Hebrew Dependency Parsing: Initial Results

What it is about

The authors derive a Hebrew dependency treebank (DepTB) from version 2 of the Hebrew constituency treebank. The conversion makes conjunctions the head of coordinated structures, prepositions the head of prepositional phrases and relativizers the head of clauses, and it fixes some errors found in the source. They train two parsers, MSTParser (first and second order) and MaltParser, on sections 2–12 and test unlabelled attachment on section 1. With gold segmentation and tags, the best model (second order, full lexicon) reaches 84.31% and MaltParser 80.77%. With automatic segmentation and tagging, the best falls to 76.38%. Graph-based models do better, partly because coordination needs many attachment decisions to wait for the final conjunction. Morphological features add little when the morphology is gold. Their metric counts a predicted pair as wrong when one of its segments is missing from the gold parse.

Ideas

• I1. Report two numbers for every parser run, one on gold segmentation and one on the automatic segmenter, and count a predicted segment absent from gold as an error. This makes a segmenter gain or loss visible in the parse score.
• I2. Test the conjunction-as-head representation as an option in the Modern Hebrew UD conversion, and measure it on coordinated clauses separately. Their coordination examples (figures 1 and 2) are the test cases: the subject hangs on the coordinator, not on a verb.
• I3. Run an ablation on gparse: with and without morphological features, and with rare lexical items replaced by an unknown token at thresholds 20 and 100. The paper finds that a second-order graph model with pruned lexicon can beat the full one. Check this on Mova's data before adopting it, and keep the full run as the control.

For Mova

I1 is a reporting rule for tools/hbo evaluation. I2 goes into 11. Modern Hebrew in Universal Dependencies and 7. Syntax. I3 is a gparse experiment that must run twice with identical results before any claim.

Topics: Dependency parsing

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.3115/1697236.1697261

@inproceedings{goldberg2009hebrew,
  title = {{Hebrew Dependency Parsing: Initial Results}},
  author = {Goldberg, Yoav and Elhadad, Michael},
  booktitle = {Proceedings of the 11th International Conference on Parsing Technologies (IWPT'09)},
  year = {2009},
  doi = {10.3115/1697236.1697261},
  url = {https://doi.org/10.3115/1697236.1697261}
}

References

14 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Meni Adler. 2007. Hebrew Morphological Disambiguation: An Unsupervised Stochastic Word-based Approach. Ph.D. thesis, Ben-Gurion University of the Negev, Beer-Sheva, Israel. find
2. Sabine Buchholz and Marsi Erwin. 2006. CoNLL-X shared task on multilingual dependency parsing. In Proc. of CoNLL. find our card
3. Marie-Catherine de Marneffe, Bill MacCartney, and Christopher D. Manning. 2006. Generating typed dependency parses from phrase structure parses. In Proc. of LREC. find our card
4. Yoav Goldberg, Meni Adler, and Michael Elhadad. 2006. Noun phrase chunking in hebrew: Influence of lexical and morphological features. In Proc. of COLING/ACL. find our card
5. Yoav Goldberg, Meni Adler, and Michael Elhadad. 2008. EM can find pretty good HMM POS-Taggers (when given a good start). In Proc. of ACL. find
6. Yoav Golderg, Reut Tsarfaty, Meni Adler, and Michael Elhadad. 2009. Enhancing unlexicalized parsing performance using a wide coverage lexicon, fuzzy tag-set mapping, and EM-HMM-based lexical probabilities. In Proc of EACL. find
7. Noemie Guthmann, Yuval Krymolowski, Adi Milea, and Yoad Winter. 2009. Automatic annotation of morpho-syntactic dependencies in a modern hebrew treebank. In Proc of TLT. find
8. Ryan McDonald and Joakim Nivre. 2007. Characterizing the errors of data-driven dependency parsing models. In Proc. of EMNLP. find our card
9. Ryan McDonald. 2006. Discriminative Training and Spanning Tree Algorithms for Dependency Parsing. Ph.D. thesis, University of Pennsylvania. find
10. Joakim Nivre, Johan Hall, and Jens Nillson. 2006. MaltParser: A data-driven parser-generator for dependency parsing. In Proc. of LREC. find
11. Joakim Nivre, Johan Hall, Sandra Kubler, Ryan Mc-Donald, Jens Nilsson, Sebastian Riedel, and Deniz Yuret. 2007. find
12. The CoNLL 2007 shared task on dependency parsing. In Proc. of the EMNLP-CoNLL. find
13. Reut Tsarfaty and Khalil Sima’an. 2008. Relational-realizational parsing. In Proc. of CoLING, August. find
14. Reut Tsarfaty. 2006. Integrated morphological and syntactic disambiguation for modern hebrew. In Proceedings of ACL-SRW. find our card

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/lingbulk-2009-goldberg-hebrew-dependency-parsing-initial-results/