Identifying Phrase Boundaries in the Samaritan Pentateuch with Machine Learning

What it is about

The full text is not in the library (the publisher's page refused access); this card rests on the abstract as indexed by OpenAlex, and its ideas are provisional. The authors add the first syntactic layer to an open, morphologically annotated dataset of the Samaritan Pentateuch: phrase atoms and phrase boundaries. The markers were produced by a machine-learning parser earlier used for the morphology of Hebrew and Syriac texts (the ETCBC tradition), then corrected by hand because the output was not always right. They present the result as a resource for comparing corpora and for studying how the text was transmitted, and the corrections as better training data for later models.

Ideas

• I1. The Samaritan Pentateuch is the Torah in another tradition: once its data are available (licence to check), align it verse by verse with our MACULA Torah and list the differences that touch grammar (word order, as in the עוֹד examples of Joosten 2025; plene spelling; forms), as a dated table.
• I2. Their workflow (automatic layer, then hand correction, with the corrections kept as training data) is the same as Mova's teacher loop; record the share of phrase boundaries they had to correct, if the paper gives it, as a reference figure for how good a phrase chunker on ancient Hebrew can be.
• I3. Phrase atoms are a level our UD trees do not have; derive them from the trees (a head and its non-clausal dependents) and compare with ETCBC phrase boundaries on the MT Torah to see where our head rules disagree.

For Mova

I1 is for 12. Biblical, Mishnaic and Modern Hebrew compared and 01-script.md (spelling). I2 is a note for chapter 13 (measures). I3 is a check in tools/macula against BHSA phrases, a cheap second gold beside PTNK (BHSA is CC BY-NC: a measure only, never copied).

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.3390/rel17020192

@inproceedings{cantanhede2026identifying,
  title = {{Identifying Phrase Boundaries in the Samaritan Pentateuch with Machine Learning}},
  author = {Cantanhede, Saulo de Oliveira and Naaijer, Martijn and Højgaard, Christian Canu and Glanz, Oliver},
  booktitle = {Religions},
  year = {2026},
  doi = {10.3390/rel17020192},
  url = {https://doi.org/10.3390/rel17020192}
}

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/hbo-2026-cantanhede-identifying-phrase-boundaries-samaritan-pentateuch/