--- title: "Annotating Verbal Multiword Expressions in — Mohamed 2022" description: "The paper extends the PARSEME framework for verbal multiword expressions (VMWEs) to Modern Standard Arabic." authors: ["Najet Hadj Mohamed", "Cherifa Ben Khelil", "Agata Savary", "Iskandar Keskes", "Jean-Yves Antoine", "Lamia Hadrich-Belguith"] year: 2022 venue: "Proceedings of the Thirteenth Language Resources and Evaluation Conference" source: "https://aclanthology.org/2022.lrec-1.196/" licence: "CC BY-NC 4.0 (ELRA, LREC proceedings)" licence_url: "https://creativecommons.org/licenses/by-nc/4.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/mohamed-2022-annotating-verbal-multiword-expressions-arabic/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/mohamed-2022-annotating-verbal-multiword-expressions-arabic/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2022](https://md.research.elamur.ai/free/en/papers/2022/) / Annotating Verbal Multiword Expressions in Arabic: Assessing the Validity of a Multilingual Annotation Procedure # Annotating Verbal Multiword Expressions in Arabic: Assessing the Validity of a Multilingual Annotation Procedure **Authors:** Najet Hadj Mohamed, Cherifa Ben Khelil, [Agata Savary](https://md.research.elamur.ai/free/en/authors/agata-savary/), Iskandar Keskes, [Jean-Yves Antoine](https://md.research.elamur.ai/free/en/authors/jean-yves-antoine/), [Lamia Hadrich-Belguith](https://md.research.elamur.ai/free/en/authors/lamia-hadrich-belguith/) · **Year:** 2022 · **Venue:** Proceedings of the Thirteenth Language Resources and Evaluation Conference · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/) **Links:** [Publisher's page](https://aclanthology.org/2022.lrec-1.196/) · [PDF](https://aclanthology.org/2022.lrec-1.196.pdf) · [OpenAlex](https://openalex.org/W4283836407) **Citations:** 2 citations (openalex-snapshot · 2026-10-14) · 3 in our library, of the 140884 papers whose bibliographies we read ## Summary The paper extends the PARSEME framework for verbal multiword expressions (VMWEs) to Modern Standard Arabic. An earlier Arabic PARSEME corpus was never openly released, so Arabic was absent from the PARSEME shared tasks. The authors rebuilt the resource from scratch on top of the Prague Arabic Dependency Treebank (PADT). Two native speakers independently annotated a sample of 1,062 sentences with the multilingual guidelines. Their agreement was measured at the very start of the campaign as a test of whether the guidelines fit Arabic. The current corpus has over 1,250 VMWEs, which already outnumbers the smallest PARSEME corpora. Arabic's agreement scores land mid-table among PARSEME languages, even though other languages measured agreement only after their annotators had matured. Light verb constructions and verbal idioms dominate the categories. Reflexive and particle-verb types do not occur. Arabic VMWEs are unusually discontinuous, second only to German. The conclusion is that the PARSEME guidelines carry over to Arabic without major changes. The result is an open, reproducible Arabic VMWE resource that also lets Arabic join cross-lingual MWE studies. ## Key points - The corpus is built on PADT and contains 1,847 sentences and 70,498 tokens. It holds 1,252 VMWEs, about 0.68 per sentence, and is released under CC-BY v4 in the PARSEME repository together with the double-annotated agreement sample. - Category counts: - LVC.full: 673 - VID: 345 - IAV: 165 - LVC.cause: 68 - MVC: 1 - IRV and VPC: 0 - Among 26 PARSEME agreement estimations, Arabic has the second-largest number of VMWEs used for the estimate (after Chinese). It ranks 12th on F span, 14th on κ span and 12th on κ cat. - The average VMWE length is 2.26 tokens, and over 73% of VMWEs have two tokens. - Only 42.17% of VMWEs are continuous, and 17.33% have gaps longer than 3. The mean of 1.97 gaps per VMWE is second only to German (2.96). ## Interesting ideas - Agreement was measured at the start of annotation rather than at the end, so it served as an early check on guideline soundness and annotator readiness. — For Mova (learn): Before scaling up silver data, run the witness parsers and LLM teacher tiers on a small pilot sample and measure their agreement, so that weak guidelines or witnesses are caught early. - For span-based kappa, the chance-agreement term was approximated by the number of verbs, since nearly every VMWE has a verbal head. — For Mova (learn): When comparing witnesses on frame or MWE detection, compute a kappa whose chance baseline is the count of candidate heads (verbs), instead of reporting raw overlap. - Category agreement (κ cat) was scored only on items where both annotators chose the exact same span, and partial overlaps counted as full errors. — For Mova (frames): Evaluate semantic-role labels separately from argument-span detection, so that boundary errors and labeling errors can be told apart. - Each corpus was profiled by VMWE length and by gap distribution, treating discontinuity as a key property of the language. — For Mova (tree): Measure how often English particle verbs and light-verb constructions are split by intervening words (e.g. "pick the box up"), which would show whether MWE detection must work over UD tree links rather than linear windows. - The shared typology merges only phenomena that are truly alike, and fine distinctions stay separate, such as LVC.full versus LVC.cause. — For Mova (frames): When mapping light verb constructions like "take a walk" onto frames, keep the causative variants distinct so the noun's frame and the causer role are assigned correctly. *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](https://md.research.elamur.ai/free/en/shelves/linguistics/) ## Cite The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{mohamed2022annotating, title = {{Annotating Verbal Multiword Expressions in Arabic: Assessing the Validity of a Multilingual Annotation Procedure}}, author = {Mohamed, Najet Hadj and Khelil, Cherifa Ben and Savary, Agata and Keskes, Iskandar and Antoine, Jean-Yves and Hadrich-Belguith, Lamia}, booktitle = {Proceedings of the Thirteenth Language Resources and Evaluation Conference}, year = {2022}, doi = {10.63317/5drvhtqy8gsh}, url = {https://doi.org/10.63317/5drvhtqy8gsh} } ``` - [← previous ANEC: An Amharic Named Entity Corpus and Transformer Based Recognizer](https://md.research.elamur.ai/free/en/papers/jibril-2022-anec-amharic-named-entity-corpus/) - [next → Annotators with Attitudes: How Annotator Beliefs And Identities Bias Toxic Language Detection](https://md.research.elamur.ai/free/en/papers/sap-2022-annotators-attitudes-how-annotator-beliefs/) - [↑ back to papers of 2022](https://md.research.elamur.ai/free/en/papers/2022/) ## References 2 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. Abdou, A. (2012). Arabic Idioms: A Corpus Based Study. Routledge. [find](https://scholar.google.com/scholar?q=Abdou%2C+A%2E+%282012%29%2E+Arabic+Idioms%3A+A+Corpus+Based+Study%2E+Routledge%2E) 2. Al-Badrashiny, M., Hawwari, A., Ghoneim, M., and Diab, M. (2016). SAMER: a semi-automatically created lexical resource for Arabic verbal multiword expressions tokens paradigm [https://gitlab.com/parseme/parseme_corpus_ar](https://gitlab.com/parseme/parseme_corpus_ar) https://creativecommons.org/licenses/by/4.0/https://gitlab.com/parseme/utilities/-/blob/master/st-organizers/corpus-statistics/mwe-stats-simple.py and https://gitlab.com/parseme/utilities/-/blob/master/st-organizers/corpus-statistics/mwe-stats.py [↗](https://gitlab.com/parseme/parseme_corpus_ar)