Parsing Arabic Dialects
What it is about
Levantine Arabic (LA) is spoken, unwritten, and has almost no treebank, while MSA has the Penn Arabic Treebank (about 17,600 training sentences). The authors parse transcribed LA without any LA training treebank and without a parallel LA–MSA corpus, using only a small LA–MSA lexicon and explicit knowledge of how the two differ. They try three routes: translate the LA sentence into MSA and parse it (sentence transduction); rewrite the MSA treebank into an LA-like treebank and train on it (treebank transduction); and rewrite the MSA grammar of a synchronous tree-substitution grammar into an LA grammar (grammar transduction). On the LA test set, grammar transduction is best: labelled F1 rises from 53.5 to 60.2 without POS tags (+6.7) and from 60.2 to 67.1 with gold tags (+6.9). The negation rule helps; the verb-subject-object reordering and the demonstrative rule do not; the morphological rules do not help. The paper has no treebank of LA beyond a 33,000-word development and test set.
Ideas
- I1. Grammar transduction applied to a sister variety: keep the parent-language trees and add a small set of explicit rewrites (here: negation as a postverbal particle, and a lexically specific construction for "want"). For Mishnaic Hebrew, the same kind of rewrite on the Biblical PTNK trees (negative particles, a "want/need" construction) would give a first Mishnaic treebank without new annotation; the rules would be a table in
tools/macula. - I2. Ablation by rule: each transformation is tested alone and in combination on the development set, and only those that help are kept. The paper shows which rules hurt (reordering, demonstrative switching, morphology) even though they were linguistically motivated. Use the same discipline for
languages/arc-x-targum/andlanguages/tmr/before any rewrite enters the release. - I3. Start the lexicon small: the 321-entry list of closed-class words and frequent items gave most of the gain, and the 1,560-entry extension added ambiguity that lowered F1 with gold tags. For a Biblical-to-Mishnaic (or Aramaic) pair table, measure each addition on held-out verses before keeping it.
For Mova
I1 belongs in 11. Mishnaic Hebrew (hbo-x-mishnaic) as a plan for a Mishnaic treebank built from PTNK; the negative particles of I1 go into Pronouns, particles and prepositions. I2 is a rule for the transduction step itself, to be written into 11. Mishnaic Hebrew (hbo-x-mishnaic). I3 is a measurement for The lexicon.
Cite
@inproceedings{chiang2006parsing,
title = {{Parsing Arabic Dialects}},
author = {Chiang, David and Diab, Mona and Habash, Nizar and Rambow, Owen and Shareef, Safiullah},
booktitle = {11th Conference of the European Chapter of the Association for Computational Linguistics},
year = {2006},
doi = {10.7916/d85t3tzg},
url = {https://doi.org/10.7916/d85t3tzg}
}References
- Daniel M. Bikel. 2002. Design of a multi-lingual, parallel-processing statistical parsing engine. In Proceedings of International Conference on Human Language Technology Research (HLT). find
- John Chen. 2001. Towards Efficient Statistical Parsing Using Lexicalized Grammatical Information. Ph.D. thesis, University of Delaware. find
- David Chiang. 2000. Statistical parsing with an automatically-extracted tree adjoining grammar. In 38th Meeting of the Association for Computational Linguistics (ACL’00), pages 456–463, Hong Kong, China. find
- Mona Diab, Kadri Hacioglu, and Daniel Jurafsky. 2004. Automatic tagging of arabic text: From raw text to base phrase chunks. In 5th Meeting of the North American Chapter of the Association for Computational Linguistics/Human Language Technologies Conference (HLT-NAACL04), Boston, MA. find
- David Graff. 2003. Arabic Gigaword, LDC Catalog No.: LDC2003T12. Linguistic Data Consortium, University of Pennsylvania. find
- Nizar Habash. 2004. Large scale lexeme based arabic morphological generation. In Proceedings of Traitement Automatique du Langage Naturel (TALN-04). Fez, Morocco. find
- Jan Hajic, Jan Hric, and Vladislav Kubon. 2000. Machine Translation of very close languages. In 6th Applied Natural Language Processing Conference (ANLP’2000), pages 7–12, Seattle. find
- Rebecca Hwa, Philip Resnik, Amy Weinberg, Clara Cabezas, and Okan Kolak. 2004. Bootstrapping parsers via syntactic projection across parallel texts. Natural Language Engineering. find
- Mohamed Maamouri, Ann Bies, and Tim Buckwalter. 2004. The Penn Arabic Treebank: Building a large-scale annotated Arabic corpus. In NEMLAR Conference on Arabic Language Resources and Tools, Cairo, Egypt. find
- Mohamed Maamouri, Ann Bies, Tim Buckwalter, Mona Diab, Nizar Habash, Owen Rambow, and Dalila Tabessi. 2006. Developing and using a pilot dialectal Arabic treebank. In Proceedings of the Fifth International Conference on Language Resources and Evaluation, LREC’06, page to appear, Genoa, Italy. find
- Owen Rambow, K. Vijay-Shanker, and David Weir. 2001. D-Tree Substitution Grammars. Computational Linguistics, 27(1). find
- Owen Rambow, David Chiang, Mona Diab, Nizar Habash, Rebecca Hwa, Khalil Sima’an, Vincent Lacey, Roger Levy, Carol Nichols, and Safi ullah Shareef. 2005. Parsing arabic dialects. find
- Final Report, 2005 JHU Summer Workshop. find
- Yves Schabes. 1990. Mathematical and Computational Aspects of Lexicalized Grammars. Ph.D. thesis, Department of Computer and Information Science, University of Pennsylvania. find
- Khalil Sima’an. 2000. Tree-gram parsing: Lexical dependencies and structural relations. In Proceedings of 38th Annual Meeting of the Association for Computational Linguistics (ACL’00), Hong Kong, China. find
- David A. Smith and Noah A. Smith. 2004. Bilingual parsing with factored estimation: Using english to parse korean. In Proceedings of the 2004 Conference on Empirical Methods in Natural Language Processing (EMNLP04). find
- Fei Xia, Martha Palmer, and Aravind Joshi. 2000. A uniform method of grammar extraction and its applications. In Proc. of the EMNLP 2000, Hong Kong. find
This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/aclbulk-2006-chiang-parsing-arabic-dialects/