CamelParser: A system for Arabic Syntactic Analysis and Morphological Disambiguation

What it is about

CamelParser is a public system that builds Arabic dependency trees (in the CATiB style) with the same morphological features that MADAMIRA produces, and then uses the syntax to correct MADAMIRA's choice of case and state. The parser is MaltParser. Its feature set is tuned in two stages: MaltOptimizer picks the parsing algorithm and learner, then a greedy search adds morphological features one at a time and keeps each one only if labelled attachment improves. The parser is trained on gold morphology and tested on predicted morphology, which needed a modified MaltOptimizer that accepts separate train and test sets. The system outputs the tree, a TrEd file, a PDF tree picture and a traditional Arabic grammatical analysis in plain text. On the test set, labelled attachment rises from 81.6% to 83.8%, and full diacritization of words from 88.1% (MADAMIRA) to 90.8%.

Ideas

• I1. The greedy feature-set search as a tool for gparse: start with no morphological features, add each candidate in turn, keep it only when the labelled score on a development split rises, and write the chosen set to a file so the result is repeatable. Run it with sorted (BTreeSet) candidate lists so that the order is deterministic.
• I2. Train with gold morphology, test with the morphology that tools/hbo predicts. This is the honest setting for a pipeline from raw text, and it shows how much of the parser's score depends on perfect tagging.
• I3. A traditional grammatical analysis generated from the tree, as plain text: for each word, its role (subject, object of a preposition, head of an idafa) and its case, produced by rules. This is an explanation a learner can read and it is a direct form of the "say why" that Mova aims for.

For Mova

I1 is a training procedure for gparse used with tools/hbo, and the chosen feature sets are recorded in its run output. I2 is the test protocol for tools/hbo once it feeds gparse. I3 goes into Syntax: the clause and the sentence as an output format for the explanation of case and construct, and into prag as a rule-based explanation.

Cite

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

@inproceedings{shahrour2016camelparser,
  title = {{CamelParser: A system for Arabic Syntactic Analysis and Morphological Disambiguation}},
  author = {Shahrour, Anas and Khalifa, Salam and Taji, Dima and Habash, Nizar},
  booktitle = {Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: System Demonstrations},
  year = {2016},
  url = {https://aclanthology.org/C16-2048/}
}

References

15 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Sarah Alkuhlani, Nizar Habash, and Ryan Roth. 2013. Automatic morphological enrichment of a morphologically underspecified treebank. In HLT-NAACL, pages 460–470. find
2. Miguel Ballesteros and Joakim Nivre. 2012. Maltoptimizer: an optimization tool for maltparser. In Proceedings of the Demonstrations at the 13th Conference of the European Chapter of the Association for Computational Linguistics, pages 58–62. Association for Computational Linguistics. find
3. Tim Buckwalter. 2002. Buckwalter Arabic Morphological Analyzer Version 1.0. Linguistic Data Consortium, University of Pennsylvania, 2002. LDC Catalog No.: LDC2002L49. find
4. Mona Diab, Nizar Habash, Owen Rambow, and Ryan Roth. 2013. LDC Arabic treebanks and associated corpora: Data divisions manual. arXiv preprint arXiv:1309.5652. ↗
5. Nizar Habash and Ryan M Roth. 2009. CATiB: The Columbia Arabic Treebank. In Proceedings of the ACL-IJCNLP 2009 Conference Short Papers, pages 221–224. Association for Computational Linguistics. find our card
6. Nizar Habash, Ryan Gabbard, Owen Rambow, Seth Kulick, and Mitchell P Marcus. 2007a. Determining case in Arabic: Learning complex linguistic behavior requires complex linguistic features. In EMNLP-CoNLL, pages 1084–1092. find
7. Nizar Habash, Abdelhadi Soudi, and Tim Buckwalter. 2007b. On Arabic Transliteration. In A. van den Bosch and A. Soudi, editors, Arabic Computational Morphology: Knowledge-based and Empirical Methods. Springer. find
8. Nizar Habash. 2010. Introduction to Arabic Natural Language Processing. Morgan & Claypool Publishers. find
9. Seth Kulick, Ryan Gabbard, and Mitch Marcus. 2006. Parsing the Arabic Treebank: Analysis and Improvements. In Proceedings of the Treebanks and Linguistic Theories Conference, pages 31–42, Prague, Czech Republic. find
10. Mohamed Maamouri, Ann Bies, Tim Buckwalter, and Wigdan Mekki. 2004. The Penn Arabic Treebank: Building a Large-Scale Annotated Arabic Corpus. In NEMLAR Conference on Arabic Language Resources and Tools, pages 102–109, Cairo, Egypt. find
11. Yuval Marton, Nizar Habash, and Owen Rambow. 2013. Dependency parsing of modern standard Arabic with lexical and inflectional features. Computational Linguistics, 39(1):161–194. find our card
12. Joakim Nivre, Johan Hall, Jens Nilsson, Atanas Chanev, Gülsen Eryigit, Sandra Kübler, Svetoslav Marinov, and Erwin Marsi. 2007. MaltParser: a language-independent system for data-driven dependency parsing. Natural Language Engineering, 13(02):95–135. find
13. Petr Pajas. 2008. Tred: Tree editor. http://ufal.mff.cuni.cz/ pajas/tred. ↗
14. Arfath Pasha, Mohamed Al-Badrashiny, Ahmed El Kholy, Ramy Eskander, Mona Diab, Nizar Habash, Manoj Pooleery, Owen Rambow, and Ryan Roth. 2014. MADAMIRA: A Fast, Comprehensive Tool for Morphological Analysis and Disambiguation of Arabic. In In Proceedings of LREC, Reykjavik, Iceland. find our card
15. Anas Shahrour, Salam Khalifa, and Nizar Habash. 2015. Improving Arabic diacritization through syntactic analysis. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, EMNLP 2015, Lisbon, Portugal, September 17-21, 2015, pages 1309–1315. find our card

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/aclbulk-2016-shahrour-camelparser-system-arabic-syntactic-analysis/