hmcgovern/original-language-bibles-greek

About this dataset

This data is compiled from STEP Bible (https://www.stepbible.org/), produced by Biblical scholars at Tyndale House. Usage under CC BY 4.0.

This dataset contains high-quality, annotated text of the Greek New Testament with accompanying word-level translations. Each line in the dataset contains the columns:

• reference: the unique location of the word as <book>.<chapter>.<verse>.<word_number>
• text: the Koine Greek text
• transliteration: the English transliteration
• translation: word-level translation
• dStrong: Strong's number, disambiguated as described at https://www.tinyURL.com/STEP-Disambiguated as well as grammatical parsing
• manuscript_source: which manuscript(s) record the word. A key to the options, from STEP's documentation, follows:

Summary of words included in the Amalgamated Greek NT:

Types of Words	Numbers Found in Types of Manuscripts	Usage in Types of Bibles	Significance	Other Variants
NKO / NK(O) / NK(o)	133,608 words that are identical in virtually all manuscripts	Translated in almost all Bibles	94% of the total words	714 of which differ in Other manuscripts
N(K)(O) / N(k)(o) ...	3,084 words that are different in Traditional manuscripts	Might be translated differently	470 of which alter the translation	14 of which differ in Other manuscripts
K(O) / k(o)	4,164 words found in Traditional but not Ancient manuscripts	Translated in KJV-style Bibles	2,347 of which alter the translation	33 of which differ in Other manuscripts
N(O) / n(o)	896 words found in Ancient but not Traditional manuscripts	Translated in modern Bibles	274 of which alter the translation	3 of which differ in Other manuscripts
O / o	284 words in manuscripts different from Ancient and Traditional	Rarely translated in any Bible	111 of which alter the translation

Motivation

There is great interest in applying NLP tasks (author attribution, question answering, topic modeling, etc.) to Biblical text; however, progress is hampered by several issues:

• More often than not, I see Biblical NLP work done in translation, specifically with the KJV, presumably because it is readily available online and is no longer under copyright. The KJV lacks input from early manuscript evidence, uses outdated language, and leaves something to be desired in terms of translation accuracy. Furthermore, many tasks such as author attribution, require original language data to be meaningful.
• Digitized corpora of ancient manuscript texts are often incomplete or contain duplications or misalignments.
• No other digitized original language corpus that I have seen takes into account different versification traditions (Hebrew vs. English) and contains word-level annotations for manuscript attestation.

STEP Bible provides high-quality data, carefully assembled by subject matter experts, and it is my hope that a readily accessible format as a huggingface dataset will encourage new directions and depth of research into Biblical NLP.

Citation Information

If you use this data, please cite:

BibTeX:

@dataset{original_language_bibles,
  author       = {Hope McGovern},
  title        = {Original Language Bible Corpus: Ancient Hebrew Bible and Ancient Greek New Testament with Word-Level Translations},
  year         = {2024},
  publisher    = {Hugging Face Hub},
  url          = {https://huggingface.co/hmcgovern/original-language-bibles-greek},
  note         = {Includes word-level annotations and English translations for the Ancient Hebrew Bible (Old Testament) and Ancient Greek New Testament based on earliest manuscript evidence, compiled from STEP Bible (https://www.stepbible.org/).}
}