Elamur Research⌕
☰
home / papers / 2003 / Towards semantic composition of Arabic: a \ensuremathłambda-DRT based approach

Towards semantic composition of Arabic: a \ensuremathłambda-DRT based approach

Paper · 2003 · Bassam Haddad and Mustafa Yaseen

Summary

The paper argues that Arabic NLP has focused on morphology and syntax and lacks a compositional semantic layer. It sketches one. The authors' earlier λ-calculus model treated Arabic determiners (the definite article, "all", "some") as generalized quantifiers, formalised in their AGQL language. Such Montague-style logical forms work for single sentences but cannot handle discourse phenomena such as anaphora. Discourse Representation Theory (DRT) does handle them, but its usual construction is top-down, depends on processing order, and is not compositional.

The proposed remedy is λ-DRT, following Bos et al. (1994). DRSs are lambda-abstracted and joined with a merge operator. λ-conversion is simulated by unifying feature structures that carry LAMBDA and DRS attributes inside an HDPSG-style grammar. Worked examples cover nested definite quantifiers, a two-clause discourse whose verb-incorporated pronoun links back to earlier referents, and the sentence "each student studies-hard".

There is no implementation or evaluation. The contribution is a design proposal meant to steer Arabic computational linguistics toward deep, unification-based syntax–semantics interfaces.

Key points

  • A quantifier's meaning is a two-place λ-term over a restriction R and a scope S. For "the boy studies the Arabic", composition yields two nested The1 quantifiers over boy(x) and Arabic(y), with study(x,y) as the scope.
  • A DRS is a pair of discourse referents and conditions. The merge ⊗ takes the union of the referent sets and of the condition sets separately.
  • In "Ayman studies a language, he-likes-it", the DRS gains an event e, a speech time n and a state s. The pronoun variables are equated with earlier referents (z = x, w = y), even though the pronouns are morphologically incorporated into the verb.
  • Lexical DRSs carry sorts (Individual, Event, Any) and thematic roles such as agent. Simple temporal-anaphora rules are added on top of the Bos et al. approach.
  • In the feature-structure encoding, LAMBDA lists the arguments and DRS holds the body. Unifying LAMBDA values with argument structures plays the role of λ-conversion. Syntactic features are omitted for simplicity, and no quantitative results are reported.

Interesting ideas

  • I1. Beta-reduction can be replaced by unification: a lexical entry exposes argument slots, and composition fills them by unifying with sister constituents. This keeps semantic construction declarative and traceable.
  • I2. Definites are pointers to existing referents, indefinites introduce fresh referents, and pronouns are labels resolved to earlier referents rather than locally bound variables. This gives a clean, explainable policy for when an entity is created.
  • I3. Events and the speech time are first-class referents, linked by explicit inclusion conditions (event inside speech-time interval, event inside state). Tense and ordering then become ordinary conditions rather than tags.
  • I4. Sorted referents (Individual vs Event) with roles attached at lexical entry time act as built-in type checks during composition.

For Mova

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

  • I1 (frames): verb and determiner templates in prag could expose slots that are unified with UD dependents, so each frame records exactly which tree node filled which role.
  • I2 (world): the world builder could create entities only on indefinites and proper names, and require definites and pronouns to link to an existing entity or flag a failure.
  • I3 (world): event ordering and state holdings could be stored as explicit temporal conditions between event, state and utterance-time referents, which makes "when/before" questions answerable by lookup.
  • I4 (frames): VerbNet role fillers could be checked against sort constraints during frame building, rejecting parses that put an event where an agent individual is required.

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 · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

🧭 Topics 1Research subjects and languages the work is about.Semitic languages · 3,066 works

Cite

Cite as Bassam Haddad and Mustafa Yaseen. 2003. Towards semantic composition of Arabic: a \ensuremathłambda-DRT based approach. Workshop on Machine Translation for Semitic languages: issues and approaches.

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

@inproceedings{haddad2003semantic,
  title = {{Towards semantic composition of Arabic: a ensuremathłambda-DRT based approach}},
  author = {Bassam Haddad and Mustafa Yaseen},
  booktitle = {Workshop on Machine Translation for Semitic languages: issues and approaches},
  year = {2003},
  url = {https://aclanthology.org/2003.mtsummit-semit.8/}
}

References

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

  1. Al-Fedaghi and Al-Anzi. 1989 A New Algorithm to Generate Arabic Root-Pattern Forms. Proceedings of the 11th National Computer Conference, Saudi Arabia, 1989, pp. 391-400. find
  2. Badr AL-Johar and Jim McGregor. 1997. A Logical Meaning Representation for Arabic (LMRA). Proceedings of the 15th National Computer Conference, Riyadh, Saudi Arabia, 1997, pp. 31-40. find
  3. Husni Al-Muhtaseb and Chris Mellish. 1997. Towards an Arabic Upper Model: A proposal. Proceeding of the 15 th National Conference, Riyadh, Saudi Arabia 1997. find
  4. Beesley. 2001. Finite-State Morphological Analysis and Generation of Arabic at Xerox Research: Status and Plans 2001. ACL/EACL01: Conference of the European Chapter, Workshop: Arabic Language Processing: Status and Prospects. find
  5. Bende-Farkas and Kamp. 2001. Indefinites and Binding: From Specificity to Incorporation, 13 th European Summer School In Logic, Language and Information, ESSLLI, 2001. find
  6. Johan Bos, E. Mastenbroek, S. McGlashan, S. Millies and M. Pinkal. 1994. A Compositional DRS-based Formalism for NLP Applications. Report 59, VerbMobil, Universität des Saarlandes. ↗
  7. Everhard Ditters. 2001. A Formal Grammar for the Description of Sentences Structures in Modern Standard Arabic. ACL/EACL01: Conference of the European Chapter, Workshop: Arabic Language Processing: Status and Prospects. find
  8. El-Dessouki, Nazif, Ahmad. 1988 1 An Expert System for Understanding Arabic Sentences. Proceeding of the 10 th National Computer Conference, Jeddah, Saudi Arabia, 1988, pp 745-759. find
  9. Haddad and Yaseen. 2001. Towards Understanding Arabic: A Logical Approach for Semantic Representation. ACL/EACL01: Conference of the European Chapter, Workshop: Arabic Language Processing: Status and Prospects. find
  10. Haddad. 2002. AGQL: An Arabic generalized Quantifiers Language. Internal Artificial Intelligence Projects at F.I.T, Amman University. find
  11. Hans Kamp. 1981. A Theory of Truth and Semantics Representation. In J. Groendijek, T. J. Stokhof, eds., Formal Methods in Study of Languages. Mathematish Centrum, Amsterdam. find
  12. Hans Kamp and Uwe Reyle. 1993. From Discourse to Logic. Kluwer Academic Publishers. ↗
  13. Karsten Konard, H. Maier, M. Pinkal. 1996. CLEARS Ein Werkzeug für Ausbildung und Forschung in Computerlinguistik. In Natural Language Processing and Speech Technology, 3 rd KONVENS Conference, Bielefeld. find
  14. Mouton de Gruyter Ingrid Fischer. 1993 Die Kompositionale Bildung von Diskursrepräsentationsstrukturen über einer Chart. Masters thesis, University of Erlangen, Germany Richard Montague. 1973. The Proper Treatment of Quantification in Ordinary English. In: Philosophy, Language , and Artificial Intelligence, ed., J. Kulas, J. H. Fetzer and T. Rankin, Kluwer Academic Publishers, 1988. find
  15. Riadh Ouersighni. 2001. A major offshoot of the Dinar-MBC project: AraParse, a morphosyntactic analyzer for unvowelled Arabic texts. ACL/EACL01: Conference of the European Chapter, Workshop: Arabic Language Processing: Status and Prospects. find
  16. Carl Pollard and Ivan Sag. 1994. Head-Driven Phrase Structure Grammar. University of Chicago Press, Chicago. find
  17. John Nerbonne. 1993. A feature-based Syntax/ semantics Interface. In Annals of Mathematics and Artificial Intelligence 8(1993), J.C. Baltzer AG, Science Publishers. ↗
  18. Has Uszkoreit, Flickinger and Ivan Sag 2000. Deep Linguistic Analysis with HDPSG. In Verbmobile: Foundation of Speech-to-Speech Translation, Ed. W. Wahlster, Springer. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/haddad-2003-semantic-composition-arabic-ensuremath-lambda/