Pith. sign in

REVIEW 3 major objections 5 minor 49 references

Towards Universal Semantics With Large Language Models

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read Fine-tuned 1B and 8B language models generate accurate, cross-translatable NSM explications and outperform GPT-4o on the paper's benchmark.

desk verdict First LLM-NSM paper: real contribution, but the accuracy claim is entangled in a circular evaluation loop. read the letter →

arxiv 2505.11764 v3 pith:2F2EZZFI submitted 2025-05-17 cs.CL cs.AI

classification cs.CLcs.AI
keywords NaturalSemanticMetalanguageprimesexplicationgenerationlargelanguagemodelslow-resourcelanguagescross-translatabilitysubstitutabilitytestuniversalrepresentation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that large language models can be trained to write Natural Semantic Metalanguage (NSM) explications: short paraphrases of a word's meaning built from a small universal vocabulary of semantic primes. It reports that small fine-tuned models, DeepNSM-1B and DeepNSM-8B, produce higher-quality explications than much larger general-purpose models such as GPT-4o, and that these explications survive round-trip translation into low-resource languages with less semantic drift than dictionary definitions. If true, this removes the main bottleneck to a universal semantic representation: the slow manual drafting of explications. The paper also contributes the first automatic evaluation metrics for legality, descriptive accuracy, and cross-translatability, plus a quality-filtered dataset of roughly 44,000 entries that makes training possible. The payoff would be a practical, human-readable intermediate layer for translation and semantic analysis.

What carries the argument

The central object is the NSM explication: a paraphrase of a target word's meaning composed of semantic primes, a fixed inventory of roughly 65 primitive word-meanings claimed to be lexicalized across most languages, with occasional non-prime 'semantic molecules' allowed for brevity. The argument runs through three automatic metrics. The Legality Score is $\alpha \cdot (\text{primes}-\text{molecules})/\text{total words}$, rewarding prime-heavy, non-circular text. The Substitutability Score averages, over three grader LLMs and four ambiguous passages, the capped log-probability gain $\min(\beta, \Delta_{\mathrm{baseline}}-\Delta_{\mathrm{min}}+\Delta_{\mathrm{ent}})$ for recovering a masked target word when the explication is supplied, with a cap $\beta=40$. Cross-translatability is measured by round-tripping text through a machine translation service into five low-resource languages and comparing back-translations to originals with BLEU and embedding similarity. These metrics select the highest-scoring candidates from hundreds of thousands of LLM-generated explications to form the training dataset, and the same metrics are then used to compare DeepNSM to general-purpose baselines.

What would settle it

An independent blind evaluation by human linguistic experts on the same 149 benchmark words would settle the claim: if native-speaker experts rank GPT-4o's explications as equal to or more accurate than DeepNSM's, or if a held-out set of expert-authored NSM explications scores lower on the substitutability metric than obviously wrong paraphrases, the central performance claim is refuted.

Watch

Extended reading notes

Core claim

The paper claims that NSM explication generation is not only automatable but works best in small, task-specific models: DeepNSM-1B and DeepNSM-8B achieve the highest overall Explication Scores on a hand-curated 149-word benchmark, beating GPT-4o and Gemini-2.0-Flash while also using more semantic primes, fewer non-prime molecules, and less circularity. The same explications are shown to be more robust than WordNet dictionary definitions when translated into five low-resource languages and back, supporting the framework's universality premise. The authors also show that filtering the training dataset by their automatic quality scores is what drives much of the gain: unfiltered fine-tuned models score noticeably worse.

Load-bearing premise

The substitutability score, which uses grader language models to measure whether an explication helps recover a masked target word, is assumed to track true descriptive accuracy; because the same score filters the training data and then evaluates the final models, the headline performance gap would be an artifact of the metric rather than a real semantic improvement if that assumption fails.

Editorial extensions

If this is right

  • Dedicated fine-tuning can substitute for model scale in a linguistically constrained generation task: 1B- and 8B-parameter models match or beat far larger general models on NSM explication quality.
  • NSM explications could serve as an intermediate representation in low-resource translation pipelines, since they survive round-trip translation with less semantic drift than dictionary-style definitions.
  • Automated legality, substitutability, and cross-translatability scores make it feasible to screen hundreds of thousands of candidate explications, enabling the first large-scale NSM dataset.
  • Because the filtered dataset is public, future work can fine-tune other model families or sizes without repeating the expensive candidate-generation and scoring pipeline.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The riskiest spot is the evaluation loop: the same substitutability metric filters the training data and then judges the final models, so an independent measure of meaning preservation, such as human paraphrase acceptability testing, would be needed to confirm the metric is not rewarding a proxy that only looks like accuracy.
  • If explication generation generalizes beyond English, semantic primes could become an interlingua for machine translation into low-resource languages, but that would require testing the prime inventory directly in each target language rather than only through round-trip machine translation.
  • A testable extension is to use DeepNSM-generated prime paraphrases as an auxiliary training signal for low-resource language models, letting those models learn word meanings by their distribution near known prime expressions, as the paper sketches in its appendix.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. This paper presents DeepNSM, the first attempt to adapt large language models to generate Natural Semantic Metalanguage (NSM) explications. The authors propose three automatic evaluation methods: a legality score based on prime/molecule counts (Eq. 1), a substitutability score using LLM log-probability recovery of masked target words (Eq. 5), and a round-trip cross-translatability test. They build a dataset of approximately 44,000 entries by having Gemini-2.0-Flash generate candidate explications for WordNet senses and filtering them with their Explication Score (Eq. 6), then fine-tune 1B and 8B Llama models. On a hand-curated 149-word benchmark, the filtered DeepNSM models achieve the highest Explication Score and competitive cross-translatability across five low-resource languages; a brief human ranking is also reported. The central claim is that 1B and 8B models outperform GPT-4o in producing accurate, cross-translatable explications.

Significance. If the central claim held, this would be a valuable contribution: it is the first open-sourced pipeline for NSM explication generation, with code, models, and dataset released; it introduces automatic metrics that, if validated, would enable scalable research on NSM; and the cross-translatability results support the practical relevance of NSM-style representations for low-resource translation. The paper also makes concrete falsifiable predictions, such as the ill/sick analysis in Appendix B. However, the significance is conditional: because the main evaluation metric is also the data-filtering criterion, the headline accuracy advantage over GPT-4o needs an independent validation before the contribution can be accepted as stated.

major comments (3)
  1. [§4, §5.2, Table 1] The Explication Score (Eq. 6) is used both to select training data in §4 (threshold ≥35, passed by fewer than 15% of candidates) and to evaluate models in Table 1. Fine-tuning on data selected to maximize a metric is expected to improve that metric even when the metric does not track real semantic quality, so the reported accuracy advantage of DeepNSM over GPT-4o is partly self-fulfilling. The filtered-versus-unfiltered comparison (DeepNSM-* vs DeepNSM-*†) shows the filter moves the metric, but it does not show that the movement reflects genuine accuracy; the human ranking that could break this loop is reported only as aggregate percentages with no rater count, item count, expertise, or inter-annotator agreement. Please break the circularity, for example by validating the substitutability score against expert human judgments on a held-out set, or by reporting a fully specified human evaluation with agreement statistics.
  2. [§5.2, 'Metrics Align with Qualitative Judgements'] The human evaluation is too underspecified to serve as an independent validation. The text states that DeepNSM explications received top rankings 46% of the time, versus 28% for GPT, 21% for Gemini, and 5% for Llama, but it does not report how many raters judged how many items, what expertise the raters had, what instructions or rubrics were used, or any inter-annotator agreement statistic. Without this information the claimed alignment between automatic metrics and human judgment is not verifiable, and the human evaluation cannot break the circularity of using the same metric for filtering and evaluation.
  3. [§3.2, §5.2, Limitations] The substitutability score (Eq. 5) is an unvalidated LLM-based heuristic: the Limitations section concedes that the grader LLMs' alignment with human judgments is unverified. Section 5.2 further shows that dictionary definitions score highest on substitutability (12.14) yet are dismissed as 'cheating' because they reuse the target word; this demonstrates that the raw substitutability metric does not enforce non-circularity, and only the explicit zeroing in Eq. 6 suppresses this artifact. The central 'accurate explications' claim therefore depends on a metric that is known to be gameable in a direction that favors definitions over NSM explications; a validation study against expert-authored explications is needed.
minor comments (5)
  1. [Table 2, §5.2] The claim that NSM explications 'consistently achieve the highest cross-translatability scores across all languages tested' is not supported: DeepNSM-8B obtains BLEU 23.3 for Dzongkha, below Llama-3.1-8B-it's 23.8, and embedding similarity 91.7 for Kinyarwanda, below Gemini-2.0-Flash's 93.0. Please qualify this statement.
  2. [Eqs. 3–4, Appendix C] Equations 3 and 4 are described as averages but contain no division by k, whereas Algorithm 1 in Appendix C divides by k after the loop; please make the notation consistent.
  3. [Eq. 5] The subscripts and superscripts in Eq. 5 (e.g., Δ(g,p) baseline− Δ(g,p) min + Δ(g,p) ent) obscure the intended grouping; please rewrite with a clearer notation such as Δ_base(g,p).
  4. [§3.1] The legality scoring counts 'words' without defining tokenization or handling of inflected and multiword exponents, and it does not address the polysemy of English prime exponents (e.g., 'above' in a non-spatial sense, as noted in Appendix A); this makes the legality component of the Explication Score potentially noisy.
  5. [Appendix Table 3] The reported standard errors in Table 3 (e.g., 0.0008 for several proportions) appear implausibly small; please specify how they were computed and what resampling procedure was used.

Circularity Check

1 steps flagged · score 6.0 of 10

The headline accuracy advantage is partly circular: the same Explication Score (Eq. 6) used to filter the training data is also the main metric used to show DeepNSM outperforms GPT-4o; the paper's own filtered-vs-unfiltered comparison shows filtering drives the metric.

  1. fitted input called prediction [Section 4 (Quality Filtering and Final Dataset) and Section 5.1/5.2 (Evaluation)]
    "Legality and substitutability scores are then combined to compute the overall Explication Score (Equation 6). This score serves as a quality indicator that can be used to filter out low-quality entries from the final dataset. ... After generating and scoring candidate explications, we filter out those with scores below 35, a quality threshold met by fewer than 15% of candidates ... The generated explications are evaluated using the legality, substitutability, and cross-translatability tests described in Section 3, as well as qualitative human ranking."

    The Explication Score (Eq. 6 = γ(Substitutability + Legality)) is used twice: first as the training-data quality filter (threshold ≥ 35, passed by fewer than 15% of candidates), then as the principal evaluation metric in Table 1 supporting the claim that DeepNSM 1B/8B 'consistently outperform or match' GPT-4o. Fine-tuning a model on data pre-selected to score highly on a metric is expected to raise that same metric even when the metric does not track semantic fidelity. The paper itself shows the filter changes exactly the reported quantity: quality-filtered DeepNSM-1B and 8B achieve a 34% and 15% relative increase in substitutability over their unfiltered counterparts.

full rationale

This paper is not a self-citation chain: NSM theory is imported from external linguistics literature, and the automatic metrics are newly defined in the paper itself. The main circularity is metric double-use. Eq. 6 is computed from Eq. 1 (legality) and Eq. 5 (substitutability); Section 4 uses Eq. 6 to filter which Gemini-generated explications enter the 43,000-entry training set, and Section 5 uses Eq. 6 to rank DeepNSM against GPT-4o in Table 1. Selecting training data to maximize a metric and then evaluating on that metric partially forces the result. The paper's own comparison to unfiltered models quantifies the effect (34%/15% relative substitutability gains), confirming the loop is load-bearing. Independent evidence partially mitigates the circularity: cross-translatability (Table 2) does not use Eq. 6 for filtering and favors DeepNSM on most low-resource languages, and the blind human ranking gives DeepNSM 46% top rankings vs GPT-4o's 28%. However, the human ranking is reported only as aggregate percentages with no annotator-expertise or inter-annotator detail, and the Limitations section concedes that substitutability relies on 7-12B grader models whose alignment with human judgments is unverified. Thus the accuracy half of the central claim is partially circular; the cross-translatability half is independent. Overall score 6.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The paper's central empirical claims rest on several hand-tuned evaluation constants and on the validity of LLM-based and translation-based metrics. These are domain assumptions or ad hoc choices that the paper does not independently validate against a gold standard. No new physical or conceptual entities are introduced beyond the fine-tuned model itself.

free parameters (5)
  • alpha (legality weighting) = 10
    Set by hand in Section 3.1 to scale the Legality Score; affects the Explication Score and therefore dataset filtering and final rankings.
  • beta (substitutability cap) = 40
    Set in Section 3.2 to cap extreme log-probability deltas from grader LLMs; affects Explication Score.
  • gamma (overall score scaling) = 2
    Set in Section 3.3 to normalize the maximum score to 100.
  • k (truncation depth) = 2
    Number of lines or sentences removed in minimality and entailment tests; chosen because explications usually have 3+ lines.
  • quality filtering threshold = 35
    Explication Score cutoff for dataset filtering in Section 4; fewer than 15% of candidates pass, and this threshold influences training data and downstream results.
assumptions (4)
  • domain assumption Semantic primes are universal and lexicalized across languages
    Basis of NSM theory, cited from [23,35,24,45]; the paper's universal semantics framing relies on it.
  • ad hoc to paper LLM log-probability substitutability is a valid measure of descriptive accuracy
    Proposed in Section 3.2 without external validation; used to filter training data and to evaluate models.
  • domain assumption Round-trip translation with Google Translate measures cross-translatability
    Proposed in Section 3.4; assumes the translation quality for the five low-resource languages is sufficient to reveal semantic drift, and that BLEU plus embedding similarity captures meaning preservation.
  • domain assumption WordNet sense inventory provides a representative distribution of word meanings
    Dataset builds on WordNet synsets, as described in Section 4.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Universal Semantics With Large Language Models." pith.science (2026). https://pith.science/paper/2F2EZZFI

@misc{pith2026250511764,
  author       = {Pith},
  title        = {Pith review of: Towards Universal Semantics With Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2F2EZZFI}},
  note         = {Machine review of arXiv:2505.11764}
}
read the original abstract

The Natural Semantic Metalanguage (NSM) is a linguistic theory based on a universal set of semantic primes: simple, primitive word-meanings that have been shown to exist in most, if not all, languages of the world. According to this framework, any word, regardless of complexity, can be paraphrased using these primes, revealing a clear and universally translatable meaning. These paraphrases, known as explications, can offer valuable applications for many natural language processing (NLP) tasks, but producing them has traditionally been a slow, manual process. In this work, we present the first study of using large language models (LLMs) to generate NSM explications. We introduce automatic evaluation methods, a tailored dataset for training and evaluation, and fine-tuned models for this task. Our 1B and 8B models outperform GPT-4o in producing accurate, cross-translatable explications, marking a significant step toward universal semantic representation with LLMs and opening up new possibilities for applications in semantic analysis, translation, and beyond. Our code is available at https://github.com/OSU-STARLAB/DeepNSM.

Figures

Figures reproduced from arXiv: 2505.11764 by the authors.

Figure 1
Figure 1. A list of the proposed Natural Semantic Primes, along with an example demonstrating how [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Envisioned interaction where a user prompts an LLM to generate an NSM-style explication [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Example outputs from instruction-tuned LLMs when prompted to generate an NSM-style [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Illustration of the cross-translatability test described in Section 3.4. Semantic descriptions [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Our dataset generation process described in Section 4. Target words are selected from [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Comparing the meaning of "ill" and "sick" using their NSM explications. Explications are [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: Distribution of Explication Scores assigned by different "grader" LLMs during substi [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: Model-generated explications for the word "apologize." Both the Llama-3.1-8B-Instruct [PITH_FULL_IMAGE:figures/full_fig_p020_8.png]
Figure 9
Figure 9. Figure 9: Model-generated explications for the word "elbow." Similarly to Figure 8, the Llama [PITH_FULL_IMAGE:figures/full_fig_p021_9.png]
Figure 10
Figure 10. Figure 10: Model-generated explications for the word "hypothesis." The word hypothesis refers to a [PITH_FULL_IMAGE:figures/full_fig_p022_10.png]
Figure 11
Figure 11. Figure 11: Model-generated explications for the word "rough." As with previous examples, the [PITH_FULL_IMAGE:figures/full_fig_p023_11.png]
Figure 12
Figure 12. Figure 12: Prompt for generating additional usage examples. [PITH_FULL_IMAGE:figures/full_fig_p024_12.png]
Figure 13
Figure 13. Figure 13: Prompt for generating additional ambiguous examples for the substitutability testing [PITH_FULL_IMAGE:figures/full_fig_p024_13.png]
Figure 14
Figure 14. Figure 14: Explication Generation System Prompt. 24 [PITH_FULL_IMAGE:figures/full_fig_p024_14.png]
Figure 15
Figure 15. Figure 15: Few-Shot Examples used for the Explication Generation Prompt. [PITH_FULL_IMAGE:figures/full_fig_p025_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 41 canonical work pages

  1. [1]

    A. B. Abacha, W.-w. Yim, Y . Fu, Z. Sun, M. Yetisgen, F. Xia, and T. Lin. Medec: A benchmark for medical error detection and correction in clinical notes. arXiv preprint arXiv:2412.19260, 2024

  2. [2]

    Achiam, S

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat, et al. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023

  3. [3]

    K. Allan. Linguistic Semantics, Vols I and II. London: Routledge & Kegan Paul, 1986

  4. [4]

    N. Arnawa. The implementation of natural semantic metalanguage and semantic field in language teaching: A case study. Journal of Language Teaching and Research, 8(3):507, 2017

  5. [5]

    J. T. Ball, S. Rodgers, R. Schvaneveldt, and A. Ball. Creating meaningful word vectors and examining their use as representations of word meaning. Proceedings of the Annual Meeting of the Cognitive Science Society, 46, 2024

  6. [6]

    S. Bird. Nltk: the natural language toolkit. In Proceedings of the COLING/ACL 2006 interactive presentation sessions, pages 69–72, 2006. 11

  7. [7]

    Bromhead

    H. Bromhead. The reign of truth and faith: epistemic expressions in 16th and 17th century English. Number 62 in Topics in English linguistics. Mouton de Gruyter, Berlin, 2009

  8. [8]

    Bromhead

    H. Bromhead. The semantics of bushfire in Australian English. In H. Bromhead and Z. Ye, editors, Meaning, Life and Culture: In conversation with Anna Wierzbicka , pages 115–134. ANU Press, 1st edition, Dec. 2020

Show all 49 references
  1. [9]

    Bromhead and C

    H. Bromhead and C. Goddard. Applied semantics and climate communication. Australian Review of Applied Linguistics, July 2023

  2. [10]

    J. Chen, S. Xiao, P. Zhang, K. Luo, D. Lian, and Z. Liu. Bge m3-embedding: Multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation.arXiv preprint arXiv:2402.03216, 2024

  3. [11]

    Conneau, K

    A. Conneau, K. Khandelwal, N. Goyal, V . Chaudhary, G. Wenzek, F. Guzmán, E. Grave, M. Ott, L. Zettlemoyer, and V . Stoyanov. Unsupervised cross-lingual representation learning at scale. arXiv preprint arXiv:1911.02116, 2019

  4. [12]

    Dettmers, A

    T. Dettmers, A. Pagnoni, A. Holtzman, and L. Zettlemoyer. Qlora: Efficient finetuning of quantized llms. arXiv preprint arXiv:2305.14314, 36, 2023

  5. [13]

    Dimakis, S

    A. Dimakis, S. Markantonatou, and A. Anastasopoulos. Dictionary-aided translation for handling multi-word expressions in low-resource languages. In L.-W. Ku, A. Martins, and V . Srikumar, editors,Findings of the Association for Computational Linguistics: ACL 2024 , pages 2588–...

  6. [14]

    Elsner and J

    M. Elsner and J. Needle. Translating a low-resource language using gpt-3 and a human-readable dictionary. In Proceedings of the 20th SIGMORPHON workshop on Computational Research in Phonetics, Phonology, and Morphology, pages 1–13, 2023

  7. [15]

    Fähndrich

    J. Fähndrich. Semantic decomposition and marker passing in an artificial representation of meaning. PhD Thesis, Technische Universität Berlin, 2018

  8. [16]

    Giulianelli, I

    M. Giulianelli, I. Luden, R. Fernandez, and A. Kutuzov. Interpretable Word Sense Repre- sentations via Definition Generation: The Case of Semantic Change Analysis, July 2023. arXiv:2305.11993 [cs]

  9. [17]

    C. Goddard. Testing the translatability of semantic primitives into an australian aboriginal language. Anthropological Linguistics, 33:31–56, 1991

  10. [18]

    C. Goddard. Semantic Analysis: A Practical Introduction. Oxford University Press, Oxford, first edition edition, 1998

  11. [19]

    C. Goddard. 1. Natural Semantic Metalanguage: The state of the art. In C. Goddard, editor, Studies in Language Companion Series, volume 102, pages 1–34. John Benjamins Publishing Company, Amsterdam, Apr. 2008

  12. [20]

    Goddard, editor

    C. Goddard, editor. Cross-linguistic semantics. Number v. 102 in Studies in language companion series. John Benjamins Pub. Co, Amsterdam ; Philadelphia, 2008. OCLC: ocn192134491

  13. [21]

    C. Goddard. The natural semantic metalanguage approach. In B. Heine and H. Narrog, editors, The Oxford Handbook of Linguistic Analysis, pages 459–484. Oxford University Press, Oxford, 2009

  14. [22]

    Goddard and A

    C. Goddard and A. Wierzbicka. Cultural scripts: What are they and what are they good for? Intercultural Pragmatics, 1(2), Jan. 2004

  15. [23]

    Goddard and A

    C. Goddard and A. Wierzbicka, editors. Semantic and lexical universals: theory and empir- ical findings. Number v. 25 in Studies in language companion series (SLCS) 0165-7763. J. Benjamins, Amsterdam Philadelphia, 2010

  16. [24]

    Goddard and A

    C. Goddard and A. Wierzbicka. Semantic fieldwork and lexical universals. Studies in Language, 38(1):80–127, May 2014. 12

  17. [25]

    Goddard and A

    C. Goddard and A. Wierzbicka. Words and meanings: lexical semantics across domains, languages, and cultures. Oxford linguistics. Oxford University Press, Oxford, first edition edition, 2014. OCLC: ocn830367893

  18. [26]

    Virus”, “bacteria

    C. Goddard and A. Wierzbicka. Semantics in the time of coronavirus: “Virus”, “bacteria”, “germs”, “disease” and related concepts. Russian Journal of Linguistics, 25(1):7–23, Dec. 2021

  19. [27]

    Goddard, A

    C. Goddard, A. Wierzbicka, and H. Fabréga. Evolutionary semantics: using NSM to model stages in human cognitive evolution. Language Sciences, 42:60–79, Mar. 2014

  20. [28]

    Litaina, A

    T. Litaina, A. Soularidis, G. Bouchouras, K. Kotis, and E. Kavakli. Towards llm-based semantic analysis of historical legal documents. 2024

  21. [29]

    H. Man, N. T. Ngo, V . D. Lai, R. A. Rossi, F. Dernoncourt, and T. H. Nguyen. LUSIFER: Language Universal Space Integration for Enhanced Multilingual Embeddings with Large Language Models, Jan. 2025. arXiv:2501.00874 [cs]

  22. [30]

    S. Maskova. Semantic Analysis of Snow-related Words in Danish and Kalaallisut (West Greenlandic). Scandinavian Studies in Language, 13(1):225–248, Dec. 2022

  23. [31]

    J. D. McCawley. A program for logic. In Semantics of natural language, volume 40, pages 498–544. Springer, Dordrecht, 1972

  24. [32]

    R. Merx, A. Mahmudi, K. Langford, L. A. de Araujo, and E. Vylomova. Low-resource machine translation through retrieval-augmented llm prompting: a study on the mambai language. arXiv preprint arXiv:2404.04809, 2024

  25. [33]

    G. A. Miller. Wordnet: a lexical database for english. Communications of the ACM, 38(11):39– 41, 1995

  26. [34]

    Otomo and A

    A. Otomo and A. Torii. An nsm approach to the meaning of tear and its japanese equivalents. In Selected Papers from the 2005 Conference of the Australian Linguistic Society. Citeseer, 2006

  27. [35]

    B. Peeters. Nsm approach: Natural semantic metalanguage resource base. https://www. nsm-approach.net, 2025. Accessed: 2025-05-15

  28. [36]

    Periti and S

    F. Periti and S. Montanelli. Lexical Semantic Change through Large Language Models: a Survey. ACM Computing Surveys, 56(11):1–38, Nov. 2024

  29. [37]

    Reimers and I

    N. Reimers and I. Gurevych. Sentence-bert: Sentence embeddings using siamese bert-networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, 11 2019

  30. [38]

    Sadow and S

    L. Sadow and S. S. Fernández. Pedagogical pragmatics: Natural semantic metalanguage applications to language learning and teaching. Scandinavian Studies in Language, 13(1):53–66, 2022

  31. [39]

    Siino, M

    M. Siino, M. Falco, D. Croce, and P. Rosso. Exploring llms applications in law: A literature review on current legal nlp approaches. IEEE Access, 2025

  32. [40]

    Y . Song, L. Li, C. Lothritz, S. Ezzini, L. Sleem, N. Gentile, R. State, T. F. Bissyandé, and J. Klein. Is llm the silver bullet to low-resource languages machine translation? arXiv preprint arXiv:2503.24102, 2025

  33. [41]

    Touvron, T

    H. Touvron, T. Lavril, G. Izacard, X. Martinet, M.-A. Lachaux, T. Lacroix, B. Rozière, N. Goyal, E. Hambro, F. Azhar, et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023

  34. [42]

    Wierzbicka

    A. Wierzbicka. Understanding Cultures through Their Key Words English, Russian, Polish, German, and Japanese. Oxford University Press, Oxford, first edition edition, 1997

  35. [43]

    Wierzbicka

    A. Wierzbicka. Why there are no ‘colour universals’ in language and thought. Journal of the Royal Anthropological Institute, 14(2):407–425, June 2008. 13

  36. [44]

    Wierzbicka

    A. Wierzbicka. Experience, evidence, and sense: the hidden cultural legacy of English. Oxford university press, Oxford, 2010

  37. [45]

    semantic primitives

    A. Wierzbicka. “semantic primitives”, fifty years later.Russian Journal of Linguistics, 25(2):317– 342, 2021

  38. [46]

    B. J. Wilson and G. M. Farese. What did adam smith mean? the semantics of the opening key principles in the wealth of nations. In P. Sagar, editor,Interpreting Adam Smith: Critical Essays, pages 77–95. Cambridge University Press, Cambridge, 2023

  39. [47]

    Yu, C.-H

    Y . Yu, C.-H. H. Yang, J. Kolehmainen, P. G. Shivakumar, Y . Gu, S. R. R. Ren, Q. Luo, A. Gourav, I.-F. Chen, Y .-C. Liu, T. Dinh, A. G. D. Filimonov, S. Ghosh, A. Stolcke, A. Rastow, and I. Bulyko. Low-rank adaptation of large language model rescoring for parameter-efficient ...

  40. [48]

    Zamblera

    F. Zamblera. Computational nsm: a prolog-based notation. Online]. SÍNTESIS CURRICULAR, 2010

  41. [49]

    above” is a valid prime in the spatial sense (“the sky is above the ground

    M. Zeifert. Natural semantic (legal?) metalanguage. what can legal theory learn from anna wierzbicka? Between Text, Meaning and Legal Languages, page 173, 2023. 14 Appendix This appendix offers supplementary materials, background information, and extended discussions that coul...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.