The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
concepts: string
labeller: string
words: int64
sets: struct<hebrew-nt: struct<words: int64, one concept (no call): int64, answered by jev: int64, no answ (... 499 chars omitted)
child 0, hebrew-nt: struct<words: int64, one concept (no call): int64, answered by jev: int64, no answer: int64, none of (... 183 chars omitted)
child 0, words: int64
child 1, one concept (no call): int64
child 2, answered by jev: int64
child 3, no answer: int64
child 4, none of the concepts: int64
child 5, confidence >= 0.9: int64
child 6, confidence >= 0.8: int64
child 7, confidence >= 0.8 with a concept: int64
child 8, chosen by opus (jev below 0.8): int64
child 9, opus differs from jev: int64
child 1, septuagint: struct<words: int64, one concept (no call): int64, answered by jev: int64, no answer: int64, none of (... 183 chars omitted)
child 0, words: int64
child 1, one concept (no call): int64
child 2, answered by jev: int64
child 3, no answer: int64
child 4, none of the concepts: int64
child 5, confidence >= 0.9: int64
child 6, confidence >= 0.8: int64
child 7, confidence >= 0.8 with a concept: int64
child 8, chosen by opus (jev below 0.8): int64
child 9, opus differs from jev: int64
quoted words: struct<words: int64, same concept: double, confidence >= 0.9: struct<words: int64, same concept: dou (... 280 chars omitted)
child 0, words: int64
child 1, same concept: double
child 2, confidence >= 0.9: struct<words: int64, same concept: double, jev's own: double>
child 0, words: int64
child 1, same concept: double
child 2, jev's own: double
child 3, confidence 0.8 to 0.9: struct<words: int64, same concept: double, jev's own: double>
child 0, words: int64
child 1, same concept: double
child 2, jev's own: double
child 4, confidence 0.6 to 0.8: struct<words: int64, same concept: double, jev's own: double>
child 0, words: int64
child 1, same concept: double
child 2, jev's own: double
child 5, confidence 0 to 0.6: struct<words: int64, same concept: double, jev's own: double>
child 0, words: int64
child 1, same concept: double
child 2, jev's own: double
code: string
licence: string
version: string
built: timestamp[s]
schema: int64
counts: struct<words: int64, labeller: struct<jev: int64, one concept: int64>, set: struct<hebrew-nt: int64, (... 33 chars omitted)
child 0, words: int64
child 1, labeller: struct<jev: int64, one concept: int64>
child 0, jev: int64
child 1, one concept: int64
child 2, set: struct<hebrew-nt: int64, septuagint: int64>
child 0, hebrew-nt: int64
child 1, septuagint: int64
child 3, none: int64
models: string
layer: string
inputs: struct<concepts: string, hebrew_nt: string, words: string>
child 0, concepts: string
child 1, hebrew_nt: string
child 2, words: string
to
{'schema': Value('int64'), 'version': Value('string'), 'built': Value('timestamp[s]'), 'code': Value('string'), 'licence': Value('string'), 'layer': Value('string'), 'models': Value('string'), 'inputs': {'concepts': Value('string'), 'hebrew_nt': Value('string'), 'words': Value('string')}, 'counts': {'words': Value('int64'), 'labeller': {'jev': Value('int64'), 'one concept': Value('int64')}, 'set': {'hebrew-nt': Value('int64'), 'septuagint': Value('int64')}, 'none': Value('int64')}}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 544, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3052, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 558, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 361, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
concepts: string
labeller: string
words: int64
sets: struct<hebrew-nt: struct<words: int64, one concept (no call): int64, answered by jev: int64, no answ (... 499 chars omitted)
child 0, hebrew-nt: struct<words: int64, one concept (no call): int64, answered by jev: int64, no answer: int64, none of (... 183 chars omitted)
child 0, words: int64
child 1, one concept (no call): int64
child 2, answered by jev: int64
child 3, no answer: int64
child 4, none of the concepts: int64
child 5, confidence >= 0.9: int64
child 6, confidence >= 0.8: int64
child 7, confidence >= 0.8 with a concept: int64
child 8, chosen by opus (jev below 0.8): int64
child 9, opus differs from jev: int64
child 1, septuagint: struct<words: int64, one concept (no call): int64, answered by jev: int64, no answer: int64, none of (... 183 chars omitted)
child 0, words: int64
child 1, one concept (no call): int64
child 2, answered by jev: int64
child 3, no answer: int64
child 4, none of the concepts: int64
child 5, confidence >= 0.9: int64
child 6, confidence >= 0.8: int64
child 7, confidence >= 0.8 with a concept: int64
child 8, chosen by opus (jev below 0.8): int64
child 9, opus differs from jev: int64
quoted words: struct<words: int64, same concept: double, confidence >= 0.9: struct<words: int64, same concept: dou (... 280 chars omitted)
child 0, words: int64
child 1, same concept: double
child 2, confidence >= 0.9: struct<words: int64, same concept: double, jev's own: double>
child 0, words: int64
child 1, same concept: double
child 2, jev's own: double
child 3, confidence 0.8 to 0.9: struct<words: int64, same concept: double, jev's own: double>
child 0, words: int64
child 1, same concept: double
child 2, jev's own: double
child 4, confidence 0.6 to 0.8: struct<words: int64, same concept: double, jev's own: double>
child 0, words: int64
child 1, same concept: double
child 2, jev's own: double
child 5, confidence 0 to 0.6: struct<words: int64, same concept: double, jev's own: double>
child 0, words: int64
child 1, same concept: double
child 2, jev's own: double
code: string
licence: string
version: string
built: timestamp[s]
schema: int64
counts: struct<words: int64, labeller: struct<jev: int64, one concept: int64>, set: struct<hebrew-nt: int64, (... 33 chars omitted)
child 0, words: int64
child 1, labeller: struct<jev: int64, one concept: int64>
child 0, jev: int64
child 1, one concept: int64
child 2, set: struct<hebrew-nt: int64, septuagint: int64>
child 0, hebrew-nt: int64
child 1, septuagint: int64
child 3, none: int64
models: string
layer: string
inputs: struct<concepts: string, hebrew_nt: string, words: string>
child 0, concepts: string
child 1, hebrew_nt: string
child 2, words: string
to
{'schema': Value('int64'), 'version': Value('string'), 'built': Value('timestamp[s]'), 'code': Value('string'), 'licence': Value('string'), 'layer': Value('string'), 'models': Value('string'), 'inputs': {'concepts': Value('string'), 'hebrew_nt': Value('string'), 'words': Value('string')}, 'counts': {'words': Value('int64'), 'labeller': {'jev': Value('int64'), 'one concept': Value('int64')}, 'set': {'hebrew-nt': Value('int64'), 'septuagint': Value('int64')}, 'none': Value('int64')}}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Greek NT Concepts (model)
A Hebrew concept for each content word of the Greek New Testament: which concept of hebrew-word-concepts-model a word expresses in its verse. The Greek word and the Hebrew Bible then meet on one key, so that a word list or a translation built for the Hebrew Bible can be read from the New Testament too.
Every label here was chosen by a language model. No person has checked them. Jev (TypeSafe System One) chose for every word with more than one concept. Use the confidence: see Quality.
| Version | v0.1 |
| Built | 2026-11-16 by macula/release_greek_nt_concepts_model.py (bcv-query e07ff2c) |
| Words | 65,808 Greek NT words (MACULA Greek, Nestle 1904) |
| Concepts | bcv-commons/hebrew-word-concepts-model v0.5 (5a7af303) |
| Licence | CC BY-SA 4.0 |
How a word was labelled
Each word was put to the model in its Greek verse, with its lemma and morphology, and a numbered list of concepts to choose from (the concept's definition and the UBS lexicon sense it belongs to). No translation, gloss or Hebrew text was shown. The list depends on the word's set:
hebrew-nt: a Hebrew New Testament (Salkinson-Ginsburg or Delitzsch, links of hebrew-nt-bridge v0.1) renders the word by a Hebrew lexeme that has concepts: the list is that lexeme's concepts (both editions' lexemes where they differ).septuagint: no such lexeme: the list is the ten concepts the Septuagint most often renders by this Greek lexeme, and one more choice, "none of the concepts above". Many of these words are function words or names;none= true marks that choice.
A word whose list holds one concept took it without a model (labeller = one concept).
Files
| Path | What it holds |
|---|---|
data/words.parquet |
one row per Greek NT word labelled |
report.json |
counts by set and confidence, and the check on quoted words |
build.json |
inputs, models, counts |
Columns of data/words.parquet:
| Column | |
|---|---|
greek_id |
MACULA Greek word id (n40001001001) |
greek_strong, greek_lemma |
the word's Strong's number and lemma (MACULA Greek) |
set |
hebrew-nt or septuagint (above) |
concept |
the concept id of hebrew-word-concepts-model; empty where the model chose "none of the concepts" |
hebrew_strong |
the Hebrew lexeme the concept belongs to |
n_concepts |
how many concepts the model chose from |
confidence |
Jev's probability for its choice (0 to 1); empty for one concept |
none |
the model chose "none of the concepts" |
labeller |
jev, opus (Opus's choice, where Jev's confidence was below 0.8), or one concept |
jev_concept |
Jev's own choice (the same as concept unless labeller = opus) |
opus_sure |
for opus rows: Opus said its choice was clear (false: two concepts fit about equally well) |
Quality
There is no hand-checked answer key for the New Testament. The check uses the words where the New Testament quotes the Hebrew Bible and the Greek word stands for a Tanakh word with a concept of its own: there the Tanakh word's concept is taken as the answer. On these 766 words:
- all: 77.3% the same concept as the Tanakh word
- Jev's confidence >= 0.9: 341 words, 95.0%
- Jev's confidence 0.8 to 0.9: 102 words, 88.2%
- Jev's confidence 0.6 to 0.8: 134 words, 64.2%
- Jev's confidence 0 to 0.6: 189 words, 48.7%
These words are easier than the average word (the quotation keeps the Hebrew close), so read the figures as upper figures. Jev's confidence separates good from weak choices well; most of Jev's labels on the quoted words were Jev's own labels in the Tanakh, which favours Jev here.
In make-tanakh-research's drafting tests (2026-11-16), using these labels to choose New Testament words by concept improved drafts of New Testament-only engines in Indonesian, Swahili and Tagalog (chrF3 +0.8 to +1.0 with the labels at confidence 0.8 or above); with the labels below 0.8 included the gain was smaller. Those figures are research's, not part of this release.
Recommended use: confidence >= 0.8 (or labeller = opus with opus_sure) where precision matters; all labels where reach
matters more.
Sources and licence
| Part | Source | Licence |
|---|---|---|
| Greek words, ids, lemmas, morphology | MACULA Greek (Clear Bible), Nestle 1904 | CC BY 4.0 |
| Concepts and their definitions | bcv-commons/hebrew-word-concepts-model v0.5 (concepts within the UBS senses) | CC BY-SA 4.0 |
| Hebrew lexemes of the Hebrew NTs | bcv-commons/hebrew-nt-bridge v0.1 | CC BY 4.0 |
| Septuagint candidates | bcv-commons resources (MACULA Hebrew's Septuagint links; GLAUx, KU Leuven) | CC BY 4.0 / CC BY-SA 4.0 |
| Labels | Jev (jev-latest, TypeSafe System One), one run | model output |
The whole is CC BY-SA 4.0.
Versions
- v0.1: Jev on every word with more than one concept (one run, 2026-11-16).
- Downloads last month
- 57