README.md

metadata

license: cc-by-sa-4.0
language:
  - hbo
  - arc
  - grc
pretty_name: Clause columns for the Hebrew and Greek clause layouts
tags:
  - bible
  - hebrew
  - greek
  - clauses
  - rrg
  - discourse
configs:
  - config_name: hebrew
    data_files: data/hebrew.parquet
  - config_name: greek-nt
    data_files: data/greek-nt.parquet

Clause columns (v1.0)

Grammar and discourse facts for every clause of the two bcv-commons clause layouts, keyed by their clause keys: the main verb and its form, voice, stem and person; negation and its scope; force; quotation and speaker; the narrative line; actor and undergoer with the verb's argument frame; subject and participant tracking; fronting and focus; and, for the Hebrew, the Septuagint's form of the verb.

config	layout (pinned)	clauses
hebrew	bcv-commons/hebrew-clause-layout v0.37.1 (32f89b71, CC BY-SA 4.0)	87,582
greek-nt	bcv-commons/greek-nt-clause-layout v0.7 (9168fac3, CC BY 4.0)	29,682

Join a row to its layout by clause_key (Hebrew clause_key; Greek first). Every clause of the layout has exactly one row. The layouts' own columns (rule, relation, parent, depth and so on) are not copied here: read them from the layout.

What changed from the previous version

The previous table (revision 8a132e44) was built by make-tanakh-research. This version is rebuilt by bcv-query from the same free sources, under the same column names and value names, and research has agreed to retire its builder. Differences, all measured in checks.md (the contract check against the previous table, which joins on every clause both ways):

• Split clauses. A Hebrew clause the layout splits into several segments (2,263 keys) is now read whole. The previous table read its verb from one segment and its word order from another. Every other Hebrew difference comes from this.
• Undergoer. Both testaments now use the verb's frame (A1 inside the clause is overt, A1 elsewhere is implicit), with three new values: clause (the verb's patient is a clause), complement (a word with an object role but no A1 in the frame: complements, double and cognate objects, indefinite objects; in Hebrew also objects the frame misses) and, in undergoer_basis, what made each undergoer present. Hebrew: an object with no frame A1 but with the object marker אֵת counts as overt (role_o_et).
• Greek actor reads the frame too, as the Hebrew does.
• Force is the layouts' own force codes. In the Greek a question covers the whole question within its quotation, so about 5% of Greek clauses differ from the previous table.
• Septuagint parse: the parse of the form in its own GLAUx verse comes first (lxx_basis).
• New columns: speaker_id, lxx_basis, undergoer_basis; Greek args, subj, subj_switch.
• Not carried any more: akt, akt_p, juss_p (model answers, not source facts), juncture and nexus (to come with the clause relation work), heb_form (built from a private alignment), legal_future, and the copies of layout columns (last, rule, relation, discourse, via, role, junction, parent, depth).

Columns

column	in	description
clause_key	both	the layout's clause key (Hebrew clause_key, one row per key: a clause the layout splits into several segments is read whole; Greek: the segment's first word id, the layout's first)
ref	both	BOOK c:v.seg of the clause's first segment
verb	both	the main verb's word id (first finite verb; else participle / infinitive); null in a verbless clause
conj	both	Hebrew: qatal
voice	both	active
neg	both	the negator's class: lo
kind	both	the layout's clause kind (VC verbal, NC nominal)
force	both	prohibitive
interrog	both	true when the clause carries the layout's question code (fc.q)
person, number, gender	both	the main verb's (1 2 3; s p d; m f c n); null where the form has none
stem	both	Hebrew / Aramaic stem of the verb (qal, niphal ... peal, aphel ...); Greek: null
causative	both	true for hiphil, hophal and the causative stems of both languages; Greek: false
strong	both	the verb's Strong's number
speech	both	Hebrew: the quotation level from bcv-commons/hebrew-speaker-layer (0 narration); Greek: 1 inside a quotation
speaker	both	Hebrew: the layer's speaker (the Hebrew lemma[s] of the words naming the speaker); Greek: the layout's speaker_id
speaker_id	hebrew	the layer's character id of the speaker (the ids the Greek layout also uses)
frame	hebrew	the speaker of a text that is speech from its first line (a psalm's psalmist), from the layer
genre	both	the book's genre (one code per book)
line	both	main
actor	both	overt
undergoer	both	overt
undergoer_basis	both	what made the undergoer present: np_a1
args	both	the verb frame's argument labels in order (A0+A1+A2 ...)
subj	both	the subject's referent word id(s): MACULA's subjref of the verb, else the A0's referent, else the A0 word
subj_switch	both	same
lxx_form	hebrew	the Septuagint form MACULA Hebrew gives for the clause's verb
lxx_tense, lxx_mood	hebrew	that form's tense and mood: its GLAUx parse in the same verse when found there, else the commonest GLAUx parse of the form
lxx_basis	hebrew	verse
neg_scope	both	verb (the negator right before a verb)
subj_form	both	noun
part	both	the subject's participant: what its reference chain resolves to (Hebrew: Strong's number; Greek: lemma)
part_switch	both	same
front	both	the role of the first constituent before the verb; '' when the verb comes first; verbless
focus	both	sentence (presentational)
m_trans	both	number of macroroles in the verb's frame: 2 (A0 and A1), 1, 0
subj_role	both	actor
recip	both	1 when the frame has an A2

Undergoer

value	meaning	Hebrew	Greek NT
overt	a patient inside the clause (the frame's A1, or an object with אֵת)	23,993	8,124
clause	every A1 of the verb lies under an object clause	1,396	3,425
complement	an object-role word with no A1 in the frame	8,650	3,168
implicit	the frame's A1 is named outside the clause	3,211	1,479
none	no patient	50,332	13,486

MACULA calls its Hebrew frames tentative and they list fewer patients than the Greek ones, so Hebrew complement includes real objects the frame misses. A second human analysis (used for figures only) finds a patient for 0.75 of Greek overt clauses, 0.61 of clause, 0.34 of complement and 0.29 of none.

How it was made, and how far to trust it

Built by code (shoresh/macula/build_clause_columns.py in bcv-query, contract check clause_columns_contract.py) from: MACULA Hebrew (WLC) and MACULA Greek (Nestle 1904) with their trees, frames and references (CC BY 4.0); the two layouts; bcv-commons/hebrew-speaker-layer (CC BY 4.0); and GLAUx's Septuagint parses (CC BY-SA 4.0; only lxx_tense and lxx_mood). Nothing comes from BHSA or TBTA, and no English text is used as evidence. Every value is a rule's output; no column was checked by hand row by row. A human review of samples is planned for later and will be reported here.

Licence

CC BY-SA 4.0: the Hebrew layout's clause keys are CC BY-SA 4.0, and so are GLAUx's parses. Credit: MACULA (Clear Bible / Biblica), GLAUx (KU Leuven, Alek Keersmaekers), bcv-commons (the layouts, the speaker layer), make-tanakh-research (the column design and the previous version), and this dataset.