Elamur Research⌕
☰
home / papers / 2016 / Explicit Fine grained Syntactic and Semantic Annotation of the Idafa Construction in Arabic

Explicit Fine grained Syntactic and Semantic Annotation of the Idafa Construction in Arabic

Summary

In Arabic, Idafa is a broad construction that joins two nominal elements into one syntactic unit. It covers what English expresses as noun-noun compounds, 's-genitives and of-genitives, and it also appears in quantifier, quasi-preposition and adjectival patterns. The LDC Arabic Treebank (ATB) contains these structures only implicitly, as (NP (NN NP)) or (NP (JJ NP)). As a result, "grammar book" and "all topics" receive the same analysis.

The authors propose a typology based on the actual part of speech of the first element. It has three top-level classes, False, Grammatical and True Idafa, which split further into 10 syntactic subclasses. True Idafa cases also receive semantic relation labels, chosen after reviewing existing noun-compound semantic schemes. They build an automatic procedure that adds these explicit labels to the ATB and compare its output with the IDF relation in the CATiB treebank. Accuracy is on par with CATiB, and the labels are considerably finer-grained. The resource matters for parsing, MT, MWE detection and information extraction, all of which depend on how nominal compounds compose meaning.

Key points

  • The typology has three main syntactic types: FIC (adjective-headed), GIC (function-word-like heads such as quantifiers and quasi-prepositions) and TIC (true noun-noun). Together they have 10 syntactic subtypes.
  • FIC has three variants: qualitative (the adjective agrees with the preceding noun), elative (always masculine singular, no agreement) and nominalized (the adjective takes noun positions). FIC accounts for 1.4% of ATB Idafa cases.
  • A pilot semantic annotation of 1,946 randomly sampled TIC instances from the ATB uses 26 semantic types, reported by frequency.
  • Prior work cited shows how frequent Idafa is:
    • an average of 3.06 Idafa per sentence, present in 84.3% of trees (Green & Manning);
    • over 75% of Idafa are single-level, with nesting up to 6 levels, and 40.9% are modified by adjectives (Gabbard & Kulick).
  • The automatic labeling is evaluated against CATiB's IDF label, one of its 8 relations, and reaches the same accuracy. The source data is Arabic Treebank Part 3 v3.2 (LDC2010T08).

Interesting ideas

  • I1. When one treebank bracket hides several constructions, the authors separate them by the real category of the head rather than adding new tree shapes. This keeps the original structure and still exposes distinctions that matter downstream. — For Mova (tree): Mova could post-classify English compound, nmod:poss and nmod "of" arcs by head type (quantifier "all of", partitive, true noun-noun) so that later stages can tell them apart.
  • I2. Annotation is layered: syntactic typing comes first, and semantic relations are assigned only to the class that genuinely links two nouns. This avoids forcing a meaning relation onto quantifier or adjective-headed cases. — For Mova (world): only true noun-noun compounds would get a relation such as location, purpose or part-whole when Mova builds entities, while quantifier constructions would go to counting logic instead.
  • I3. Agreement behaviour serves as a diagnostic for subtype: in Arabic, the qualitative adjective agrees with the preceding noun while the elative never does. Cheap morphological signals can therefore stand in for costly semantic judgments. — For Mova (tree): number agreement and plural marking on the first element (e.g. "sports car" vs "sales department") could feed rule-based features that help the parser decide between compound and modifier readings.
  • I4. A richer automatic labeling is checked against an independent coarse label from another treebank. This tests that adding granularity did not cost coverage or accuracy. — For Mova (learn): when Mova refines a UD relation into subtypes, it could collapse them back and score against the original gold arcs, using that as a negative control against regressions.
  • I5. Genitive and compound forms are compared across languages: Arabic Idafa, Persian, Urdu and Kurdish Ezafe, Hebrew smikhut, and English 's, of-phrases and compounds. A single underlying relation can surface in several ways, chosen by semantic type. — For Mova (back): the generator could pick between "X's Y", "Y of X" and "X Y" from the semantic relation stored in the meaning graph rather than copying the surface form.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Topics: Semitic languages

Institutions

🎓Dublin City UniversityDublin · Ireland48 works · 8 753 citations🎓George Washington UniversityWashington · United States16 works · 4 701 citations

Cite

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

@inproceedings{hawwari2016explicit,
  title = {{Explicit Fine grained Syntactic and Semantic Annotation of the Idafa Construction in Arabic}},
  author = {Hawwari, Abdelati and Attia, Mohammed and Ghoneim, Mahmoud and Diab, Mona},
  booktitle = {Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC'16)},
  year = {2016},
  doi = {10.63317/2mf9ys5gfpj9},
  url = {https://doi.org/10.63317/2mf9ys5gfpj9}
}

References

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

  1. Asghari, H., Maleki, J., & Faili, H. (2014, April). A Probabilistic Approach to Persian Ezafe Recognition. In EACL (pp. 138-142). find
  2. Barker, K., & Szpakowicz, S. (1998, August). Semi-automatic recognition of noun modifier relationships. In Proceedings of the 36th Annual Meeting of the Association for Computational Linguistics and 17th International Conference on Computational Linguistics-Volume 1 (pp. 96-102). Association for Computational Linguistics. find
  3. Bögel, T., & Butt, M. (2013). Possessive clitics and ezafe i n U r d u . Morphosyntactic Categories and the Expression of Possession, 199(291), 86-129. find
  4. Boujelben, I., Mesfar, S., & Hamadou, A. B. (2010). Arabic Compound Nouns processing: inflection and tokenization. In Proceedings of Nooj International Conference 2010, 40. find
  5. Downing, P. (1977). On the creation and use of English compound nouns.Language, 810-842. find
  6. Gabbard, R., & Kulick, S. (2008, June). Construct state modification in the Arabic treebank. In Proceedings of the 46th Annual Meeting of the Association for Computational Linguistics on Human Language Technologies: Short Papers (pp. 209-212). Association for Computational Linguistics. find our card
  7. Gadalla, H. A., & Abdel-Hamid, A. M. E. (2000). Genitive Constructions in English and Arabic: A Contrastive Study. In the Bulletin of the Faculty of Arts, Assiut University, Egypt. Vol. 6, Dec. 2000, pp. 1-64. find
  8. Girju, R., Moldovan, D., Tatu, M., & Antohe, D. (2005). On the semantics of noun compounds. Computer speech & language, 19(4), 479-496. find
  9. Girju, R. (2007, June). Improving the interpretation of noun phrases with cross-linguistic information. In the Proceedings of the 45 th Annual Meeting of the Association for Computational Linguistic (Vol. 45, No. 1, p. 568). find
  10. Green, S., & Manning, C. D. (2010, August). Better Arabic parsing: Baselines, evaluations, and analysis. In Proceedings of the 23rd International Conference on Computational Linguistics (pp. 394-402). Association for Computational Linguistics. find our card
  11. Habash, N., Gabbard, R., Rambow, O., Kulick, S., & Marcus, M. P. (2007, June). Determining Case in Arabic: Learning Complex Linguistic Behavior Requires Complex Linguistic Features. In EMNLP-CoNLL (pp. 1084-1092). find
  12. Habash, N., & Roth, R. M. (2009, August). Catib: The columbia arabic treebank. In Proceedings of the ACL-IJCNLP 2009 conference short papers (pp. 221-224). Association for Computational Linguistics. find our card
  13. Lauer, M. (1995). Designing statistical language learners: experiments on noun compounds. Macquarie University, Australia. find
  14. Lees, R. B. (1960). The Grammar of English Nominalizations. The Hague-Paris, Mouton de Gruyter. find
  15. Levi, J. N. (1978) The Syntax and Semantics of Complex Nominals. New York: Academic Press. find
  16. Lotfi, A. R. (2014). Persian Ezafe as a ‘figure’marker: A unified analysis.California Linguistic Notes, 39, 1. find
  17. Linh, N. M. (2010). Noun-noun Combinations in Technical English. A Master's Thesis, Suranaree University of Technology, Thailand Megerdoomian, K. (2000). A computational analysis of the Persian noun phrase. Memoranda in Computer and Cognitive Science MCCS-00-321, Computing Research Lab, New Mexico State University. find
  18. Maamouri, M., Bies, A., & Kulick, S. (2008, May). Enhancing the Arabic Treebank: a Collaborative Effort toward New Annotation Guidelines. In LREC. find
  19. Moldovan, D., Badulescu, A., Tatu, M., Antohe, D., & Girju, R. (2004, May). Models for the semantic classification of noun phrases. In Proceedings of the HLT-NAACL Workshop on Computational Lexical Semantics (pp. 60-67). Association for Computational Linguistics. find
  20. Samvelian, P. (2007). The Ezafe as a head-marking inflectional affix: Evidence from Persian and Kurmanji Kurdish. Aspects of Iranian Linguistics: Papers in Honor of Mohammad Reza Bateni, 339-361. find
  21. Shilon, R., Habash, N., Lavie, A., & Wintner, S. (2012). Machine translation between Hebrew and Arabic. Machine translation, 26(1-2), 177-195. find
  22. Strunk, J. (2003). The structure of the Kurdish noun phrase. unpublished ms. find
  23. Tratz, S., & Hovy, E. (2010, July). A taxonomy, dataset, and classifier for automatic noun compound interpretation. In Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics ( p p . 6 7 8 - 6 8 7 ) . A s s o c i a t i o n f o r Computational Linguistics. find
  24. Vanderwende, L. (1994, August). Algorithm for automatic interpretation of noun sequences. In P r o c e e d i n g s o f t h e 1 5 t h c o n f e r e n c e o n Computational linguistics-Volume 2 (pp. 782-788). Association for Computational Linguistics. find
  25. Warren, B. (1978). Semantic patterns of noun-noun compounds. Acta Universitatis Gothoburgensis. Gothenburg Studies in English Goteborg, 41, 1-266. 9. find
  26. Language Resource References Maamouri, M., Bies, A., Kulick, S., Krouna, S., Gaddeche, F. & Zaghouani, W. (2010). Arabic treebank: Part 3 v 3.2. Distributed by the Linguistic Data Consortium. LDC Catalog No.: LDC2010T08 . find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/hawwari-2016-explicit-fine-grained-syntactic-semantic/