Complete DSS corpus with biblical and non-biblical scrolls
Rich linguistic annotations including lemmas, parts of speech, and morphological features
Text-critical annotations (corrections, uncertainties, alternatives)
Multiple transcription formats (Unicode, ETCBC, original)
Text Fabric data converted to Hugging Face Dataset format

The DSS is one of the most important manuscript sources for ancient Hebrew, providing invaluable insights into Biblical Hebrew grammar and textual transmission. This dataset is developed as part of the CACCHT project.

Dataset Statistics

Total words: 500,991
Rows: 500,991 word-level rows
Coverage: Complete ETCBC/DSS 1.9 edition
Source: Text Fabric (TF) format from ETCBC (dss/tf/1.9)
Languages: Ancient Hebrew, Aramaic, and other ancient languages

Features

Feature	Type	Description
`scroll`	`string`	Scroll identifier (e.g., `1Q5`, `1QS`).
`fragment`	`string`	Fragment identifier within the scroll.
`line`	`string`	Line identifier within the fragment.
`line_nr`	`string`	Line number label (mirrors `line` in DSS 1.9).
`book_id`	`string`	Canonical OSIS book ID when the fragment is biblical.
`book_num`	`int32`	Standard numeric book code when the fragment is biblical.
`book_name`	`string`	Canonical short book name when the fragment is biblical.
`book_name_long`	`string`	Canonical long-form book name when defined for biblical material.
`chapter`	`string`	Biblical chapter label when applicable.
`verse`	`string`	Biblical verse label when applicable.
`halfverse`	`string`	Biblical half-verse marker (`a` or `b`) when applicable.
`word_position`	`int32`	Position within line (1-indexed).
`text`	`string`	Primary Hebrew Unicode word form.
`text_original`	`string`	Original-source full word form.
`text_etcbc`	`string`	ETCBC transliterated full word form.
`glyph`	`string`	Unicode display glyph.
`glyph_etcbc`	`string`	ETCBC transliteration glyph.
`glyph_original`	`string`	Original transcription glyph.
`lemma`	`string`	Dictionary lemma.
`lemma_etcbc`	`string`	ETCBC lemma format.
`pos`	`string`	Part of speech.
`person`	`string`	Person annotation.
`gender`	`string`	Gender annotation.
`number`	`string`	Number annotation.
`verbal_tense`	`string`	Verbal tense/aspect annotation.
`state`	`string`	Noun or adjective state annotation.
`person_2`	`string`	Person annotation for morpheme slot 2.
`person_3`	`string`	Person annotation for morpheme slot 3.
`gender_2`	`string`	Gender annotation for morpheme slot 2.
`gender_3`	`string`	Gender annotation for morpheme slot 3.
`number_2`	`string`	Number annotation for morpheme slot 2.
`number_3`	`string`	Number annotation for morpheme slot 3.
`pos_etcbc`	`string`	ETCBC part-of-speech code.
`person_etcbc`	`string`	ETCBC person code.
`gender_etcbc`	`string`	ETCBC gender code.
`number_etcbc`	`string`	ETCBC number code.
`verbal_tense_etcbc`	`string`	ETCBC verbal tense/aspect code.
`punctuation`	`string`	Punctuation markers.
`punct_original`	`string`	Original punctuation encoding.
`punct_etcbc`	`string`	ETCBC punctuation encoding.
`alternative`	`string`	Alternative material marker.
`correction`	`string`	Correction status.
`uncertain`	`string`	Uncertainty markers.
`language`	`string`	Language code.
`biblical_source`	`string`	Biblical source marker from TF.
`space_after`	`string`	Spacing information after the word.
`interlinear`	`string`	Interlinear line marker when present.
`script`	`string`	Script identifier.
`source_line`	`string`	Source data line number (`srcLn`).
`morphology_tag`	`string`	Original compact morphology tag (`morpho`).
`occurrence`	`string`	Occurrence number (`occ`).
`type`	`string`	Token/type classification from TF.

Usage

from datasets import load_from_disk

# Load the dataset
dataset = load_from_disk("dead-sea-scrolls-dataset")

# Filter by scroll
scroll_1q5 = dataset.filter(lambda x: x["scroll"] == "1Q5")
print(f"Scroll 1Q5 has {len(scroll_1q5):,} words")

# Filter by part of speech
verbs = dataset.filter(lambda x: x["pos"] == "verb")

# Filter biblical content
biblical = dataset.filter(lambda x: x["biblical_source"] == "1")
print(f"Biblical content: {len(biblical):,} words")

# Filter by biblical reference
num_23_5 = dataset.filter(
  lambda x: x["book_id"] == "Num" and x["chapter"] == "23" and x["verse"] == "5"
)
print(f"Num 23:5 rows: {len(num_23_5):,}")

# Filter by language
hebrew = dataset.filter(lambda x: x["language"] == "h")

# Print dataset info
print(f"Total words: {len(dataset):,}")
print(dataset)

Data Quality

The DSS dataset includes:

Text-critical annotations marking corrections and uncertainties
Multiple transcription formats for scholarly study
Precise morphological parsing by ETCBC experts
ETCBC and original transcriptions for comparison
Comprehensive coverage of all known DSS fragments

Morphological Features

The dataset includes both nominal and verbal morphological features:

Nominal features:

Gender (masculine, feminine, common)
Number (singular, dual, plural)
Case (Hebrew has fewer case distinctions than related languages)

Verbal features:

Person (1st, 2nd, 3rd)
Gender (masculine, feminine)
Number (singular, dual, plural)
Tense/Mood (indicated through verbal form and conjugation)
ETCBC code variants are included in *_etcbc fields

Morpheme-level suffix features:

person_2, person_3
gender_2, gender_3
number_2, number_3

Parts of Speech:

noun - nouns and nominal forms
verb - verbs and verbal forms
adj - adjectives
det - determiners (articles)
prep - prepositions
conj - conjunctions
adv - adverbs
intj - interjections
pron - pronouns
other

License

This dataset follows the CC BY-NC 4.0 data license metadata distributed with the TF corpus (dss/tf/1.9/otext.tf).

The original DSS texts and encodings come from:

ETCBC/DSS project: https://github.com/ETCBC/dss
Martin Abegg for providing the original data files
Text Fabric infrastructure: https://github.com/annotation/text-fabric/

Note: the repository also contains an MIT license file for software components. For dataset redistribution/use, the TF data metadata specifies CC BY-NC 4.0.

References

Jacobs, J., Naaijer, M., & Roorda, D. (2020). Dead Sea Scrolls in Text-Fabric. https://github.com/ETCBC/dss
CACCHT Project: Creating Annotated Corpora of Classical Hebrew Texts https://github.com/CACCHT
Roorda, D. (2018). Text-Fabric: text as a fabulous data structure. https://github.com/annotation/text-fabric/
Abegg Jr., M. M., Nitzan, B. (1999-2012). The Dead Sea Scrolls Concordance. https://www.deadseascrolls.org.il/

Citation

If you use this dataset, please cite:

@dataset{etcbc_dss_2020,
  title={Dead Sea Scrolls (DSS) Morphology Dataset},
  author={Jacobs, Jarod and Naaijer, Martijn and Roorda, Dirk},
  year={2020},
  publisher={ETCBC and DANS},
  url={https://github.com/ETCBC/dss},
  license={CC-BY-NC-4.0}
}

Notes

The Text Fabric source data uses slot-based indexing; this dataset converts it to row-based format for accessibility
Word positions are calculated relative to line boundaries
Empty or whitespace-only entries have been filtered out
Multiple transcription formats are provided for scholarly flexibility
Text-critical annotations help identify reconstructed, corrected, and uncertain material
The dataset reunites some scrolls that were split between biblical and non-biblical source files
A small number of duplicate lines (14 lines) are resolved with preference given to the biblical version
The glyph variants (glyph_etcbc, glyph_original) provide different scholarly transcription conventions
Biblical references are exposed in book_id, chapter, verse, and halfverse where available

Related Datasets

BHSA - Biblical Hebrew Semantic Alignment
Hebrew Lexicon Dataset
Septuagint (LXX)

Downloads last month: 84

Total file size:

360 MB