The Development of Hebrew in Antiquity – A Computational Linguistic Study

What it is about

This is a project report on Qumran Hebrew (QH), the language of the Dead Sea Scrolls, and whether it is a literary imitation of Biblical Hebrew or a spoken dialect that continues it. The authors take lemma-level data from Accordance (commercial, not publishable) for the Hebrew Bible, the Mishnah and the scrolls. They represent each book or tractate as a lemma-frequency vector, compare them with cosine similarity after raw-frequency and TF–IDF weighting, and cluster them with Ward linkage. They also compare verb-stem (binyanim) distributions by Euclidean distance, and extract verb valency by looking up to four words after each verb for a prepositional particle, object marker or pronominal suffix. Preliminary results match known groupings (1–2 Samuel, Kings, Chronicles), and place Tamid, Middot and Yoma with the Qumran texts, which fits their early date. Under TF–IDF, Miqsat Ma'ase ha-Torah (4QMMT) is closer to the Mishnah than to other scrolls. The valency tool misses some complements, and the authors say so; the stem analysis is still early.

Ideas

• I1. A valency extractor as a rule: after a verb, look at the next up to four tokens and classify the complement (prepositional particle, object marker, pronominal suffix, direct object). Check it against the UD trees from MACULA with macula compare, and count how many complements are missed.
• I2. Lemma TF–IDF clustering as a stylistic measure for the macula corpora: split the Torah, the Prophets and the Writings into books, cluster them, and check whether the grouping matches the known order. This is a deterministic, testable baseline that needs no model.
• I3. Binyanim distribution per text: count the stems of all verbs in a text, normalise by the number of verbs, and compare texts by Euclidean distance. The paper's measure is a small table that Mova can compute from the hbo tags.

For Mova

I1 goes into The verb (valency) and is measured by macula compare in tools/macula. I2 is a measure in tools/macula (on the MACULA corpus, with macula dict lemmas) and a dating test for The lexicon. I3 goes into The verb under binyanim. The data is not publishable, so only the method is carried over; any test must use open Hebrew Bible data.

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.18653/v1/2025.alp-1.7

@inproceedings{baitner2025development,
  title = {{The Development of Hebrew in Antiquity – A Computational Linguistic Study}},
  author = {Baitner, Hallel and Duchovny, Dimid and Gottlieb, Lee-Ad and Yorav, Amir and Dershowitz, Nachum and Ratzon, Eshbal},
  booktitle = {Proceedings of the Second Workshop on Ancient Language Processing},
  year = {2025},
  doi = {10.18653/v1/2025.alp-1.7},
  url = {https://doi.org/10.18653/v1/2025.alp-1.7}
}

References

1 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Joseph L. Angel. 2012. Maskil, community, and religious experience in the Songs of the Sage (4Q510– 511). Dead Sea Discoveries, 19:1–27. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/aclbulk-2025-baitner-development-hebrew-antiquity-computational-linguistic/