Dataset Viewer
Auto-converted to Parquet Duplicate
part_id
int64
400B
660B
word_id
int64
40B
66B
n
int64
1
5
start
int64
0
16
end
int64
1
20
text
stringlengths
1
16
kind
stringclasses
7 values
features
stringlengths
3
25
⌀
400,010,010,011
40,001,001,001
1
0
4
Βίβλ
stem
null
400,010,010,012
40,001,001,001
2
4
6
ος
ending
N-NSF
400,010,010,021
40,001,001,002
1
0
5
γενέσ
stem
null
400,010,010,022
40,001,001,002
2
5
8
εως
ending
N-GSF
400,010,010,031
40,001,001,003
1
0
5
Ἰησοῦ
stem
null
400,010,010,041
40,001,001,004
1
0
5
Χριστ
stem
null
400,010,010,042
40,001,001,004
2
5
7
οῦ
ending
N-GSM
400,010,010,051
40,001,001,005
1
0
2
υἱ
stem
null
400,010,010,052
40,001,001,005
2
2
4
οῦ
ending
N-GSM
400,010,010,061
40,001,001,006
1
0
6
Δαυεὶδ
stem
null
400,010,010,071
40,001,001,007
1
0
2
υἱ
stem
null
400,010,010,072
40,001,001,007
2
2
4
οῦ
ending
N-GSM
400,010,010,081
40,001,001,008
1
0
6
Ἀβραάμ
stem
null
400,010,020,011
40,001,002,001
1
0
6
Ἀβραὰμ
stem
null
400,010,020,021
40,001,002,002
1
0
1
ἐ
augment
augment=syllabic
400,010,020,022
40,001,002,002
2
1
7
γέννησ
stem
null
400,010,020,023
40,001,002,002
3
7
9
εν
ending
V-AAI-3S
400,010,020,031
40,001,002,003
1
0
3
τὸν
stem
null
400,010,020,041
40,001,002,004
1
0
5
Ἰσαάκ
stem
null
400,010,020,061
40,001,002,006
1
0
2
δὲ
particle
null
400,010,020,051
40,001,002,005
1
0
5
Ἰσαὰκ
stem
null
400,010,020,071
40,001,002,007
1
0
1
ἐ
augment
augment=syllabic
400,010,020,072
40,001,002,007
2
1
7
γέννησ
stem
null
400,010,020,073
40,001,002,007
3
7
9
εν
ending
V-AAI-3S
400,010,020,081
40,001,002,008
1
0
3
τὸν
stem
null
400,010,020,091
40,001,002,009
1
0
5
Ἰακώβ
stem
null
400,010,020,111
40,001,002,011
1
0
2
δὲ
particle
null
400,010,020,101
40,001,002,010
1
0
5
Ἰακὼβ
stem
null
400,010,020,121
40,001,002,012
1
0
1
ἐ
augment
augment=syllabic
400,010,020,122
40,001,002,012
2
1
7
γέννησ
stem
null
400,010,020,123
40,001,002,012
3
7
9
εν
ending
V-AAI-3S
400,010,020,131
40,001,002,013
1
0
3
τὸν
stem
null
400,010,020,141
40,001,002,014
1
0
4
Ἰούδ
stem
null
400,010,020,142
40,001,002,014
2
4
6
αν
ending
N-ASM
400,010,020,151
40,001,002,015
1
0
3
καὶ
particle
null
400,010,020,161
40,001,002,016
1
0
4
τοὺς
stem
null
400,010,020,171
40,001,002,017
1
0
5
ἀδελφ
stem
null
400,010,020,172
40,001,002,017
2
5
8
οὺς
ending
N-APM
400,010,020,181
40,001,002,018
1
0
3
αὐτ
stem
null
400,010,020,182
40,001,002,018
2
3
5
οῦ
ending
P-GSM
400,010,030,021
40,001,003,002
1
0
2
δὲ
particle
null
400,010,030,011
40,001,003,001
1
0
4
Ἰούδ
stem
null
400,010,030,012
40,001,003,001
2
4
6
ας
ending
N-NSM
400,010,030,031
40,001,003,003
1
0
1
ἐ
augment
augment=syllabic
400,010,030,032
40,001,003,003
2
1
7
γέννησ
stem
null
400,010,030,033
40,001,003,003
3
7
9
εν
ending
V-AAI-3S
400,010,030,041
40,001,003,004
1
0
3
τὸν
stem
null
400,010,030,051
40,001,003,005
1
0
5
Φαρὲς
stem
null
400,010,030,061
40,001,003,006
1
0
3
καὶ
particle
null
400,010,030,071
40,001,003,007
1
0
3
τὸν
stem
null
400,010,030,081
40,001,003,008
1
0
4
Ζαρὰ
stem
null
400,010,030,091
40,001,003,009
1
0
2
ἐκ
particle
null
400,010,030,101
40,001,003,010
1
0
3
τῆς
stem
null
400,010,030,111
40,001,003,011
1
0
5
Θάμαρ
stem
null
400,010,030,131
40,001,003,013
1
0
2
δὲ
particle
null
400,010,030,121
40,001,003,012
1
0
5
Φαρὲς
stem
null
400,010,030,141
40,001,003,014
1
0
1
ἐ
augment
augment=syllabic
400,010,030,142
40,001,003,014
2
1
7
γέννησ
stem
null
400,010,030,143
40,001,003,014
3
7
9
εν
ending
V-AAI-3S
400,010,030,151
40,001,003,015
1
0
3
τὸν
stem
null
400,010,030,161
40,001,003,016
1
0
5
Ἐσρώμ
stem
null
400,010,030,181
40,001,003,018
1
0
2
δὲ
particle
null
400,010,030,171
40,001,003,017
1
0
5
Ἐσρὼμ
stem
null
400,010,030,191
40,001,003,019
1
0
1
ἐ
augment
augment=syllabic
400,010,030,192
40,001,003,019
2
1
7
γέννησ
stem
null
400,010,030,193
40,001,003,019
3
7
9
εν
ending
V-AAI-3S
400,010,030,201
40,001,003,020
1
0
3
τὸν
stem
null
400,010,030,211
40,001,003,021
1
0
4
Ἀράμ
stem
null
400,010,040,021
40,001,004,002
1
0
2
δὲ
particle
null
400,010,040,011
40,001,004,001
1
0
4
Ἀρὰμ
stem
null
400,010,040,031
40,001,004,003
1
0
1
ἐ
augment
augment=syllabic
400,010,040,032
40,001,004,003
2
1
7
γέννησ
stem
null
400,010,040,033
40,001,004,003
3
7
9
εν
ending
V-AAI-3S
400,010,040,041
40,001,004,004
1
0
3
τὸν
stem
null
400,010,040,051
40,001,004,005
1
0
8
Ἀμιναδάβ
stem
null
400,010,040,071
40,001,004,007
1
0
2
δὲ
particle
null
400,010,040,061
40,001,004,006
1
0
8
Ἀμιναδὰβ
stem
null
400,010,040,081
40,001,004,008
1
0
1
ἐ
augment
augment=syllabic
400,010,040,082
40,001,004,008
2
1
7
γέννησ
stem
null
400,010,040,083
40,001,004,008
3
7
9
εν
ending
V-AAI-3S
400,010,040,091
40,001,004,009
1
0
3
τὸν
stem
null
400,010,040,101
40,001,004,010
1
0
7
Ναασσών
stem
null
400,010,040,121
40,001,004,012
1
0
2
δὲ
particle
null
400,010,040,111
40,001,004,011
1
0
7
Ναασσὼν
stem
null
400,010,040,131
40,001,004,013
1
0
1
ἐ
augment
augment=syllabic
400,010,040,132
40,001,004,013
2
1
7
γέννησ
stem
null
400,010,040,133
40,001,004,013
3
7
9
εν
ending
V-AAI-3S
400,010,040,141
40,001,004,014
1
0
3
τὸν
stem
null
400,010,040,151
40,001,004,015
1
0
6
Σαλμών
stem
null
400,010,050,021
40,001,005,002
1
0
2
δὲ
particle
null
400,010,050,011
40,001,005,001
1
0
6
Σαλμὼν
stem
null
400,010,050,031
40,001,005,003
1
0
1
ἐ
augment
augment=syllabic
400,010,050,032
40,001,005,003
2
1
7
γέννησ
stem
null
400,010,050,033
40,001,005,003
3
7
9
εν
ending
V-AAI-3S
400,010,050,041
40,001,005,004
1
0
3
τὸν
stem
null
400,010,050,051
40,001,005,005
1
0
4
Βόες
stem
null
400,010,050,061
40,001,005,006
1
0
2
ἐκ
particle
null
400,010,050,071
40,001,005,007
1
0
3
τῆς
stem
null
400,010,050,081
40,001,005,008
1
0
5
Ῥαχάβ
stem
null
400,010,050,101
40,001,005,010
1
0
2
δὲ
particle
null
End of preview. Expand in Data Studio

Word Parts: Greek New Testament and Hebrew Bible

This dataset cuts every word of the Greek New Testament and of the Hebrew Bible into its parts. Each part has a stable id keyed to MACULA's ids (Clear Bible). Read in order, a word's parts give its printed form exactly.

  • Greek: preverb, augment, reduplication, stem and ending. Stems follow the grammar books.
  • Hebrew: MACULA's own parts (prefixed particles, stem, pronominal suffix), cut once more inside: the yiqtol's person prefix, the verb's person and number ending, and the gender, number and state ending of nouns, adjectives and participles.

What it is for:

  • aligning a translation to the source by parts, not only by whole words;
  • keying data that belongs to a part, such as an ending, a prefix or a suffix;
  • drafting a word's form from its stem and its endings;
  • looking up words that share a stem.
Version v1.0 (schema 2), built 2026-11-22 by shoresh/macula/release_word_parts_v1.py (bcv-query f9d056f)
Greek 137,779 words of the Nestle 1904 New Testament (MACULA Greek), 220,871 parts, 12-digit ids
Hebrew 469,476 MACULA parts of the Westminster Leningrad Codex (MACULA Hebrew), 578,786 parts, 13-digit ids
Licence CC BY 4.0

The id standard is in ID-STANDARD.md: the shape of the ids, the kinds of part, the letter spans, and the rules for how a later release may change them. In short:

  • Greek: a part id is MACULA's word id (11 digits, without the n) followed by one part digit: 400010010011.
  • Hebrew: a part id is MACULA's 12-digit part id followed by one sub-part digit: 0100100100311 אֱלֹהִ, 0100100100312 ים. Where a MACULA part is not cut, its sub-part is 1. MACULA's id is therefore always the first 12 digits, so data keyed to MACULA Hebrew ids joins unchanged.

Files

file one row per columns
greek/parts.tsv part of a Greek word part_id, word_id, n (order in the word, from 1), start, end (letters within the word), text, kind, features
greek/words.tsv Greek word word_id, lead, surface, trail (punctuation), lemma, morph (Robinson's code as MACULA gives it), basis (which source gave the stem), cut (the parts joined by |), flags
hebrew/parts.tsv piece of a MACULA Hebrew part part_id, macula_id (MACULA's part id), word_id (its first 11 digits), n (order in the MACULA part, from 1), start, end (letters within the MACULA part), text, kind, features (MACULA's morph code on prefixes, endings and fused pieces)
*/checks.md acceptance figures
greek/sample_200.tsv, greek/sample_review.md; hebrew/sample_200.tsv, hebrew/review.md a seeded sample of 200, read by hand, and the review
greek/recheck/ the Greek stems compared with greek-inflexion (verbs) and Dodson (nouns)
hebrew/checks_bhsa.md the Hebrew pieces measured against BHSA's morpheme marks: counts only
*/churn-from-v0.1/ id_map.tsv (v0.1 part id -> the v1.0 part ids sharing a letter with it, and the kind of change), changed_words.tsv, churn.md
*/build.json, build.json sources with commits or hashes, code, counts
LICENSES/ the licence notices of greek-inflexion (MIT) and Dodson's lexicon (CC0)
SHA256SUMS the sha256 of every file

A letter is a base character with its combining marks. Spans count from 0, the end is not included, and punctuation is not counted.

Kinds of part:

kind language what
preverb Greek the prefix of a compound verb, as the form writes it (ἀπ-, ἀφ-, συμ-); features preverb=
augment Greek the syllabic augment ἐ-; a temporal augment (a lengthened vowel) is never cut
reduplication Greek consonant + ε, ε or ει
prefix Hebrew the yiqtol's person prefix א ת י נ
stem both the stem
ending both Greek: the ending, with Robinson's code. Hebrew: the verb's person / number ending; the gender / number / state ending of nouns, adjectives and participles
fused both stem and ending merged in the letters, one part carrying the parse (Greek ποιῶ, ὄνομα; Hebrew כָּרַתִּי, where the ending's letter is the root's last letter written once)
particle both a Greek function word (one part), a Hebrew prefixed particle
suffix Hebrew a pronominal suffix (MACULA's)

Not cut, on purpose.

  • Hebrew:
    • the stem formatives: hiphil ה-, niphal נ-, hitpael הִתְ-, and the participle's מְ-;
    • the infinitive absolute;
    • proper nouns, pronouns and particles;
    • the Aramaic parts (5,503), which a later release may cut.
  • Greek:
    • the article and the relative pronouns;
    • indeclinable words, crasis forms, elided forms and adverbs.

How the cut is made

Greek. The stems follow the grammar books:

  • greek-inflexion's stems for verbs;
  • Dodson's lexicon for nouns and adjectives;
  • a short written list for a few closed-class words.

Where no source stem fits the form, our own rule from v0.1 is used. The basis column of greek/words.tsv says which source gave each word's stem.

Hebrew. The cut follows the grammar books' tables, written out in shoresh/macula/build_hebrew_parts.py and applied to MACULA's morph code for each part. A cut never falls inside a letter, since a vowel sign belongs to the letter it is written on. Two rules decide the harder cases:

  • An ending begins with ו or י only when that letter is a vowel letter.
  • A feminine ת is an ending unless the root in BDB's root groups ends in ת: בַּ|ת, דֶּלֶ|ת, but שַׁבָּת stays whole.

Checks

  • Greek:
    • 0 rebuild failures (lead + parts + trail against the surface), 0 duplicate ids, at most 5 parts per word;
    • the same stem boundary as greek-inflexion in 98.5% of the comparable verbs, and as Dodson in 94.7% of the comparable nouns. Most of the remaining nouns are Ἰησοῦς, which we cut by our rule on purpose.
  • Hebrew:
    • 0 rebuild failures, 0 duplicate ids;
    • against BHSA's morpheme marks, as written: the prefix has the same letters in 100.0% of 272,468 compared parts, and the ending in 98.9% of 272,078. The deliberate differences are counted in hebrew/checks_bhsa.md and explained in hebrew/review.md.
  • Both: the seeded sample of 200 per language was read by hand.

Changes from v0.1

Every v0.1 part id maps onto v1.0 through id_map.tsv, by the letters each part covers.

  • Greek: Words compared: 137,779. Words whose cut changed: 9,889 (7.2%). Old part ids whose meaning changed (anything but same): 19,047 of 223,317 (8.5%).
  • Hebrew: MACULA parts compared: 469,476. MACULA parts whose cut changed: 101,338 (21.6%). Old part ids whose meaning changed (anything but same): 101,338 of 469,476 (21.6%). In v0.1 a Hebrew part was MACULA's part (12 digits). In v1.0 a part that is not cut is that id followed by 1, which the map counts as same. A part that is cut is split. The 117 fused parts stay one piece.

v0.1 stays downloadable at tag v0.1. Its sister dataset bcv-commons/word-parts-sa (route b, CC BY-SA 4.0) is retired with this release: v1.0 takes its stems from greek-inflexion (MIT) and Dodson (public domain) instead of MorphGNT and Morpheus, so the share-alike version is no longer needed.

Stability

  • A release is frozen: pin a tag and its revision.
  • Word ids and MACULA Hebrew part ids never change.
  • A correction comes only as a new release, which ships an id_map.tsv from the previous one.

The full rules are in ID-STANDARD.md.

Licence and credit

CC BY 4.0. Credit:

BHSA (CC BY-NC 4.0) was read only to measure the Hebrew cut. No BHSA value is in the data, and hebrew/checks_bhsa.md holds counts only.

Downloads last month
52