without qere (e.g. Ruth 3:12).", |
| "type": [ |
| "string", |
| "null" |
| ] |
| } |
| }, |
| "description": "One entry per ketiv/qere. A ketiv over two qere words, or the reverse, gives one entry per word key (e.g. Gen 30:11)." |
| } |
| }, |
| "discourse_attach": { |
| "description": "(new, optional) Where the dependency is on a word or phrase inside the discourse_parent clause (e.g. a relative clause on its antecedent), the word key of that word, or of the phrase's head. Null when the clause depends on the clause as a whole. Does not affect text_depth.", |
| "type": [ |
| "string", |
| "null" |
| ], |
| "pattern": "^[0-9]{12}$" |
| }, |
| "relation": { |
| "description": "(new) The type of the link to discourse_parent; null when discourse_parent is null. The set is listed in build.json (relations.all), e.g. continuation, coordination, asyndetic, speech, inquit, causal, purpose, temporal, conditional, relative, circumstantial, complement, vocative, embedded. Not an enum, so the set can grow.", |
| "type": [ |
| "string", |
| "null" |
| ], |
| "examples": [ |
| "continuation", |
| "coordination", |
| "speech", |
| "conditional", |
| "relative" |
| ] |
| }, |
| "discourse_rule": { |
| "description": "(new, optional) The rule that set discourse_parent, e.g. embedded, main, sub_back, sub_fwd, circ. Names and set follow the rule version in build.json (discourse_rules).", |
| "type": [ |
| "string", |
| "null" |
| ], |
| "examples": [ |
| "embedded", |
| "main", |
| "sub_back", |
| "sub_fwd", |
| "circ" |
| ] |
| } |
| } |
| } |
|
|
