Linguistic Patterns in European Public Organization Names

This paper presents an analysis of the names of public organizations with the purpose of classifying them into broad classes. The paper provides a very detailed (perhaps overdetailed) analysis of a number of algorithms. There is not really much novelty in the methodologies, but this study provides enough experimental depth to still be of interest to readers of the journal.

The dataset is of limited use, but would support reproducibility of the results.

The authors tend to capitalize terms for no reason. Here are some of the terms that should not be capitalized: natural language processing, natural language inference, large language models, entity linking, entity type classification, named entity recognition, (zero-shot) hierarchical text classification, entity normalization, entity classification, (supervised) contrastive learning, open government data, knowledge graphs, "Temporal Dynamics of Administration, Absence of Evolution-Aware Ontologies, Metadata Quality Deficits, Linguistic Heterogeneity and Organizational Disambiguation", information extraction, cosine similarity, (universal) classifier, logistic regression, support vector machine, principal component analysis

The definition $\mathcal{Y}(x) = \{ y_k \in \mathcal{Y} | y_k = 1 \}$ is totally incorrect. It defines a function of $x$, where the RHS doesn't use $x$; as $\mathcal{Y}$ is a function $y_k$ cannot be a member of this (functions are not sets); the RHS can be trivially simplified to $\{ 1 \}$ if $1 \in \mathcal{Y}$ or $\emptyset$ otherwise. In the same section, the authors should define what $\preceq$ means in this context.

"∃ k ∈ K such that their dot product yi · y j ⩾ 1.": This doesn't make sense as $k$ is not used. I think you just mean to say $y_{ik} = 1$ and $y_{jk} = 1"
"∀x ∈ X , P(i) denotes" - surely should be $P(x)$

Minor:
p2 generative Generative
p4 LaTeX backticks for `Ltd', `hospital'
p6 l13: commas between method names, they should probably also be cited
p6 l45: Paragraph that just says "Finally"
p10 l48: space before period
p11 l9: space before period
p15 l27: missing reference
p19 l29: "only six hospitals exist" (remove "do")
p25 l42: `others'
p27 l37-38: an em-dash is matched with a comma
p29 l37: "Then comes Logistic Regression." - weird sentence
p35: two conclusions sections, one which just contains "latex"

This version of the paper on multilingual detection of European Public Organization names shows significant improvement both in scope and in narrative. The first part of the manuscript argues about the need of such work and research and bring convincing arguments for the value such research and results bring. The related work section added provides extensive description of the problems and the approaches that can and have been adopted to tackle them successfully. While the terminology aspects of the three domains covered by the experiments remain sparsam from a linguistic stand point, it is important to emphasize the analysis of the different language characteristics and the evaluation of the different outcomes of the experiments with the three different methods that are subject of the study, and provides conclusions about the suitability of the methods for different classes of languages, and presents extensive data supporting the evidence. The multilingual aspects and the ambition to provide multilingual organization names detection have been not only well covered, but also very well exemplified with experiments result tables and discussions. The experiments settings and the technical details on the three methods applied have also been presented in greater detail than in the first version of he paper that brings clarity in the narrative. The generalizations of their performance and the analysis of the reasons for this performance is a contribution to the study of natural language processing methods in multilingual contexts. The obtained results and their analysis bring complementary evidence and arguments for the selection of a given approach for a particular task. This can also be regarded as a sensible contribution in the field of multilingual natural language processing.
Overall, the paper shows a large amount of work carried out. This version of the paper is mature enough to be presented to the audience interested in language models, natural language processing and multiliguality on the one hand and in application of natural language processing methods in real world settings, addressing real world information acquisition needs.