A Universal Dependencies Treebank of Ancient Hebrew

What it is about

The authors build the first UD treebank of Biblical Hebrew, the books of Genesis and Ruth (1,579 sentences). The source is the BHSA, a morphologically tagged table of the Hebrew Bible that has word-level features but only a flat phrase and clause structure. The BHSA tokenization is adjusted to UD (pronominal suffixes are not split in the BHSA, and must be), its tags are mapped to UD, and a rule-based parser of about 300 rules, written in Constraint Grammar, attaches the arcs; punctuation is attached afterwards and every tree is checked by hand. As a consistency test, a UDPipe parser trained on the whole corpus does not reproduce it perfectly; its errors cluster on nmod versus obl attachment after a noun and on where a quotation ends. The treebank is released with the UD project.

Ideas

I1. Suffix splitting as a table in tools/hbo: the BHSA leaves pronominal suffixes attached (the form for "to guard him" is one token), while UD splits them. Write the split as a rule with a test list of suffixed verbs, prepositions and nouns taken from the treebank, and report the count per class.
I2. Use the paper's error analysis as a list of hard classes for the parser: nmod against obl after a noun, and the end of a quotation. Each class gets one measure on held-out sentences, so that a gain in one does not hide a loss in another.
I3. Train gparse on the treebank and parse it back as a sanity check, as the authors do with UDPipe; a model that cannot reproduce its own training data shows the annotation is not yet consistent enough to be memorised.

For Mova

I1 goes into tools/hbo (segmenter tables, the suffix split) and The Noun, the Adjective and the Numeral for the noun suffixes. I2 is a set of tests for tools/hbo and gparse on the hard classes, with the attachment rule itself in Syntax: the clause and the sentence. I3 is a check on treebanks/hbo/ptnk, used only as a test, never as training data for a release (licence NC).

Topics: Universal Dependencies · Treebanks

Cite

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

@inproceedings{swanson2022universal,
  title = {{A Universal Dependencies Treebank of Ancient Hebrew}},
  author = {Swanson, Daniel G. and Tyers, Francis M.},
  booktitle = {Proceedings of the Thirteenth Language Resources and Evaluation Conference},
  year = {2022},
  url = {https://aclanthology.org/2022.lrec-1.252/}
}

References

14 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

Antonsen, L., Trosterud, T., and Wiechetek, L. (2010). Reusing grammatical resources for new languages. In LREC. find
Barry, R. K. (1997). ALA-LC romanization tables-transliteration schemes for non-roman scripts. In Library of Congress, 1997. find
Bick, E. and Didriksen, T. (2015). CG-3—beyond classical constraint grammar. In Proceedings of the 20th Nordic Conference of Computational Linguistics (NODALIDA 2015), pages 31–39. find
Bick, E. (2005). Turning constraint grammar data into running dependency treebanks. In Civit, Montserrat & Kübler, Sandra & Martı́, Ma. Antònia (red.), Proceedings of TLT 2005 (4th Workshop on Treebanks and Linguistic Theory, Barcelona, pages 19–27. find
Eckhoff, H., Bech, K., Bouma, G., Eide, K., Haug, D., Haugen, O. E., and Jøhndal, M. (2018). The proiel treebank family: a standard for early attestations of indo-european languages. Language Resources and Evaluation, 52(1):29–65. find
Haug, D. T. and Jøhndal, M. (2008). Creating a parallel treebank of the old indo-european bible translations. In Proceedings of the second workshop on language technology for cultural heritage data (LaTeCH 2008), pages 27–34. find
Luukko, M., Sahala, A., Hardwick, S., and Lindén, K. (2020). Akkadian treebank for early neo-assyrian royal inscriptions. In Proceedings of the 19th International Workshop on Treebanks and Linguistic Theories, pages 124–134, Düsseldorf, Germany, October. Association for Computational Linguistics. find our card
McDonald, R. T., Nivre, J., Quirmbach-Brundage, Y., Goldberg, Y., Das, D., Ganchev, K., Hall, K. B., Petrov, S., Zhang, H., Täckström, O., et al. (2013). Universal dependency annotation for multilingual parsing. In Proc. of ACL. find
Nivre, J., de Marneffe, M.-C., Ginter, F., Hajic, J., Manning, C. D., Pyysalo, S., Schuster, S., Tyers, F., and Zeman, D. (2020). Universal dependencies v2: An evergrowing multilingual treebank collection. In Proceedings of the 12th Language Resources and Evaluation Conference, pages 4034–4043. Peursen, W. v., Sikkel, C., and Roorda, D. (2015). Hebrew text database ETCBC4b. find our card
Straka, M. and Straková, J. (2017). Tokenizing, POS tagging, lemmatizing and parsing UD 2.0 with UD-Pipe. In Proceedings of the CoNLL 2017 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies, pages 88–99, Vancouver, Canada, August. Association for Computational Linguistics. find our card
Straka, M., Hajič, J., and Straková, J. (2016). UD-Pipe: Trainable pipeline for processing CoNLL-U files performing tokenization, morphological analysis, POS tagging and parsing. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC’16), pages 4290– 4297, Portorož, Slovenia, May. European Language Resources Association (ELRA). find
Tsarfaty, R. (2013). A unified morpho-syntactic scheme of stanford dependencies. In Proc. of ACL. find
Tyers, F. and Sheyanova, M. (2017). Annotation schemes in North Sámi dependency parsing. In Proceedings of the Third Workshop on Computational Linguistics for Uralic Languages, pages 66–75. find
Zeldes, A. and Abrams, M. (2018). The Coptic Universal Dependency Treebank. In Proceedings of the Universal Dependencies Workshop 2018, pages 192–201, Brussels. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/ud-2022-swanson-universal-dependencies-treebank-ancient-hebrew/