Word-Based or Morpheme-Based? Annotation Strategies for Modern Hebrew Clitics

What it is about

The Modern Hebrew Treebank splits pronominal clitics on prepositions, accusative markers and possessive markers into separate leaves: one orthographic word becomes two or more nodes (morpheme-based, MB). The authors propose a word-based (WB) alternative. The clitic becomes inflectional features on a special class of preposition or marker, and the dropped pronoun is a trace. They convert the 5000-sentence treebank (test on the first 500 non-empty sentences) and train a PCFG grammar on each version, then parse with Bitpar. On bare prepositions the two strategies are equal. On the cliticised forms, the WB analysis gives the same parse in 116 of 131 occurrences. In the rest it is correct where MB is wrong in 5 cases and closer to the gold tree in 8 more, and it is never worse. The gain comes from the parser no longer deciding high attachment for prepositions that exist only in the surface cliticised form.

Ideas

• I1. Represent the clitic as features on a specialised preposition or marker, with an empty-node trace for the pronoun, so the tree yield matches the surface word. Their trace device for coordinated noun phrases (figure 1c) is the part that needs care. Note that UD uses multiword tokens for this, so this is a design choice to measure, not a fix.
• I2. Use their evaluation design on Mova's own data: convert the same gold trees both ways, parse with the same model, and count per sentence whether the cliticised cases come out identical, better or worse. Report the three counts, not one averaged score, because the converted sentences are shorter.
• I3. Encode the "light" preposition as a lexical feature: a preposition fused with a pronoun takes only a pronoun, cannot attach to a complex noun phrase, and cannot scope over a conjoined bare noun phrase (their examples 6 and 7). This is a constraint the attachment rules check.

For Mova

I1 is a representation choice for 11. Modern Hebrew in Universal Dependencies and 12. Mova and Modern Hebrew: where it stands, what it would need. I2 is a measurement for any Hebrew treebank conversion under treebanks/he/. I3 goes into Prepositions, prefixes, conjunctions and particles with the pronominal clitic facts.

Cite

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

@inproceedings{tsarfaty2008word,
  title = {{Word-Based or Morpheme-Based? Annotation Strategies for Modern Hebrew Clitics}},
  author = {Tsarfaty, Reut and Goldberg, Yoav},
  booktitle = {Proceedings of the Sixth International Conference on Language Resources and Evaluation (LREC'08)},
  year = {2008},
  url = {https://aclanthology.org/L08-1536/}
}

References

27 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. M. Adler and M. Elhadad. 2006. An Unsupervised Morpheme-Based HMM for Hebrew Morphological Disambiguation. In Proceedings of COLING-ACL 2006. find our card
2. R. Bar-Haim, K. Sima’an, and Y. Winter. 2005. Choosing an Optimal Architecture for Segmentation and POS-Tagging of Modern Hebrew. In ACL Workshop on Computational Approaches to Semitic Languages. find our card
3. D. Bikel and D. Chiang. 2000. Two Statistical Parsing Models Applied to the Chinese Treebank. In Second Chinese Language Processing Workshop, Hong Kong. find
4. J. P. Blevins. 2006. Word-Based Morphology. Journal of Linguistics, 3(42):531–573. find
5. L. Bloomfield. 1933. Language. University of Chicago Press. find
6. Rens Bod, Remko Scha, and Khalil Sima’an, editors. 2003. Data-Oriented Parsing. CSLI Publications. find
7. Eugine Charniak and Mark Johnson. 2005. Coarse-to-fine N-Best Parsing and Maxent Discriminative Reranking. In Proceedings of ACL. find our card
8. E. Charniak. 1997. Statistical Parsing with a Context-Free Grammar and Word Statistics. In AAAI/IAAI, pages 598–603. find
9. S. Cohen and N. Smith. 2007. Joint Morphological and Syntactic Disambiguation. In Proceedings of EMNLP. find
10. M. Collins. 2003. Head-Driven Statistical Models for Natural Language Parsing. Computational Linguistics. find our card
11. A. Dubey and F. Keller. 2003. Probabilistic Parsing for German using Sister-Head Dependencies. In Proceedings of ACL. find
12. Y. Goldberg and M. Elhadad. 2007. Toward Better Understanding of Hebrew NP Chunks. In ISCOL/BIFSAI 2007. find
13. Y. Goldberg and R. Tsarfaty. 2008. A Single Generative Probabilistic Model for Joint Morphological Segmentation and Syntactic Parsing. In Proceedings of ACL. find
14. C. F. Hocket. 1954. Two Models of Grammatical Description. Word, (10). find
15. M. Johnson. 1998. PCFG Models of Linguistic Tree Representations. Computational Linguistics, 24(4):613– 632. find
16. D. Klein and C. Manning. 2003. Accurate Unlexicalized Parsing. In Proceedings of ACL, pages 423–430. find our card
17. R. Levy and C. Manning. 2004. Deep Dependencies from Context-Free Statistical Parsers: Correcting the Surface Dependency Approximation. In Proceedings of ACL. find
18. M. Maamouri, A. Bies, T. Buckwalter, and W. Mekki. 2004. The Penn Arabic Treebank: Building a Large-Scale Annotated Arabic Corpus. find
19. M. Marcus, G. Kim, M. Marcinkiewicz, R. MacIntyre, A. Bies, M. Ferguson, K. Katz, and B. Schasberger. 1994. The Penn Treebank: Annotating Predicate-Argument Structure. find our card
20. P. H. Matthews. 1991. Morphology. Cambridge University Press. find
21. D. Ravid. 2006. Word-level Morphology: A Psycholinguistic Perspective on Linear Formation in Hebrew Nominals. Morphology, 16(1):127–148. find
22. H. Schmid. 2004. Efficient Parsing of Highly Ambiguous Context-Free Grammars with Bit Vectors. In Proceedings of ACL. find
23. H. Schmid. 2006. Trace Prediction and Recovery with Unlexicalized PCFGs and Slash features. In Proceedings of ACL. find
24. K. Sima’an, A. Itai, Y. Winter, A. Altman, and N. Nativ. 2001. Building a Tree-Bank of Modern Hebrew Text. In Traitement Automatique des Langues. find
25. R. Tsarfaty and K. Sima’an. 2007. Three-Dimensional Parameterization for Parsing Morphologically Rich Languages. In Proceedings of IWPT. find
26. R. Tsarfaty. 2006. Integrated Morphological and Syntactic Disambiguation for Modern Hebrew. In Proceeding of SRW COLING-ACL. find our card
27. A. Zwicky. 1977. On Clitics. Indiana University Linguistics Club. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/lingbulk-2008-tsarfaty-word-based-or-morpheme-based/