Full syntactic tree annotation of the Hebrew Bible from the MACULA Hebrew Linguistic Dataset. 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).

Dataset Overview

Total rows: 3,064,499 (across all configs)
Configs: 5 (one per relational table)
Languages: Biblical Hebrew

Source

Attribute	Value
Project	MACULA Linguistic Datasets
Authors/Institution	Clear Bible, Inc. and CenterBLC
URL	https://github.com/Clear-Bible/macula-hebrew
License	CC-BY-4.0
Version	Current release

Morphologically tagged and syntactically annotated Hebrew Bible using the Westminster Leningrad Codex (WLC) base text. MACULA datasets provide relational word-level and structural annotations for biblical language study.

Config	Rows	Columns	Description
`token`	475,911	48	Word-level tokens with morphology, glosses, and semantic features
`sentence`	23,213	3	Sentence boundary containers
`wg`	340,101	14	Word groups (clauses, phrases) with syntactic class, role, and rule
`wg_wg`	316,888	3	Parent–child tree edges between word groups
`wg_token`	1,908,386	3	Leaf edges: which tokens belong to which word groups

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
token → individual words with morphology, semantics, glosses

Configs and Schemas

`token`

Column	Type
`token_id`	`large_string`
`ref`	`large_string`
`book_id`	`large_string`
`book_num`	`int64`
`book_name`	`large_string`
`book_name_long`	`null`
`chapter`	`int64`
`verse`	`int64`
`token_in_verse`	`int64`
`role`	`large_string`
`token_class`	`large_string`
`type`	`large_string`
`english`	`large_string`
`mandarin`	`large_string`
`gloss`	`large_string`
`text`	`large_string`
`after`	`large_string`
`lemma`	`large_string`
`normalized`	`large_string`
`strong`	`large_string`
`morph`	`large_string`
`person`	`large_string`
`number`	`large_string`
`gender`	`large_string`
`case_`	`null`
`tense`	`null`
`voice`	`null`
`mood`	`null`
`degree`	`null`
`domain`	`null`
`ln`	`null`
`frame`	`large_string`
`subjref`	`large_string`
`referent`	`null`
`transliteration`	`large_string`
`strongnumberx`	`large_string`
`stronglemma`	`large_string`
`sensenumber`	`large_string`
`greek`	`large_string`
`greekstrong`	`large_string`
`stem`	`large_string`
`lang`	`large_string`
`pos`	`large_string`
`state`	`large_string`
`lexdomain`	`large_string`
`contextualdomain`	`null`
`coredomain`	`large_string`
`sdbh`	`large_string`
`extends`	`null`
`participantref`	`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`	`null`
`junction`	`null`
`cltype`	`null`
`clauseType`	`large_string`
`articular`	`large_string`
`discontinuous`	`null`
`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

# Load a specific config
tokens = load_dataset("NuBerea/macula-hebrew-syntax", "token")
wg = load_dataset("NuBerea/macula-hebrew-syntax", "wg")
tree = load_dataset("NuBerea/macula-hebrew-syntax", "wg_wg")

# Get all tokens in Matthew
matt = tokens["train"].filter(lambda x: x["book_id"] == "MAT")

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

# Reconstruct a syntax tree for a sentence
sent_id = "sent:MAT:1"
groups = wg["train"].filter(lambda x: x["sentence_id"] == sent_id)
edges = tree["train"].filter(
    lambda x: x["parent_wg_id"].startswith("wg:MAT:1:")
    or x["child_wg_id"].startswith("wg:MAT:1:")
)

Attribution

This dataset is derived from MACULA Hebrew Linguistic Datasets, available at https://github.com/Clear-Bible/macula-hebrew/

Licensed under CC BY 4.0.

Citation

@misc{macula-hebrew-syntax,
  title = {MACULA Hebrew Syntax Trees},
  author = {Biblica and Clear Bible},
  year = {2024},
  url = {https://github.com/Clear-Bible/macula-hebrew/},
  note = {Westminster Leningrad Codex with syntactic tree annotations (lowfat)}
}

Downloads last month: 105

Total file size:

327 MB