Elamur Research⌕
☰
home / papers / 2011 / Construction of an HPSG Grammar for the Arabic Relative Sentences

Construction of an HPSG Grammar for the Arabic Relative Sentences

Paper · 2011 · Ines Zalila and Kais Haddar

Summary

Arabic relative constructions are hard to parse. They come in many shapes, they can nest inside one another, and they interact with ellipsis and coordination. The authors note that HPSG-based Arabic work had largely ignored them. Drawing on traditional Arabic grammar and consultations with linguists, they build a typology of relatives along two axes:

  • Conjunctive noun type: nominal (common or special) versus prepositional (acting on verbs or on nouns).
  • Following clause: a VP or PP versus an NP.

They also separate explanatory relatives, which have an antecedent, from completive ones, which have none. On this basis they extend an existing Arabic adaptation of HPSG with a type hierarchy and features for conjunctive nouns. The grammar is encoded in TDL and compiled into a parser on the LKB platform. On a 500-sentence test corpus rich in relatives, 84% of sentences receive exactly one analysis, 14% receive several, and 2% fail. The work shows that a deep, explainable grammar for a complex Arabic phenomenon can be built with standard grammar-engineering tools rather than a custom parser.

Key points

  • The grammar is written in TDL (Type Description Language), and the parser is generated by LKB (Linguistic Knowledge Building).
  • Seven TDL files make up the grammar (roots, noeuds, rsynt, type-rules, type-lex, type, and lexicon), while LKB's setup relies on five LISP files, script.lsp among them.
  • The lexicon has about 3000 entries (~2500 verbs, 450 nouns, 50 particles), drawn mostly from the test corpus.
  • Evaluation used 500 sentences that also involve ellipsis, vocatives and description. Results were 2% with no derivation tree, 84% with a single tree and 14% with two or more trees.
  • Residual ambiguity arises mainly from attachment: a relative may modify either the head noun or a whole annexed (construct-state) noun group. A Java interface lets users add words to lexique.tdl without knowing TDL, and an LMF→TDL translator can help grow the lexicon.

Interesting ideas

  • I1. The relative typology is keyed on two factors: the introducing word's class and the category of its complement (VP/PP versus NP). This turns a fuzzy phenomenon into a small, checkable table of licensed combinations.
  • I2. The authors split relatives into those that describe an antecedent and those that fill an argument slot themselves (headless or free relatives). This captures a real semantic difference, not only a syntactic one.
  • I3. The relativizer is treated as coreferent with its antecedent and must agree with it in gender and number. Agreement thus doubles as a linking constraint.
  • I4. Parses are reported as "no tree / exactly one / several", so overgeneration gets measured separately from coverage failure.
  • I5. A small tool lets non-experts add lexicon entries without touching the formal grammar syntax. This removes a bottleneck that earlier Arabic projects also hit.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (tree): English relativizers (who/which/that/what/whoever) could get a similar lexicon-driven table of allowed complement shapes, used to validate or repair acl:relcl attachments.
  • I2 (frames): Free relatives like "what the father wrote" should become an argument entity in the frame, while antecedent relatives should add facts about an existing referent.
  • I3 (frames): Relative pronouns could be resolved as coreference links to their antecedent, with number and animacy checks (who vs which) used to rank candidate heads.
  • I4 (learn): Witness-parser agreement could be logged with the same three-way split, so that ambiguity rates and outright failures are tracked as separate signals.
  • I5 (tokens): A simple entry tool for the en crate's lexicon would let contributors add lemmas and morphology while the tool keeps the internal format consistent.

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, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 2Research subjects and languages the work is about.Semitic languages · 2,908 worksSyntax & grammar formalisms · 214 works

Cite

Cite as Ines Zalila and Kais Haddar. 2011. Construction of an HPSG Grammar for the Arabic Relative Sentences. Proceedings of the International Conference Recent Advances in Natural Language Processing 2011.

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

@inproceedings{zalila2011construction,
  title = {{Construction of an HPSG Grammar for the Arabic Relative Sentences}},
  author = {Ines Zalila and Kais Haddar},
  booktitle = {Proceedings of the International Conference Recent Advances in Natural Language Processing 2011},
  year = {2011},
  url = {https://aclanthology.org/R11-1081/}
}

References

19 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Abdelkader A., Haddar K. and Ben Hamadou A., « Etude et analyse de la phrase nominale arabe en HPSG », Traitement Automatique des Langue Naturelles, Louvain, UCL Presses de Louvain: 379-388, 2006. find
  2. Abdelwahed A., « „alkalima fy „attourath „allisaany „alaraby , ًتسعنا ًواسهنا زاسرنا ًف حمهكنا », Librairie Aladin 1ère édition, Sfax – Tunisie : 1-100, 2004. find
  3. Alnajem S. and Alzhouri F., “An HPSG Approach to Arabic Nominal Sentences”, Journal of the American society for information Scienceand Technology: 422 – 434, 2008. find
  4. Aloulou C., « Analyse syntaxique de l‟Arabe: Le système MASPAR », RECITAL, NantesFrance, 2003. find
  5. Bahou Y., Hadrich Belguith L., Aloulou C. and Ben Hamedou A.. «Adaptation and implementation of HPSG grammars to parse non-voweled Arabic texts », memory of Master, Faculty of Economics and Management of Sfax. Belkacemi C., «The relative marker: a definite marker substitute? », ArOr Archiv Orientální, 66/2, 142-148, Based on Arabic dialects, 1998. find
  6. Blache P., «Les Grammaires de Propriétés: des contraintes pour le traitement automatique des langues naturelles». Hermès Sciences, Paris, 2001. find
  7. Boukedi S., Haddar K. and Abdelwahed A., « Vers une analyse des phrases arabes en HPSG et LKB ». GEI 2008, 8ème Journées Scientifiques des Jeunes Chercheurs en Génie Electrique et Informatique, Sousse, Tunisie : 487- 498, 2008. find
  8. Copestake A., « Implementing Typed Feature Structure Grammars ». CSLI Publications, Stanford University, 2002. find
  9. Dahdah A.. « خاحىن و لوادج ًف حٍتسعنا حغهنا دعاىق مجعم », Librairie de Nachirun Lebanon, 5 ème edition, 1992. find
  10. Elleuch S., « Analyse syntaxique de la langue arabe basée sur le formalisme d‟unification HPSG ». Mémoire de DEA en Système d‟information et Nouvelles Technologies, Tunisie : 55-88, 2004. find
  11. Fehri H., Loukil N., Haddar K. and Ben Hamadou A., “Un système de projection du HPSG arabisé vers la plate-forme LMF ». JETALA, Maroc, 1-11, 2006. find
  12. Garcia O., « Une introduction à l‟implémentation des relatives de l‟espagnol en HPSG–LKB », Mémoire de recherche, 2005. find
  13. Haddar K. and Ben Hamadou A., « Un système de recouvrement des ellipses de la langue arabe ». Proceedings of VEXTAL, San Servolo V.I.U. 22(11) : 159-167, 1999. find
  14. Krieger H. and Schäfer U., « TDL: A Type Description Language for HPSG». Part 1 and Part 2, Research Report, RR-94-37, 1994. find
  15. Loukam M. and Laskri M., « Vers la modélisation de la grammaire de l‟arabe standard basée sur le formalisme HPSG », Actes JED‟2007, Journées de l‟Ecole Doctorale, 27(5), Annaba/Algérie, 2007. find
  16. Maaloul H., Haddar K. and Ben Hamadou A., «La coordination arabe : étude et analyse en HPSG », MCSEAI 2004, 8ème conférence maghrébine sur le GL et l‟IA, Sousse, Tunisie : 487- 498, 2004. find
  17. Meurers W. D., «A Web-based Instructional Platform for Constraint-Based Grammar Formalisms and Parsing». In Dragomir Radev and Chris Brew (eds.), Effective Tools and Methodologies for Teaching NLP and CL, New Brunswick, NJ: The Association for Computational Linguistics: 18 – 25, 2002. find
  18. Pollard C. and Sag I., «Head-drive phrase structure grammars», CSLI series, Chicago University Press, 1994. find
  19. Tseng J., « Implémentation HPSG avec LKB: La Matrix et la Grenouille », Séminaire HPSG-UFRL, Paris 7, 14(12), 2006 find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zalila-2011-construction-hpsg-grammar-arabic-relative/