Producing a Parallel Universal Dependencies Treebank of Ancient Hebrew and Ancient Greek via Cross-Lingual Projection

What it is about

The authors build a UD treebank of Greek Genesis and Ruth from the Codex Alexandrinus Septuagint (1,576 sentences, about 39K tokens) by projecting the trees of the Hebrew treebank across the translation. Lemma sequences (punctuation left out) are word-aligned with Eflomal; every Hebrew arc whose two ends are aligned is copied to the Greek tree, a binary choice where the original work kept probabilities. Arborator-Grew rewrite queries then fix systematic mismatches, such as Hebrew smixut (compound:smixut) that is a possessive in Greek, and everything is checked by hand. Cohen's kappa between the projected and the corrected trees is 0.580 for heads and 0.503 for labels, and between two annotators on three chapters it is 0.868 for heads and 0.813 for labels. Adding more text to the aligner helps only a little.

Ideas

• I1. A cross-language disagreement list as free error analysis: align the Greek text to Mova's Hebrew parses at the lemma level, and list the arcs where both ends align but the relation differs. Each entry is a candidate error for tools/hbo or a candidate rule for the Hebrew convention. Used as a test only.
• I2. A relabel rule with a morphological guard: compound:smixut becomes nmod:poss only when the dependent is genitive. The same shape (pattern, condition, new label) can express Mova's own rewrites for construct chains and quantifier nouns, in a table rather than in code.
• I3. Report kappa per relation, not one number. The paper's own result is that the head score is much higher than the label score, and per-relation scores would show which labels a projection can be trusted with.

For Mova

I1 is a test for tools/hbo on the Hebrew parses, with the Greek side used only for measurement (the licence is NC). I2 goes into Syntax: the clause and the sentence for the construct chain and The Noun, the Adjective and the Numeral for the quantifier kol. I3 is a reporting rule for every projection or cross-language test in tools/hbo.

Topics: Universal Dependencies · Treebanks

Cite

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

@inproceedings{swanson2024producing,
  title = {{Producing a Parallel Universal Dependencies Treebank of Ancient Hebrew and Ancient Greek via Cross-Lingual Projection}},
  author = {Swanson, Daniel G. and Bussert, Bryce D. and Tyers, Francis},
  booktitle = {Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)},
  year = {2024},
  url = {https://aclanthology.org/2024.lrec-main.1145/}
}

References

11 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Željko Agić, Anders Johannsen, Barbara Plank, Héctor Martínez Alonso, Natalie Schluter, and Anders Søgaard. 2016. Multilingual projection for parsing truly low-resource languages. Transactions of the Association for Computational Linguistics, 4:301–312. find
2. Kenneth A Cherney. 2014. Allusion as translation problem: Portuguese versions of second Isaiah as test case. Ph.D. thesis, Stellenbosch: Stellenbosch University. find
3. Jacob Cohen. 1960. A coefficient of agreement for nominal scales. Educational and psychological measurement, 20(1):37–46. ↗
4. Gilles Dorival. 2022. The septuagint between textual criticism and redactional criticism. In The Hebrew Bible Manuscripts: A Millennium, pages 175–188. Brill. find
5. Hanne Eckhoff, Kristin Bech, Gerlof Bouma, Kristine Eide, Dag Haug, Odd Einar Haugen, and Marius Jøhndal. 2018. The proiel treebank family: a standard for early attestations of indo-european languages. Language resources and evaluation, 52:29–65. find
6. Gaël Guibon, Marine Courtin, Kim Gerdes, and Bruno Guillaume. 2020. When collaborative treebank curation meets graph grammars. In LREC 2020-12th Language Resources and Evaluation Conference. find
7. Jeremy G Kahn, Matthew Snover, and Mari Ostendorf. 2009. Expected dependency pair match: predicting translation quality with expected syntactic structure. Machine Translation, 23:169– 179. find
8. Joakim Nivre, Marie-Catherine de Marneffe, Filip Ginter, Jan Hajic, Christopher D Manning, Sampo Pyysalo, Sebastian Schuster, Francis Tyers, and Daniel Zeman. 2020. Universal dependencies v2: An evergrowing multilingual treebank collection. In Proceedings of the 12th Language Resources and Evaluation Conference, pages 4034–4043. find our card
9. Robert Östling and Jörg Tiedemann. 2016. Efficient word alignment with Markov Chain Monte Carlo. Prague Bulletin of Mathematical Linguistics, 106:125–146. find
10. Daniel Swanson and Francis Tyers. 2022. A universal dependencies treebank of ancient hebrew. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 2353–2361. find our card
11. Emanuel Tov. 2015. The text-critical use of the Septuagint in biblical research. Penn State Press. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/ud-2024-swanson-producing-parallel-universal-dependencies-treebank/