arxiv: 2605.11779 · v1 · submitted 2026-05-12 · 💻 cs.CL

Recognition: no theorem link

Choosing features for classifying multiword expressions

Eric Laporte

Authors on Pith no claims yet

Pith reviewed 2026-05-13 06:28 UTC · model grok-4.3

classification 💻 cs.CL

keywords multiword expressionsclassification featurescomputational linguisticsmultilingual approachesnatural language processingexpression categoriesfeature reliability

0 comments

The pith

Selecting reliable features for classifying multiword expressions yields schemes more useful for computational applications across languages.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Multiword expressions form a varied collection that requires classification to support tasks in language technology. Features available for this purpose differ in how consistently they allow expressions to be placed into categories. The paper outlines an enhanced classification that draws on prior studies covering multiple languages. This selection process aims to produce categories that prove more practical for computational systems.

Core claim

The author outlines an enhanced classification for multiword expressions by choosing features that allow reliable assignment to classes, informed by previous works that take various languages into account, in order to increase suitability for computational use.

What carries the argument

The selection of features for multiword expression classification, where each feature is assessed by how reliably it supports consistent class assignment.

If this is right

Resulting classifications become more fruitful for computational use.
The classification gains greater suitability across many languages.
Class assignments for expressions achieve higher reliability.
Computational systems can apply the categories with more consistent outcomes.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Improved classifications could support more accurate phrase handling in translation or parsing systems.
The feature selection approach might extend to categorizing other variable linguistic constructions.
Direct comparison of classification performance on held-out language data would test the gain in reliability.

Load-bearing premise

Features for multiword expressions vary in reliability for class assignment, and prior multilingual works can guide the choice of better ones.

What would settle it

Apply the enhanced classification to multiword expressions in several languages and check whether class assignments show greater consistency or yield better results in computational tasks than earlier classifications.

Figures

Figures reproduced from arXiv: 2605.11779 by Eric Laporte.

read the original abstract

Multiword expressions (MWEs) are a heterogeneous set with a glaring need for classifications. Designing a satisfactory classification involves choosing features. In the case of MWEs, many features are a priori available. Not all features are equal in terms of how reliably MWEs can be assigned to classes. Accordingly, resulting classifications may be more or less fruitful for computational use. I outline an enhanced classification. In order to increase its suitability for many languages, I use previous works taking into account various languages.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper outlines an enhanced MWE classification by prioritizing reliable features from prior multi-language studies, but stays at a high conceptual level without specifics or tests.

read the letter

The main thing to know is that Laporte argues for choosing features more carefully when classifying multiword expressions so that assignments to classes become more consistent and the results more useful for computational tasks. He draws on earlier studies across languages to support a synthesis that should work better in practice for things like parsing or translation systems. This is a reasonable point because features really do vary in how reliably they can be applied, and ignoring that can lead to brittle classifications. The paper does well by keeping the focus on computational fruitfulness rather than pure theory and by treating prior work as a resource instead of reinventing the wheel. That kind of synthesis can save time for people already working in the area. The soft spots are the lack of concrete detail. There is no list of which features were selected, no clear criteria for deciding reliability, and no example or comparison showing that the enhanced version actually performs better than what already exists. Without those pieces the claim that it will be more fruitful remains untested. This is aimed at specialists already handling MWEs in NLP who might want ideas on feature selection. A reader looking for new data, quantitative results, or a ready-to-apply system will not find much here. I would send it for peer review because the underlying reasoning is straightforward and the topic matters to the subfield, even though the current version needs more development to stand on its own.

Referee Report

2 major / 1 minor

Summary. The manuscript outlines an enhanced classification for multiword expressions (MWEs) by selecting features from prior works across multiple languages. It argues that not all a priori available features are equally reliable for assigning MWEs to classes and that the resulting classifications therefore vary in their fruitfulness for computational applications. The central contribution is conceptual: a synthesis intended to increase cross-linguistic suitability.

Significance. If the feature-selection criteria and the resulting taxonomy can be shown to be more reliable and more portable than existing schemes, the work could supply a practical foundation for MWE-aware NLP pipelines in parsing, translation, and semantic analysis. At present the contribution remains a high-level proposal without concrete feature lists, reliability metrics, or empirical comparisons, so its significance is prospective rather than demonstrated.

major comments (2)

The manuscript never states the concrete set of features ultimately chosen or the explicit reliability criteria used to rank them. Without this information the central claim that the classification is 'enhanced' cannot be evaluated.
No comparison—qualitative or quantitative—is provided against any of the prior classifications referenced in the abstract. It is therefore impossible to verify the assertion that the new synthesis is more fruitful for computational use.

minor comments (1)

The abstract would benefit from a single concrete example illustrating one feature, its reliability assessment, and the class to which it assigns a sample MWE.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which highlight areas where the manuscript can be strengthened. We address each major comment below and indicate the revisions we will make.

read point-by-point responses

Referee: The manuscript never states the concrete set of features ultimately chosen or the explicit reliability criteria used to rank them. Without this information the central claim that the classification is 'enhanced' cannot be evaluated.

Authors: We agree that the current version presents the classification primarily at a conceptual level. In the revised manuscript we will add an explicit enumeration of the selected features, drawn from the referenced prior multi-language studies, together with the reliability criteria (e.g., cross-linguistic consistency of assignment and demonstrated utility in downstream computational tasks) used to rank and retain them. revision: yes
Referee: No comparison—qualitative or quantitative—is provided against any of the prior classifications referenced in the abstract. It is therefore impossible to verify the assertion that the new synthesis is more fruitful for computational use.

Authors: The manuscript is framed as a high-level synthesis rather than an empirical study. To address the concern we will insert a new qualitative comparison section that contrasts the proposed feature-based taxonomy with the main prior schemes cited in the abstract, focusing on gains in cross-linguistic portability and computational applicability. Quantitative benchmarking lies outside the scope of this conceptual paper, but we will include illustrative examples showing how the selected features improve class assignment reliability. revision: partial

Circularity Check

0 steps flagged

No significant circularity

full rationale

The manuscript is a conceptual outline of an enhanced MWE classification synthesized from prior multi-language studies. Its central claim asserts that feature reliability varies and cross-linguistic synthesis improves computational utility. No equations, fitted parameters, predictions, or self-referential definitions appear. The argument draws on external prior works without reducing any load-bearing step to a self-citation chain, tautology, or renaming of known results. The derivation is self-contained against external benchmarks and exhibits no circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The abstract does not specify any free parameters, axioms, or invented entities. The claim rests on the domain assumption that prior multi-language studies provide suitable features for an enhanced classification.

axioms (1)

domain assumption Previous works on MWEs across languages provide a basis for choosing reliable features that improve computational usefulness.
Invoked in the statement that the classification uses previous works to increase suitability for many languages.

pith-pipeline@v0.9.0 · 5356 in / 1040 out tokens · 52129 ms · 2026-05-13T06:28:21.671357+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

9 extracted references · 9 canonical work pages

[1]

In Jan Svartik (ed.), Directions in Corpus Linguistics: Proceedings of Nobel Symposium 82, Stockholm, 4 -8 August 1991, 35-

Corpus Linguistics or Computer -aided Armchair Linguistics. In Jan Svartik (ed.), Directions in Corpus Linguistics: Proceedings of Nobel Symposium 82, Stockholm, 4 -8 August 1991, 35-

work page 1991
[2]

SMIL, Journal of Linguistic Calculus 1978(3-4)

Syntax and lexicon: blessure ‘wound’, nœud ‘knot’, caresse ‘caress’... SMIL, Journal of Linguistic Calculus 1978(3-4). 55–72. Stockholm: Skriptor. Grégoire, Nicole

work page 1978
[3]

Revue Québécoise de Linguistique (RQL) 11(2)

Une classification des phrases « figées » du français. Revue Québécoise de Linguistique (RQL) 11(2). 151–185. Montréal: UQAM. Gross, Maurice. 1986 . Grammaire transformatio nnelle du français. Vol. 3, Syntaxe de l ’ adverbe. Paris : 28 ASSTRIL. Gross, Maurice

work page 1986
[4]

In Ronald Asher (ed.), The Encyclopaedia of language and linguistics, v ol

The lexicon-grammar of a language: Application to Fren ch. In Ronald Asher (ed.), The Encyclopaedia of language and linguistics, v ol. 4, 2195 –2205. Oxford/NewYork/Seoul/Tokyo: Pergamon. Gross, Maurice.1996. Les formes être Prép X du français. Lingvisticae Investigationes 20(2). 217–270. Gross, Maurice

work page 1996
[5]

Reuse of verb theoretical linguistic descriptions

Automatic construction and validation of Fren ch large lexical resources. Reuse of verb theoretical linguistic descriptions. In Proceedings of the Language Resources and Evaluation Conference, 627–636. Granada. Jespersen, Otto . 1942,

work page 1942
[6]

Noun Argument Structure for Prop Bank

Annotation Guidelines for NomBank. Noun Argument Structure for Prop Bank. https://nlp.cs.nyu.edu/meyers/nombank/nombank-specs-2007.pdf (January 2016). Unpublished manuscript. Meyers, Adam, Ruth Reeves, Catherine Macleod, Rachel Szekely, Veronika Zielinska, Brian You ng & Ralph Grishman

work page 2007
[7]

Lisboa: Instituto Nacional de Investigação Científica

Sintaxe dos P redicados Nominais com Estar (Linguística 12). Lisboa: Instituto Nacional de Investigação Científica. Ranchhod, Elisabete & Michele De Gioia, 1996, Comparative Romance Syntax. Frozen Adverbs in Italian and in Portuguese. Lingvisticae Investigationes 20(1). 33–85. Rassi, Amanda, Cristina Santos -Turati, Jorge Baptista, Nuno Mamede & Oto Vale

work page 1996
[8]

Berkeley, California: International Computer Science Institute

FrameNet II: Extended Theory and Practice . Berkeley, California: International Computer Science Institute. Sag, Ivan Andrew, Timothy Baldwin, Francis Bond, Ann Copestake & Dan Flickinger. 2002 . Multiword expressions: A pain in the neck for NLP. In Proceedings of the 3rd International Conference on Intelligent Text Processing and Computational Linguistic...

work page 2002
[9]

Stathi, Katerina

109–121. Stathi, Katerina. 2007 . A corpus -based analysis of adjectival modification in German idioms. In Christiane Fellbaum (ed.), Idioms and collocations. Corpus-based linguistic and lexicographic studies , 81–108. London/New York: Continuum. Tesnière, Lucien

work page 2007