Elamur Research⌕
☰
home / papers / 2012 / Building an Arabic Multiword Expressions Repository

Building an Arabic Multiword Expressions Repository

Paper · 2012 · Abdelati Hawwari, Kfir Bar and Mona Diab

Summary

The paper addresses the lack of a broad, morphologically annotated list of Arabic multiword expressions (MWEs). Such a list is needed for token-level MWE classification, machine translation and other meaning-oriented NLP. In Arabic, one idiom can appear in many inflected and cliticized forms, while some idioms are frozen and do not agree with their context. The authors took roughly 5,000 expressions from several Arabic dictionaries and had native linguists filter them. Each word was analyzed with MADA and then checked by hand against SAMA analyses. Each entry was assigned a syntactic class, and dictionary placeholders were normalized into a few generic types tied to semantic roles. The resulting repository has 4,209 MWE types. A deterministic pattern matcher then tagged part of Arabic Gigaword 4.0, which had been preprocessed with AMIRAN. The matcher works on lemmas, enforces clitic constraints and allows noun-phrase gaps. It found about 481K MWE instances, and a small manual evaluation showed high precision for most classes. The work provides a seed resource for later statistical MWE identification in Arabic.

Key points

  • The final repository contains 4,209 MWE types in five classes: VVC 41, VNC 1,974, VPC 670, NNC 1,239 and ANC 285.
  • Placeholder words were reduced to three generic types, each linked to a semantic role:
    • flAn for a person (Agent/Patient)
    • k*A for an object (Goal)
    • <mr for an issue (Source)
  • The corpus was preprocessed with AMIRAN, which combines AMIRA, SAMA and NER tags to supply clitics, lemma, stem, POS, base-phrase chunks and NER.
  • Matching over 250M tokens of Arabic Gigaword 4.0 produced 481,131 instances in IOB format. NNC accounted for 316,393 of them, VPC for 75,844 and VNC for 64,504.
  • A native speaker checked a subsample, with these accuracies:
    • VNC: 98% (154 of 157 correct).
    • VPC: 77.6% (125 of 161).
    • NNC: 99% (154 of 155). NNC was the only class in the sample that used gapping.
    • VVC: 2 of 111 correct. A single misfiring type accounted for 109 of those instances.

Interesting ideas

  • I1. Instead of listing idioms as fixed strings, the authors store them with typed slots (person, object, issue) mapped to semantic roles, and require any filler to agree morphologically with its context.
  • I2. Each MWE word is annotated with which features are free to inflect and which are frozen, and the matcher checks clitics (attached prepositions or particles) on top of lemma identity.
  • I3. Gaps are allowed only when a whole non-recursive noun phrase fills an MWE noun slot and its head is the word that matches, which stops arbitrary gaps from flooding the output with false positives.
  • I4. The authors deliberately use a high-precision deterministic tagger to create seed annotation for a future statistical classifier, and acknowledge that literal and idiomatic uses must later be told apart in context.
  • I5. The evaluation reports both instance counts and unique type counts for each class, which showed that one bad pattern caused nearly all VVC errors.

For Mova

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

  • I1 (frames): Idiom entries such as "[someone] turned a blind eye to [something]" could carry slots that link directly to VerbNet-style roles. This would let prag build who-did-what frames from idioms rather than from literal word meanings.
  • I2 (tokens): The en crate could match English MWEs on lemmas while still requiring specific particles or prepositions (e.g., "take advantage of," phrasal verbs), and flag frozen forms like "by and large" that never inflect.
  • I3 (tree): MWE detection could match over UD subtrees, where an MWE noun matches the head of an nsubj/obj phrase and its modifiers are absorbed, instead of relying on linear windows.
  • I4 (learn): Lexicon-matched MWE spans could act as one silver witness. A separate literal-versus-idiomatic check, such as an LLM teacher or negative controls, could then filter them before training.
  • I5 (learn): Evaluations of rules and lexicons could always break errors down by type, so that a single faulty rule or entry is caught and repaired instead of dragging down an aggregate score.

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

 Institutions

🎓Columbia UniversityNew York · United States254 works · 56,747 citations

 Fields, topics and concepts

📚 Shelves 6The 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 worksPsychology · 1,649 worksDevelopmental and Educational Psychology · 523 works

🧭 Topics 2Research subjects and languages the work is about.Semitic languages · 3,066 worksIdioms & multiword expressions · 222 works

Cite

Cite as Abdelati Hawwari, Kfir Bar and Mona Diab. 2012. Building an Arabic Multiword Expressions Repository. Proceedings of the ACL 2012 Joint Workshop on Statistical Parsing and Semantic Processing of Morphologically Rich Languages.

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

@inproceedings{hawwari2012building,
  title = {{Building an Arabic Multiword Expressions Repository}},
  author = {Abdelati Hawwari, Kfir Bar and Mona Diab},
  booktitle = {Proceedings of the ACL 2012 Joint Workshop on Statistical Parsing and Semantic Processing of Morphologically Rich Languages},
  year = {2012},
  url = {https://aclanthology.org/W12-3403/}
}

References

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

  1. Attia, Mohammed, Antonio Toral, Lamia Tounsi, Pavel Pecina and Josef van Genabith. 2010. Automatic extraction of Arabic multiword expressions. In Proceedings of the 7th Conference on Language Resources and Evaluation, LREC-2010, Valletta, Malta.
  2. Abou Saad, Ahmed. 1987. A Dictionary of Arabic Idiomatic Expressions (mu’jm altrakib wala’barat alastlahiah ala’rbiah alkdim mnha walmould). Dar El Ilm Lilmalayin. find
  3. Benajiba, Yassine, Mona Diab and Paolo Rosso. 2008. Arabic Named Entity Recognition: An SVM-based approach. In Proceedings of the Arab International Conference on Information Technology, ACIT-2008, Tunisia. ↗
  4. Buckwalter, Tim. 2002. Buckwalter Arabic Morphological Analyzer Version 1.0. LDC catalog number LDC2002L49, ISBN 1-58563-257-0. ↗
  5. Carpuat, Marine and Mona Diab. 2010. Task-based Evaluation of Multiword Expressions: a Pilot Study in Statistical Machine Translation. HLT-NAACL. find
  6. Cook, Paul, Afsaneh Fazly, and Suzanne Stevenson. 2008. The VNC-Tokens Dataset. In Proceedings of the LREC Workshop on Towards a Shared Task for Multiword Expressions (MWE 2008), Marrakech, Morocco. find
  7. Dawood, Mohammed. 2003. A Dictionary of Arabic Contemporary Idioms (mu’jm alta’bir alastlahiat). Dar Ghareeb. find
  8. Diab, Mona and Pravin Bhutada. 2009. Verb noun construction MWE token supervised classification. In Workshop on Multiword Expressions (ACL-IJCNLP), pp. 17–22. find
  9. Diab, Mona. 2009. Second Generation Tools (AMIRA 2.0): Fast and Robust Tokenization, POS tagging, and Base Phrase Chunking. MEDAR 2nd International Conference on Arabic Language Resources and Tools, Cairo, Egypt. find
  10. Fayed, Wafaa Kamel. 2007. A Dictionary of Arabic Contemporary Idioms (mu’jm alta’bir alastlahiat). Abu Elhoul. find
  11. Fellbaum, Christiane. 1998. WordNet: An Electronic Lexical Database. MIT Press. ↗
  12. Habash, Nizar and Owen Rambow. 2005. Arabic Tokenization, Morphological Analysis, and Part-of-Speech Tagging in One Fell Swoop. In Proceedings of the Conference of American Association for Computational Linguistics, Ann Arbor, MI, 578-580. find
  13. Hashimoto, Chikara and Daisuke Kawahara. 2008. Construction of an idiom corpus and its application to idiom identification based on WSD incorporating idiom-specific features. In Proceedings of the 2008 Conference on Empirical Methods in Natural Language Processing, Honolulu, Hawaii, pages 992– 1001. find
  14. Katz, Graham and Eugenie Giesbrecht. 2006. Automatic identification of non-compositional multiword expressions using latent semantic analysis. In Proceedings of the Workshop on Multiword Expressions: Identifying and Exploiting Underlying Properties, Sydney, Australia, pages 12–19. find
  15. Maamouri, Mohamed, et al. 2010. Standard Arabic Morphological Analyzer (SAMA) Version 3.1. find
  16. Linguistic Data Consortium, Philadelphia Parker, Robert, et al. 2009. Arabic Gigaword Fourth Edition LDC2009T30, ISBN 1-58563-532-4. Linguistic Data Consortium (LDC), Philadelphia Roth, R., Rambow, O., Habash, N., Diab, M. and Rudin, C. 2008. Arabic Morphological Tagging, Diacritization, and Lemmatization Using Lexeme Models and Feature Ranking. In Proceedings of Association for Computational Linguistics (ACL), Columbus, Ohio.
  17. Sag, Ivan A. and Timothy Baldwin, Francis Bond, Ann A. Copestake, and Dan Flickinger. 2002. Multiword expressions: A pain in the neck for nlp. In Proceedings of the Third International Conference on Computational Linguistics and Intelligent Text Processing, London, UK, pp. 1–15.
  18. Sieny, Mahmoud Esmail, Mokhtar A. Hussein and Sayyed A. Al-Doush. 1996. A contextual Dictionary of Idioms (almu’jm alsyaqi lelta’birat alastlahiah). Librairie du Liban Publishers. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/hawwari-2012-building-arabic-multiword-expressions-repository/