Elamur Research⌕
☰
home / papers / 2025 / The Proper Treatment of Verbal Idioms in German Discourse Representation Structure Parsing

The Proper Treatment of Verbal Idioms in German Discourse Representation Structure Parsing

Paper · 2025 · Kilian Evang, Rafael Ehren and Laura Kallmeyer

Summary

Semantic parsing datasets rarely represent potentially idiomatic expressions (PIEs) properly. These are multi-lexeme combinations that can be read either literally or idiomatically. Because of this gap, nobody can check whether a parser tells the two readings apart and builds the right meaning for each. The authors take German sentences from the Parallel Meaning Bank (PMB) and merge them with the verbal-PIE sense annotations of Ehren et al. (2024), after adjudicating those annotations. The result is a set of discourse representation structures (DRSs) in which idiomatic occurrences carry idiom-specific concepts and VerbNet roles. Non-decomposable idioms get a single concept. Decomposable idioms get component-specific senses. A state-of-the-art sequence-to-sequence DRS parser does worse on PIE sentences than on average sentences. Adding the reannotated sentences to fine-tuning narrows this gap without any change to the architecture. Choosing the right concept in context is still difficult, and gains on idiomatic readings tend to cost accuracy on literal ones. A synthetic canonical-form test set isolates the mapping from idiom to sense. The work matters because it supplies the first reusable resources for evaluating idiom handling in DRS parsing.

Key points

  • Source data: Ehren et al. (2024) released 6,187 German PMB sentences containing verbal PIEs. The 2,204 sentences that at least one annotator marked idiomatic were adjudicated. They held 4,600 annotations covering 957 distinct PIEs.
  • Adjudication made the WordNet sense and VerbNet roles consistent for each PIE type. It also enforced the guidelines on copulas, auxiliaries/modals and negation words such as nicht, since these affect token–concept alignment.
  • The annotations were merged automatically into PMB 5.1.0 through its token–concept alignment. This yields a reannotated set of 2,186 German PMB sentences that the authors plan to submit to a future PMB release.
  • The synthetic dataset has 890 idioms in canonical form with placeholders (e.g., jemand, etwas), each paired with a DRS and split evenly into dev and test. On it, the enhanced model significantly beats the baseline (permutation test).
    • The baseline scores higher on decomposable idioms.
    • The enhanced model scores higher on non-decomposable idioms, which suggests it learned their one-concept structure.
  • Smatch versus training frequency: a single training example can already give a high score, but reliably decent scores appear only at about 10 or more examples.

Interesting ideas

  • I1. Testing idioms in bare canonical form with placeholder arguments removes syntactic and argument difficulty, so the score measures almost only the idiom-to-sense mapping.
  • I2. Separating decomposable idioms (each part keeps a specialized sense) from non-decomposable ones (all parts jointly evoke one concept, with no extra concepts) fixes how many predicates a multiword expression should yield.
  • I3. During adjudication, the same idiom always received the same sense and the same role labels for corresponding arguments, and auxiliaries, copulas and negation were kept out of the idiom span.
  • I4. The authors observe that idiomatic gains came with literal losses. This means both readings should always be evaluated together, so that a model that simply always picks one side is caught.
  • I5. The authors suggest making the literal-versus-idiomatic decision an explicit step handled by a specialized model, rather than leaving it implicit in end-to-end generation.

For Mova

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

  • I1 (learn): Build a placeholder-based probe set of English phrasal idioms ("someone spills the beans") to check frame selection separately from parsing errors.
  • I2 (frames): Store each idiom entry with a decomposability flag, so prag either emits one VerbNet frame or maps components to separate roles, as in beans→secret.
  • I3 (tree): Apply the same convention when locating idiom spans in UD trees, so idiom matching does not break on "can't stand" or passives.
  • I4 (learn): Pair every idiom test with literal-use sentences as negative controls, and report the two accuracies side by side.
  • I5 (frames): Add an explainable idiom-reading decision before frame construction, with its evidence logged, so downstream world building can see why a literal or idiomatic frame was chosen.

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 1Research subjects and languages the work is about.Idioms & multiword expressions · 217 works

Cite

Cite as Kilian Evang, Rafael Ehren and Laura Kallmeyer. 2025. The Proper Treatment of Verbal Idioms in German Discourse Representation Structure Parsing. Proceedings of the 16th International Conference on Computational Semantics.

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

@inproceedings{evang2025proper,
  title = {{The Proper Treatment of Verbal Idioms in German Discourse Representation Structure Parsing}},
  author = {Kilian Evang, Rafael Ehren and Laura Kallmeyer},
  booktitle = {Proceedings of the 16th International Conference on Computational Semantics},
  year = {2025},
  url = {https://aclanthology.org/2025.iwcs-main.15/}
}

References

24 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Lasha Abzianidze, Johannes Bjerva, Kilian Evang, Hessel Haagsma, Rik van Noord, Pierre Ludmann, Duc-Duy Nguyen, and Johan Bos. 2017. The Parallel Meaning Bank: Towards a multilingual corpus of translations annotated with compositional meaning representations. In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers, pages 242–247, Valencia, Spain. Association for Computational Linguistics.
  2. Lasha Abzianidze, Rik van Noord, Chunliu Wang, and Johan Bos. 2020. The parallel meaning bank: A framework for semantically annotating multiple languages. Applied mathematics and informatics, 25(2):45–60. ↗
  3. Timothy Baldwin and Su Nam Kim. 2010. Multiword expressions. In Nitin Indurkhya and Fred J. Damerau, editors, Handbook of Natural Language Processing, second edition edition, pages 267–292. CRC Press, Boca Raton. find
  4. Laura Banarescu, Claire Bonial, Shu Cai, Madalina Georgescu, Kira Griffitt, Ulf Hermjakob, Kevin Knight, Philipp Koehn, Martha Palmer, and Nathan Schneider. 2013. Abstract Meaning Representation for sembanking. In Proceedings of the 7th Linguistic Annotation Workshop and Interoperability with Discourse, pages 178–186, Sofia, Bulgaria. Association for Computational Linguistics.
  5. Johan Bos. 2023. The sequence notation: Catching complex meanings in simple graphs. In Proceedings of the 15th International Conference on Computational Semantics, pages 195–208, Nancy, France. Association for Computational Linguistics. find
  6. Shu Cai and Kevin Knight. 2013. Smatch: an evaluation metric for semantic feature structures. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 748–752, Sofia, Bulgaria. Association for Computational Linguistics. find
  7. Ann Copestake and Dan Flickinger. 2000. An open source grammar development environment and broad-coverage English grammar using HPSG. In Proceedings of the Second International Conference on Language Resources and Evaluation (LREC‘00), Athens, Greece. European Language Resources Association (ELRA). find
  8. Ann A. Copestake, Dan Flickinger, Carl Pollard, and Ivan A. Sag. 2005. Minimal recursion semantics: An introduction. Research on Language and Computation, 3:281–332. ↗
  9. James Curran, Stephen Clark, and Johan Bos. 2007. Linguistically motivated large-scale NLP with C&C and boxer. In Proceedings of the 45th Annual Meeting of the Association for Computational Linguistics Companion Volume Proceedings of the Demo and Poster Sessions, pages 33–36, Prague, Czech Republic. Association for Computational Linguistics.
  10. Rotem Dror, Gili Baumer, Segev Shlomov, and Roi Reichart. 2018. The hitchhiker‘s guide to testing statistical significance in natural language processing. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1383–1392, Melbourne, Australia. Association for Computational Linguistics. find
  11. Rafael Ehren, Kilian Evang, and Laura Kallmeyer. 2024. To leave no stone unturned: Annotating verbal idioms in the Parallel Meaning Bank. In Proceedings of the Joint Workshop on Multiword Expressions and Universal Dependencies (MWE-UD) @ LREC-COLING 2024, pages 115–124, Torino, Italia. ELRA and ICCL.
  12. Christiane Fellbaum. 1998. WordNet: An electronic lexical database. MIT Press. find
  13. Jeffrey Flanigan, Sam Thomson, Jaime Carbonell, Chris Dyer, and Noah A. Smith. 2014. A discriminative graph-based parser for the Abstract Meaning Representation. In Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1426–1436, Baltimore, Maryland. Association for Computational Linguistics. ↗
  14. Hessel Haagsma. 2020. A Bigger Frish to Fry: Scaling up the Automatic Understanding of Idiomatic Expressions. Ph.D. thesis, Rijksuniversiteit Groningen. find
  15. Hans Kamp and Uwe Reyle. 1993. From discourse to logic - introduction to modeltheoretic semantics of natural language, formal logic and discourse representation theory. In Studies in Linguistics and Philosophy. find
  16. Rik van Noord, Lasha Abzianidze, Antonio Toral, and Johan Bos. 2018. Exploring neural methods for parsing discourse representation structures. Transactions of the Association for Computational Linguistics, 6:619–633. ↗
  17. Rik van Noord, Antonio Toral, and Johan Bos. 2020. Character-level representations improve DRS-based semantic parsing even in the age of BERT. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 4587–4603, Online. Association for Computational Linguistics. find
  18. Geoffrey Nunberg, Ivan A. Sag, and Thomas Wasow. 1994. Idioms. Language, 70(3):491–538. Publisher: Linguistic Society of America. find
  19. Jan Odijk and Martin Kroon. 2024. A canonical form for flexible multiword expressions. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 91–101, Torino, Italia. ELRA and ICCL.
  20. Carlos Ramisch, Silvio Ricardo Cordeiro, Agata Savary, Veronika Vincze, Verginica Barbu Mititelu, Archna Bhatia, Maja Buljan, Marie Candito, Polona Gantar, Voula Giouli, Tunga Güngör, Abdelati Hawwari, Uxoa Iñurrieta, Jolanta Kovalevskaitė, Simon Krek, Timm Lichte, Chaya Liebeskind, Johanna Monti, Carla Parra Escartı́n, Behrang QasemiZadeh, Renata Ramisch, Nathan Schneider, Ivelina Stoyanova, Ashwini Vaidya, and Abigail Walsh. 2018. Edition 1.1 of the PARSEME shared task on automatic identification of verbal multiword expressions. In Proceedings of the Joint Workshop on Linguistic Annotation, Multiword Expressions and Constructions (LAW-MWE-CxG-2018), pages 222–240, Santa Fe, New Mexico, USA. Association for Computational Linguistics.
  21. Zacchary Sadeddine, Juri Opitz, and Fabian Suchanek. 2024. A survey of meaning representations – from theory to practical utility. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 2877–2892, Mexico City, Mexico. Association for Computational Linguistics. find
  22. Chunliu Wang, Huiyuan Lai, Malvina Nissim, and Johan Bos. 2023. Pre-trained language-meaning models for multilingual parsing and generation. In Findings of the Association for Computational Linguistics: ACL 2023, pages 5586–5600, Toronto, Canada. Association for Computational Linguistics. find
  23. Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, and Colin Raffel. 2022. ByT5: Towards a token-free future with pre-trained byte-to-byte models. Transactions of the Association for Computational Linguistics, 10:291–306. ↗
  24. Xiao Zhang, Chunliu Wang, Rik van Noord, and Johan Bos. 2024. Gaining more insight into neural semantic parsing with challenging benchmarks. In Proceedings of the Fifth International Workshop on Designing Meaning Representations @ LREC-COLING 2024, pages 162–175, Torino, Italia. ELRA and ICCL. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/evang-2025-proper-treatment-verbal-idioms-german/