Elamur Research⌕
☰
home / papers / 2012 / A prototype for projecting HPSG syntactic lexica towards LMF

A prototype for projecting HPSG syntactic lexica towards LMF

Summary

Arabic lexica written in HPSG are hard to compare or combine. Each one was built for its own task and uses its own feature names, granularity and file format. The authors want to measure the real coverage of these resources and eventually merge them. Their approach is to map each one into a common pivot model that follows LMF (ISO 24,613), uses data categories in the spirit of ISO 12,620, and is serialised as TEI-style XML.

They first analyse how HPSG must be adapted for Arabic, adding or redefining features such as CFORM, DENUDE, DIMINUTIVE, RELATIVE, NATURE, RADICAL and NFORM. For each lexical category they then identify a matching LMF structure. On this basis they build a rule-based projection system and implement it as an object-oriented prototype with a browsing interface.

Ten heterogeneous Arabic HPSG lexica were projected, giving one LMF lexicon of 3,000 verbs, 450 nouns and 50 particles. The authors also state which conditions the source lexica must meet for the conversion to lose nothing. The work matters because it shows a standards-based route to reusing lexical resources from different formalisms.

Key points

  • Target format: LMF (ISO 24,613), developed by ISO TC 37/SC 4, with TEI-based XML serialisation. The output follows the DTD in ISO/TC37/SC4 N130 rev.9 (2006).
  • Arabic-specific HPSG features:
    • CFORM: triliteral vs. quadriliteral root.
    • DENUDE: bare vs. augmented root.
    • DIMINUTIVE and RELATIVE.
    • NATURE: agent noun, patient noun, verbal adjective.
    • RADICAL: the root.
    • NFORM: redefined to describe how a noun inflects.
  • Evaluation: 10 Arabic HPSG lexica were merged into one LMF lexicon of 3,000 verbs, 450 nouns and 50 particles.
  • Three conditions for lossless projection:
    • Source lexica with inflected forms must include features linking those forms to their canonical or derived forms.
    • Every HPSG feature used must be present in the feature base.
    • Triliteral verb schemes must be rejected, so that homographs like kharaja and kharija do not clash.
  • Limitations of the prototype: it is extensible with new rules for other formalisms, but not portable, since it is tied to Arabic and to Windows.

Interesting ideas

  • I1. Pivot model with an isomorphic serialisation: an abstract meta-model plus standard data categories serves as the pivot, and any concrete syntax maps onto it one-to-one, so resources become comparable regardless of their original format.
  • I2. Rule-based projection over loose concept mapping: the authors prefer conversion by explicit rules that act like a parser, because it always produces well-formed target structures, whereas simple concept lexica risk ill-formed output.
  • I3. Explicit conditions for lossless conversion: the authors state when conversion loses information, namely when source features are missing from the target feature base.
  • I4. Separate entries for homographs: forms that are spelled or pronounced alike but belong to different inflectional paradigms get separate lexical entries rather than one merged lemma.
  • I5. Links from inflected forms to their base: dedicated features tie each inflected form to its canonical or derived form, and merging is designed to avoid redundant entries.

For Mova

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

  • I1 (world): WordNet, FrameNet, VerbNet and Wikidata knowledge could be compiled into one shared pivot schema with registered categories, which makes merging and tracing provenance explicit.
  • I2 (tree): UD dialect and treebank conversion could be written as explicit rules that validate the target structure, rather than as label renaming tables.
  • I3 (learn): every data conversion could report which source attributes had no target slot, as a "loss audit" that makes silent information loss visible during evaluation.
  • I4 (tokens): lemma entries in crate en could be keyed by paradigm as well as by surface form, so homographs with different inflection never collapse into one lemma.
  • I5 (tokens): morphology entries could carry an explicit, checkable pointer from each form to its base lexeme, which keeps lemmatisation explainable when several lexicons are merged.

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 Syntax & grammar formalisms

 Institutions

🔬Institut national de recherche en sciences et technologies du numériqueLe Chesnay-Rocquencourt · France55 works · 16,624 citations🔬Centre Inria de SaclayPalaiseau · France2 works · 211 citations

Cite

Cite as Kais Haddar, Hela Fehri and Laurent Romary. 2012. A prototype for projecting HPSG syntactic lexica towards LMF. arXiv:1207.5328 · Journal of Language Technology and Computational Linguistics 27, 1 (2012) 21-46.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.1207.5328 · arXiv 1207.5328

@misc{haddar2012prototype,
  title = {{A prototype for projecting HPSG syntactic lexica towards LMF}},
  author = {Kais Haddar, Hela Fehri and Laurent Romary},
  year = {2012},
  doi = {10.48550/arxiv.1207.5328},
  eprint = {1207.5328},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1207.5328}
}

References

26 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Abdelkader, A. (2006). Etude et analyse de la phrase nominale arabe en HPSG. Mémoire de mastère, SINT, FSEGS, Sfax. find
  2. Ammar, S., Dichy, J. (1999). Les verbes arabes. Collection Bescherelle, Hatier, France. find
  3. Atkins S. et al. (2002) From Resources to Applications. Designing The Multilingual ISLE Lexical Entry. In Proceedings of LREC 2002, Las Palmas, Canary Islands, Spain. find
  4. Blache, P. (1995). Une introduction aix.fr/~blache/papers/Intro-hpsg.ps.gz Chabchoub, S. (2005). Etude et représentation des verbes arabes en HPSG en vue de construire un conjugueur automatique. Mémoire de mastère, SINT, FSEGS, Sfax. find
  5. Dahdah, A. (1992). Dictionnaire des règles de la langue arabe dans des tableaux. Librairie de Nachirun Lebanon, 5ème édition, Beyrouth. find
  6. Eagles (1996). EAGLES Recommendations on subcategorisation. Eagles document EAG-CLWG-SYNLEX. find
  7. Elleuch, S. (2004). Analyse syntaxique de la langue arabe sur le formalisme d’unification HPSG. Mémoire de mastère, SINT, FSEGS, Sfax. find
  8. Fehri, H. et al. (2006). Un système de projection du HPSG arabisé vers la plate-forme LMF. Dans les actes du colloque JETALA, Institut d’Etudes et de Recherches pour l’arabisation (IERA), Rabat du 05-07 juin. find
  9. Genelex (1994). Projet Eureka GENELEX, Rapport sur LE MULTILINGUISME. Version 2.0, France. find
  10. Hasni, E.et al. (2006). Un éditeur et vérificateur des contraintes lexicales pour la langue arabe. Dans les actes du colloque international sur l’informatique et ses applications, IA 2006, Oujda, 31 octobre et 1-2 novembre. à HPSG. 2LC-CNRS – http://www.lpl.univ-1S$ 3*45$678879:$6;<*$=>69?:$"*@A;+B$9CD79E$ $ Ide N., Romary L. (2004). A Registry of Standard Data Categories for Linguistic Annotation, In Proceedings of the Fourth Language Resources and Evaluation Conference (LREC). Lisbon – http://hal.inria.fr/inria-00099858 ↗
  11. Ide N. , Véronis J., (1995). Encoding Dictionaries, in Ide N. and Véronis J. (Eds.), The Text Encoding Initiative : Background and Context, Kluwer Academic Publishers, Dordrecht, pp 167–179. find
  12. ISO 12620:2009 Terminology and other language and content resources — Specification of data categories and management of a Data Category Registry for language resources. find
  13. ISO 24610-1:2006 Language resource management — Feature structures — Part 1: Feature structure representation. find
  14. ISO 24613:2008 Language resource management - Lexical markup framework (LMF). our card
  15. Kasper, R. et al. (1995). Compilation of HPSG to TAG. In proceeding of the 33rd ACL, pp 92–99. find
  16. Khemakem, A. (2006). ArabicLDB : une base lexicale normalisée pour la langue arabe. Mémoire de mastère, SINT, FSEGS, Sfax. Lemnitzer, L., Romary, L., Witt, A, “Representing human and machine dictionaries in Markup languages”, Dictionaries. An International Encyclopedia of Lexicography. Supplementary volume: Recent developments with special focus on computational lexicography Mouton de Gruyter (Ed.) (to appear) – http://hal.inria.fr/inria-00441215 ↗
  17. Levine R. D. , Meurers D. (2006). “Head-Driven Phrase Structure Grammar: Linguistic Approach, Formal Foundations, and Computational Realization”. Keith Brown (Ed.): Encyclopedia of Language and Linguistics, Second Edition. Oxford: Elsevier, pp 690-704. find
  18. Loukil, N. (2006). Une proposition de représentation normalisée des lexiques des grammaires d'unification. In Actes de TALN, Presse Universitaire de Louvain, 10 – 13 Avril 2006, Belgique. find
  19. Monachini M., Calzolari N. (1999). Standardization in the Lexicon. In: H. van Halteren (Ed.): Syntactic Wordclass Tagging. Kluwer, Dordrecht, pp 149–173. find
  20. Nguyen T. M. H. et al. (2006). A lexicon for Vietnamese language processing, Language Resources and Evaluation 40, 3-4 2006, pp 291-309 – http://hal.inria.fr/inria-00201451 ↗
  21. Pollard, C., Sag., I.A. (1994). Head-Driven Phrase Structure Grammar. Publié par la presse dans l’université de Chicago, Edition Golgoldmittu, Chicago, LSLI. find
  22. Ringersma, J., & Kemps-Snijders, M. (2007). Creating multimedia dictionaries of endangered languages using LEXUS. In H. van Hamme, & R. van Son (Eds.), Proceedings of Interspeech 2007 (pp. 65-68). Baixas, France: ISCA-Int.Speech Communication Assoc. find
  23. Romary, L., Ide, N. (2004). International Standard for a Linguistic Annotation Framework, Natural Language Engineering 10, 3-4, pp 211-225 – http://hal.inria.fr/hal-00164624 ↗
  24. Romary, L. et al. (2004). Standards going concrete: from LMF to Morphalou. In Coling 2004, August 29, Geneva, Switzerland, pp 22-28 – http://hal.inria.fr/inria-00121489 ↗
  25. Romary, L. (2001). An abstract model for the representation of multilingual terminological data: TMF - Terminological Markup Framework. In Proc. Tama 2001, Antwerpen, Belgium – http://hal.inria.fr/inria-00100405 ↗
  26. Yoshinaga, N., Miyao, Y. (2002). Grammar conversion from LTAG to HPSG. In Proc.of the sixth ESSLLI Student Session, University of Tokyo, pp 309-324. 12$ find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/haddar-2012-prototype-projecting-hpsg-syntactic-lexica/