Developing an Arabic Treebank: Methods, Guidelines, Procedures, and Tools

What it is about

The authors report two and a half years of building the Penn Arabic Treebank for unvocalized MSA news text. The pipeline is: a Buckwalter-style analyzer proposes every possible analysis of a word, an annotator chooses one and supplies the case ending, a parser (Bikel's engine) pre-parses the sentence, and a treebank annotator corrects it. The analyzer covers 99.24% of the word tokens of the second release and 98.87% of the An-Nahar tokens. On a quality test, five annotators tagged 853 words: all five agreed on 85% of them, and each pair agreed on at least 92.2%. The Arabic pre-parser reaches 75.6 recall and 77.4 precision on sentences up to 40 words, against about 90 for English. Vocalization moved from the lexical level (first corpus) to case endings at POS level (second) and to full vocalization (third). Their main finding is that the choice of where to put vocalization decides speed and consistency.

Ideas

• I1. Keep the candidate list. The analyzer returns every segmentation and vocalization of a word, and the choice is made later. For tools/hbo, the segmenter should output all splits with their niqqud and let gparse or a rule pick one, rather than emitting a single split.
• I2. Use their reasons for "no solution" as the error log of the lexicon: adjective-versus-noun confusion, typos, passive forms, dialectal forms, grammar problems, names, numerals. tools/hbo logs every unparsed word under one of these classes; the counts per class are the to-do list for the lexicon, and the dialectal class is a route to the Mishnaic segment in tools/langseg.
• I3. A quality protocol: two or more readers tag the same sample of about 800 words independently, and the full and pairwise agreement are reported before a lexicon is called gold.

For Mova

I1 goes into tools/hbo and 13. Biblical Hebrew in Mova: the tools. I2 is a log for tools/hbo and a checklist for The lexicon and 11. Mishnaic Hebrew (hbo-x-mishnaic). I3 is the acceptance test for the Hebrew lexicon in The lexicon.

Topics: Treebanks

Cite

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

@inproceedings{maamouri2004developing,
  title = {{Developing an Arabic Treebank: Methods, Guidelines, Procedures, and Tools}},
  author = {Maamouri, Mohamed and Bies, Ann},
  booktitle = {Proceedings of the Workshop on Computational Approaches to Arabic Script-based Languages},
  year = {2004},
  url = {https://aclanthology.org/W04-1602/}
}

References

8 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Elsaid Badawi, M. G. Carter and Adrian Gully, 2004. Modern Written Arabic: A Comprehensive Grammar. Routledge: New York. find
2. Daniel M. Bikel, 2002. Design of a multi-lingual, parallel-processing statistical parsing engine. Proceedings of the Human Language Technology Workshop. find
3. Bracketing Guidelines for Treebank II Style, 1995. Eds: Ann Bies, Mark Ferguson, Karen Katz, Robert MacIntyre, Penn Treebank Project, University of Pennsylvania, CIS Technical Report MS-CIS-95-06. find
4. Mohamed Maamouri and Christopher Cieri, 2002. Resources for Arabic Natural Language Processing at the Linguistic Data Consortium. Proceedings of the International Symposium on Processing of Arabic. Faculté des Lettres, University of Manouba, Tunisia. find
5. M. Marcus, G. Kim, M. Marcinkiewicz, R. MacIntyre, A. Bies, M. Ferguson, K. Katz & B. Schasberger, 1994. The Penn Treebank: Annotating predicate argument structure. Proceedings of the Human Language Technology Workshop, San Francisco. find our card
6. M. Marcus, B. Santorini and M.A. Marcinkiewicz, 1993. Building a large annotated corpus of English: the Penn Treebank. Computational Linguistics. find our card
7. Mohamed A. Mohamed, 2000. Word Order, Agreement and Pronominalization in Standard and Palestinian Arabic. CILT 181. John Benjamins: Philadelphia. find
8. Zdenek Žabokrtský and Otakar Smrž, 2003. Arabic Syntactic Trees: from Constituency to Dependency. EACL 2003 Conferenceompanion. Association for Computational Linguistics, Hungary. find our card

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/bank-2004-maamouri-developing-arabic-treebank-methods-guidelines/