Full syntactic tree annotation of the Greek New Testament from the MACULA Greek SBLGNT edition. Five relational tables: word-level tokens with morphological, semantic, and cross-language features, sentence boundaries, word groups (clauses/phrases) with syntactic rules and roles, word-group hierarchy (parent–child tree edges), and word-group membership (which tokens belong to which groups). The same data is also published as a unified Text-Fabric node graph under tf-projection/.

Source

Attribute	Value
Project	MACULA Greek (SBLGNT edition)
Authors/Institution	Clear Bible, Inc.
URL	https://github.com/Clear-Bible/macula-greek
Upstream path	`SBLGNT/lowfat/*.xml`
Base Text	SBL Greek New Testament (SBLGNT), CC-BY-4.0
License	CC-BY-4.0

MACULA Greek provides syntactically annotated treebanks of the Greek New Testament. This dataset is built from the SBLGNT lowfat XML edition, which migrates the Nestle 1904 (N1904) word-level annotations into the SBL Greek New Testament base text where word mappings exist, and supplies additional analysis from Clear Bible's SBLGNT syntax trees for readings unique to SBLGNT.

Table Relationships

sentence
  └── wg (word groups: clauses, phrases)
        ├── wg_wg (parent → child hierarchy)
        └── wg_token (word group → token membership)

token (word-level, linked to sentence via sentence_id)

sentence → top-level sentence containers
wg → syntactic word groups (clauses, phrases) within sentences
wg_wg → tree edges: which word groups contain which sub-groups
wg_token → leaf edges: which tokens belong to which word groups (transitive: every ancestor wg lists every descendant token)
token → individual words with morphology, semantics, glosses

Configs and Schemas

`token`

Column	Type
`token_id`	`large_string`
`node_id`	`large_string`
`morph_id`	`large_string`
`ref`	`large_string`
`book_id`	`large_string`
`book_num`	`int64`
`book_name`	`large_string`
`book_name_long`	`large_string`
`chapter`	`int64`
`verse`	`int64`
`token_in_verse`	`int64`
`role`	`large_string`
`token_class`	`large_string`
`type`	`large_string`
`junction`	`large_string`
`discontinuous`	`large_string`
`english`	`large_string`
`mandarin`	`large_string`
`gloss`	`large_string`
`note`	`large_string`
`text`	`large_string`
`after`	`large_string`
`unicode`	`large_string`
`normalized`	`large_string`
`lemma`	`large_string`
`strong`	`large_string`
`morph`	`large_string`
`formal_tag`	`large_string`
`person`	`large_string`
`number`	`large_string`
`gender`	`large_string`
`case_`	`large_string`
`tense`	`large_string`
`voice`	`large_string`
`mood`	`large_string`
`degree`	`large_string`
`domain`	`large_string`
`ln`	`large_string`
`frame`	`large_string`
`subjref`	`large_string`
`referent`	`large_string`
`sentence_id`	`large_string`

`sentence`

Column	Type
`sentence_id`	`large_string`
`book_id`	`large_string`
`sentence_index`	`int64`

`wg`

Column	Type
`wg_id`	`large_string`
`sentence_id`	`large_string`
`book_id`	`large_string`
`wg_index`	`int64`
`class`	`large_string`
`role`	`large_string`
`rule`	`large_string`
`type`	`large_string`
`junction`	`large_string`
`clauseType`	`large_string`
`articular`	`large_string`
`node_id`	`large_string`
`attrs_json`	`large_string`

`wg_wg`

Column	Type
`parent_wg_id`	`large_string`
`child_wg_id`	`large_string`
`ord`	`int64`

`wg_token`

Column	Type
`wg_id`	`large_string`
`token_id`	`large_string`
`ord`	`int64`

Usage

from datasets import load_dataset

tokens = load_dataset("NuBerea/macula-sblgnt-syntax", "token")
wg = load_dataset("NuBerea/macula-sblgnt-syntax", "wg")
tree = load_dataset("NuBerea/macula-sblgnt-syntax", "wg_wg")

# All tokens in Matthew
matt = tokens["train"].filter(lambda x: x["book_id"] == "Matt")

# Clause-level word groups
clauses = wg["train"].filter(lambda x: x["class"] == "cl")

Text-Fabric projection

The same annotation is also published under tf-projection/ as a single unified node graph, for tools that expect a Text-Fabric corpus rather than relational tables. No extra content — it is a re-packaging of the five configs above.

tf-projection/tf/*.tf              Text-Fabric source (load with text-fabric)
tf-projection/parquet/corpus.parquet   one row per node (DuckDB / SQL)
tf-projection/meta.json            warp + otext + feature map

Node type	Count
`word` (slots)	137,741
`wg`	101,170
`sentence`	8,010
`verse`	7,939
`chapter`	260
`book`	27

Beyond the relational tables the projection adds canonical NT book ordering (book_id sorts lexically, which would scramble reading order), synthesised book/chapter/verse section nodes, and it turns the reference-valued columns into real graph edges: wg_wg, subjref, referent, and frame (parsed into role-labelled semantic-role edges).

import duckdb

duckdb.sql("""
  select ref, text, lemma, morph
  from 'tf-projection/parquet/corpus.parquet'
  where __tf_otype = 'word' and book_id = 'Matt'
  limit 10
""")

Build it with build_tf_projection.py, which reads ./data.

Reproducibility

The dataset is reproducible from the MACULA Greek SBLGNT lowfat XML source. See create_dataset.py, which fetches upstream at a pinned commit and records it in SOURCE.json, then build_tf_projection.py for the projection.

Attribution

This dataset is derived from MACULA Greek Linguistic Datasets (SBLGNT edition), available at https://github.com/Clear-Bible/macula-greek/

Licensed under CC BY 4.0.

Citation

@misc{macula-sblgnt-syntax,
  title = {MACULA Greek (SBLGNT) Syntax Trees},
  author = {Biblica and Clear Bible},
  year = {2024},
  url = {https://github.com/Clear-Bible/macula-greek/},
  note = {SBL Greek New Testament with syntactic tree annotations (lowfat)}
}

Downloads last month: 116

Total file size:

97.1 MB