Modeling Morphosyntactic Agreement in Constituency-Based Parsing of Modern Hebrew

What it is about

The question is whether agreement features (gender, number, person, definiteness, accusativity) help a statistical parser of Hebrew, and how they should be represented. The authors compare two constituency models on the Hebrew treebank, parsing gold-tagged sentences: a state-split model that puts morphology into the category labels (SP-AGR), and an extension of the Relational-Realizational model (RR-AGR) that separates grammatical function from its realisation and marks the agreement controller and the domain (the clause) with its features. Naive state-splits with agreement features get worse as features are added, sometimes sharply with gender (for example from about 80 to 64 F1 on one setting). The RR model improves as features are added, and the best setting (grandparent labels, gender, definiteness and accusativity) reaches F1 84.13 on sentences under 40 words, the best Hebrew result reported at the time. The authors read the gap as a sparseness effect: the RR parameters capture functional patterns that do not depend on word order.

Ideas

• I1. Agreement as a checked relation over the tree: for a subject and its verb, and for a predicate and its subject, compare gender, number and person of controller and target. Run it first as a rule over the converted Hebrew trees to count violations, then use the result as a feature or as a validator in prag.
• I2. Use the morphology of all co-heads and function words in a phrase, not only of its head. The authors find that gender percolates through the head while definiteness and accusativity come from other words, and that adding only the head tag hurts. A Mova feature for a phrase should therefore be a set of morphological values gathered from the phrase, held in a BTreeSet so the result is deterministic.
• I3. Before adding a feature, measure its sparseness: report the number of distinct rule types and the F1 at each feature set, as Table 1 does. This is the cheapest guard against the drop the paper shows.

For Mova

I1 goes into 05. The verb: system, binyanim, tense and use (subject-verb agreement) and Nouns, adjectives, pronouns and numerals (predicate and copula agreement). I2 is a design point for prag and world when a phrase has several morphological carriers. I3 is a rule for any gparse or rule-based experiment on treebanks/he/. The paper parses Modern Hebrew with a constituency model, so only the ideas transfer.

Cite

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

@inproceedings{tsarfaty2010modeling,
  title = {{Modeling Morphosyntactic Agreement in Constituency-Based Parsing of Modern Hebrew}},
  author = {Tsarfaty, Reut and Sima'an, Khalil},
  booktitle = {Proceedings of the NAACL HLT 2010 First Workshop on Statistical Parsing of Morphologically-Rich Languages},
  year = {2010},
  url = {https://aclanthology.org/W10-1405/}
}

References

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

1. Steven Abney. 1997. Stochastic attribute-value grammars. Computational Linguistics, 23(4):597–618. find
2. Stephen R. Anderson. 1982. Where’s morphology? Linguistic Inquiry. find
3. Noam Chomsky. 1970. Remarks on nominalization. In R. Jacobs and P. Rosenbaum, editors, Reading in English Transformational Grammar. Waltham: Ginn. find
4. Greville G. Corbett. 2001. Agreement: Terms and boundaries. In SMG conference papers. find
5. Gabi Danon. 2008. Definiteness spreading in the hebrew construct-state. Lingua, 118(7):872–906. find
6. Edit Doron. 1986. The pronominal “copula” as agreement clitic. Syntax and Semantics, (19):313–332. find
7. Gerald Gazdar, Ewan Klein, Geoffrey K. Pullum, and Ivan A. Sag. 1985. Generalised phrase structure grammar. Blackwell, Oxford, England. find
8. Yoav Goldberg and Michael Elhadad. 2009. Hebrew dependency parsing: Initial results. In Proceedings of IWPT. find our card
9. Yoav Goldberg and Reut Tsarfaty. 2008. A single framework for joint morphological segmentation and syntactic parsing. In Proceedings of ACL. find
10. Yoav Goldberg, Meni Adler, and Michael Elhadad. 2006. Noun phrase chunking in hebrew: Influence of lexical and morphological features. In Proceedings of COLING-ACL. find our card
11. Nomie Guthmann, Yuval Krymolowski, Adi Milea, and Yoad Winter. 2009. Automatic annotation of morphosyntactic dependencies in a Modern Hebrew treebank. In Frank Van Eynde, Anette Frank, Koenraad De Smedt, and Gertjan van Noord, editors, Proceedings of TLT. find
12. Kenneth L. Hale. 1983. Warlpiri and the grammar of non-configurational languages. Natural Language and Linguistic Theory, 1(1). find
13. Julia Hockenmaier and Mark Steedman. 2003. Parsing with generative models of predicate-argument structure. In Proceedings of ACL. find
14. Dan Klein and Christopher D. Manning. 2003. Accurate unlexicalized parsing. In Proceedings of ACL. find our card
15. Mohamed Maamouri, Ann Bies, and Seth Kulick. 2008. Enhanced annotation and parsing of the arabic treebank. In Proceedings of INFOS. find
16. Yusuke Miyao and Jun’ichi Tsujii. 2008. Feature-forest models for probabilistic hpsg parsing. Computational Linguistics, 34(1):35–80. find our card
17. Lilja Øvrelid and Joakim Nivre. 2007. Swedish dependency parsing with rich linguistic features. In Proceeding of RANLP. find
18. Slav Petrov, Leon Barrett, Romain Thibaux, and Dan Klein. 2006. Learning accurate, compact, and interpretable tree annotation. In Proceedings of ACL. find our card
19. Reut Tsarfaty and Khalil Sima’an. 2008. Relational-realizational parsing. In Proceedings of CoLing. find
20. Reut Tsarfaty, Khalil Sima’an, and Remko Scha. 2009. An alternative to head-driven approaches for parsing a (relatively) free word order language. In Proceedings of EMNLP. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/aclbulk-2010-tsarfaty-modeling-morphosyntactic-agreement-constituency-based/