Pith. sign in

REVIEW 4 major objections 3 minor 23 references

Survey of Swarm Intelligence Approaches to Search Documents Based On Semantic Similarity

T0 review · 4 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A survey of swarm-intelligence text methods finds no PSO/ACO work on semantic document search.

desk verdict A readable but off-scope survey of ten SI papers that never actually surveys SI for semantic-similarity document search; the one useful gap claim needs a systematic search to be credible. read the letter →

arxiv 2507.11787 v1 pith:GFRFJED2 submitted 2025-07-15 cs.AI

Chandrashekar Muniyappa , Eunjin Kim This is my paper · ORCID
classification cs.AI
keywords swarmintelligenceparticleoptimizationantcolonysemanticsimilaritysentenceembeddingsdocumentsearchfeatureselectiontextclassification
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 is a survey, not a new algorithm. It sets out to review recent applications of swarm intelligence, chiefly particle swarm optimization (PSO) and ant colony optimization (ACO), to text problems and to see how they bear on searching documents by semantic similarity. Its stated finding is a gap: none of the studies it reviews applies PSO or ACO to advanced sentence-embedding vectors for identifying documents based on semantic similarity. The authors take this as the next research direction and say their own work on it is in progress. If the survey's scope is right, it maps where swarm intelligence already helps in text processing and identifies the retrieval problem it has not yet touched.

What carries the argument

The objects doing the work are the two classic swarm algorithms, PSO, a population search in which particles update velocities toward personal and global best solutions, and ACO, a graph search in which artificial ants lay down pheromone along good paths, along with their surveyed variants: integer and probabilistic PSO, clustering PSO, ACO-KNN, NPE-PSO, GCACO, and a two-stage multi-objective PSO. These algorithms are paired with text representations such as TF-IDF, bag-of-words, graph-coded features, and reduced feature vectors. For the survey's argument, that machinery serves as the inventory: it shows where swarm intelligence has been applied in text processing and, by absence, where it has not, namely semantic similarity search over sentence embeddings.

What would settle it

A literature search for any pre-2025 study applying PSO or ACO to sentence-embedding vectors for semantically similar document retrieval; one such study would directly falsify the survey's central conclusion.

Watch

Extended reading notes

Core claim

The paper's central claim is a negative result about the literature. After reviewing nine studies, it asserts that no existing work combines swarm intelligence, specifically PSO and ACO, with advanced sentence embedding vectors to identify documents based on semantic similarity. The reviewed applications use swarm algorithms for feature selection, clustering, recommendation, and classification, with representations such as TF-IDF, bag-of-words, and graph structures; none is a document-retrieval system that compares sentence-level semantic embeddings. The paper also draws the positive conclusion that, within these surveyed tasks, swarm-intelligence methods consistently outperform traditional similarity-measure and feature-selection baselines, which is the motivation for carrying the same algorithms into the open semantic-search problem.

Load-bearing premise

The survey's gap claim depends on the selected studies being genuinely representative of recent swarm-intelligence work on text similarity; if relevant studies were omitted or the reviewed papers do not actually address semantic document search, the conclusion does not follow.

Editorial extensions

If this is right

  • The survey defines a concrete open problem: apply PSO or ACO to sentence embedding vectors and evaluate them on semantic-similarity document search.
  • If the paper is right, researchers should test traditional similarity measures (cosine, Euclidean, Dice) on long paragraphs, where the survey predicts they will need swarm-based help.
  • Longer texts imply higher-dimensional embedding vectors; the survey's future-work list implies SI algorithms must be benchmarked on these large solution spaces and on operations that speed convergence.
  • The surveyed studies' repeated finding that SI improves text feature selection and classification is the paper's evidence that similar gains may be available in retrieval, though no retrieval result exists yet.

Reading between the lines

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

  • Beyond the paper: a quick way to test the gap claim is to search the pre-2025 literature for any PSO or ACO application to sentence embedding vectors for document retrieval; finding one would overturn the survey's conclusion.
  • Beyond the paper: the transfer from feature selection to retrieval is not automatic; cosine nearest-neighbor search over sentence embeddings may already perform well, and the paper does not include that baseline.
  • Beyond the paper: an experiment the survey's logic suggests is to run PSO/ACO on sentence embeddings over a standard retrieval benchmark and compare quality and runtime against direct similarity search, measuring whether the swarm adds value.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 3 minor

Summary. This manuscript, labeled a survey, claims in the abstract to 'review all the latest developments in Searching for documents based on semantic similarity using Swarm Intelligence algorithms.' The Main section describes ten papers that apply PSO, ACO, or hybrids to feature selection, text classification, recommendation, and clustering; the concluding section asserts that none of these applied SI to sentence embeddings for semantic-similarity document identification and lists future work. The paper provides no survey methodology, and the described papers do not, on the evidence of the paper's own summaries, address the advertised topic.

Significance. If the paper delivered what it promises, it would be a useful map of an emerging intersection and a crisply stated open problem. It has some small strengths: it organizes ten SI applications, reproduces selected formulas, and clearly separates the surveyed systems from the claimed gap. However, the central contribution is not supported. The survey is a convenience sample with no search protocol, most surveyed works are off-topic relative to the title and abstract, and the 'none applied' conclusion is an observation about the sample, not a verified research gap. The paper does not make predictive or derivational claims, so circularity is not at issue; the problem is correctness of scope and selection.

major comments (4)
  1. [Abstract and §2 Main] The abstract promises a review of 'all the latest developments in Searching for documents based on semantic similarity using Swarm Intelligence algorithms,' but the ten papers summarized in §2 are predominantly feature-selection and classification studies: [1], [3], [16], and [17] are feature-selection methods, [4] and [15] are text classification, [6] is a diet-recommendation system, and [9] clusters athlete behavior; only [7] concerns document categorization, and even that is described as classification rather than semantic-similarity ranking. The surveyed corpus does not match the advertised scope, so the central claim of the survey is not supported by its content.
  2. [§3, first bullet] The claim 'None of them applied Swarm Intelligence algorithms like PSO and ACO on advanced sentence embedding vectors text representations to identify documents based on semantic similarity' is only a statement about the ten sampled papers. The paper provides no methodology stating search queries, databases, inclusion/exclusion criteria, or coverage dates, so there is no evidence that the sample is representative of the literature; without a systematic search, the gap claim cannot bear the weight of the paper's central contribution.
  3. [§2, [1] description] The text states that 'integer programming PSO a variant of PSO [2] was applied' when describing Song et al. [1], but reference [2] is Kennedy's 'Bare bones particle swarms,' which is not an integer programming PSO variant; this citation mismatch is a concrete error in the technical description and suggests the summaries of the cited works are not fully reliable.
  4. [§2, Equations (2), (5), (7)] Several reproduced equations contain notational mistakes: Eq. (2) begins 'p^{t+1}_{i,j} = p^t_{i,j} = ...' and introduces an undefined quantity D_r; Eq. (5) uses u^k without definition; Eq. (7) writes '\tau_i \tau_i' where the Davies-Bouldin formula requires \sigma_i + \sigma_j. Because the survey reproduces these formulas as faithful records of the original works, these errors undermine confidence in the technical summaries.
minor comments (3)
  1. [Throughout] Typos including 'anf Future Work' (Section 3 heading), 'elemenated' (Section 2, [1]), 'bewteen' (Section 2, [1]), 'calcualted' (Section 2, [4]), 'deviced' (Section 2, [4]), 'luster' (Section 2, [9]), 'pheormone' (Section 2, [16]), 'ner equal' (Section 2, [1]), and 'os clusters' (Section 2, [1]) should be corrected.
  2. [Section 1 and Fig. 1] The text refers to 'Fig 1' as a sample of the Swarm Intelligence Hierarchy, but no figure is included in the manuscript; either add the figure or remove the reference.
  3. [Reference list] Several references are cited with incomplete or inconsistent bibliographic details (e.g., [17] is cited only via a ResearchGate DOI, and [10] is a book rather than a methodological reference for NLP pipelines); the reference list should be standardized.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper is a descriptive survey with no derived predictions or fitted inputs; its gap claim is an empirical literature claim, not a self-referential construction.

full rationale

This manuscript is a survey, not a derivation or prediction paper. Section 2 summarizes ten external studies with descriptive accounts of their algorithms and equations, and Section 3 draws a negative empirical claim about the surveyed literature: that none applied PSO or ACO to advanced sentence-embedding representations for semantic-similarity document search. No parameter is fitted to data and then renamed as a prediction; no equation in the paper is defined in terms of the paper's own conclusion; and no load-bearing result is justified by a self-citation, since the reference list contains no prior work by Muniyappa or Kim. The main weakness, that the ten-paper sample appears off-topic or unsystematically selected and therefore does not establish the 'none applied' gap, is a correctness and scope risk, not circularity. Because the central claim is an assertion about the external literature rather than a construction that reduces to its own inputs, the circularity score is 0.

Assumptions & free parameters 0 free parameters · 2 assumptions · 0 invented entities

The paper introduces no new parameters or entities; its content is a summary of existing algorithms. The load-bearing assumptions are about selection and faithful reporting of the cited literature, and these are questionable due to scope mismatch and equation errors.

assumptions (2)
  • domain assumption The ten cited papers constitute a relevant and representative sample of swarm intelligence for semantic document search.
    The paper provides no search protocol or inclusion criteria, and several cited works address other tasks (food recommendation, sport clustering, hate speech), so this assumption is doubtful.
  • domain assumption The equations and summaries in Section 2 accurately reproduce the cited papers.
    Multiple displayed equations are malformed (e.g., Eq. (3) with 'w = 1 - w' and Eq. (7) with 'tau_i tau_i'), which undermines confidence in the summaries.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Survey of Swarm Intelligence Approaches to Search Documents Based On Semantic Similarity." pith.science (2026). https://pith.science/paper/GFRFJED2

@misc{pith2026250711787,
  author       = {Pith},
  title        = {Pith review of: Survey of Swarm Intelligence Approaches to Search Documents Based On Semantic Similarity},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GFRFJED2}},
  note         = {Machine review of arXiv:2507.11787}
}
read the original abstract

Swarm Intelligence (SI) is gaining a lot of popularity in artificial intelligence, where the natural behavior of animals and insects is observed and translated into computer algorithms called swarm computing to solve real-world problems. Due to their effectiveness, they are applied in solving various computer optimization problems. This survey will review all the latest developments in Searching for documents based on semantic similarity using Swarm Intelligence algorithms and recommend future research directions.

Figures

Figures reproduced from arXiv: 2507.11787 by the authors.

Figure 1
Figure 1. Sample of Swarm Intelligence Hierarchy 2 Main X.-F. Song et al. [1], applied particle swarm optimization to solve feature selection problem in high-dimensional dataset. The approach has three steps: In the first step Symmetric Uncertainty (SU) method was applied to compute the relevancy score and the values less than the predetermined threshold will be elemenated as irrelevant features. In the second step, features … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 18 canonical work pages

  1. [1]

    A Fast Hybrid Feature Selection Based on Correlation-Guided Clustering and Par- ticle Swarm Optimization for High-Dimensional Data,

    X.-F. Song, Y. Zhang, D.-W. Gong, and X.-Z. Gao, “A Fast Hybrid Feature Selection Based on Correlation-Guided Clustering and Par- ticle Swarm Optimization for High-Dimensional Data, ” IEEE Transactions on Cybernetics, vol. 52, no. 9, pp. 9573–9586, Sep. 2022, doi: https://doi.org/10.1109/TCYB.2021.3061152

    work page arXiv 2022

  2. [3]

    Feature selection with clustering probabilistic particle swarm optimization,

    J. Gao, Z. Wang, Z. Lei, R.-L. Wang, Z. Wu, and S. Gao, “Feature selection with clustering probabilistic particle swarm optimization, ” International journal of machine learning and cybernetics, Mar. 2024, doi: https://doi.org/10.1007/s13042-024-02111-9

    work page doi:10.1007/s13042-024-02111-9 2024

  3. [16]

    Integration of graph clustering with ant colony optimization for feature selection

    P. Moradi, M. Rostami, "Integration of graph clustering with ant colony optimization for feature selection", Knowledge-Based Systems, vol. 84, pp. 144-161, 2015, doi: 10.1016/j.knosys.2015.04.007

    work page doi:10.1016/j.knosys.2015.04.007 2015

  4. [17]

    Londt, X

    T. Londt, X. Gao, B. Xue, Particle Swarm Optimisation based Two-Stage Feature Selection in Text Mining: A Multi-Objective Approach, 2020, doi: 10.13140/RG.2.2.26656.61446

    work page arXiv 2020

  5. [4]

    Rohaidah Ahmada, A

    S. Rohaidah Ahmada, A. Abubakar and M. Ridzwan Yaakubb, Ant colony optimization for text feature selection in sentiment analysis. Intelligent Data Analysis 23. 2019. pp.123-158, doi: 10.3233/IDA-173740

    work page doi:10.3233/ida-173740 2019

  6. [15]

    Textual Feature Extraction Using Ant Colony Optimization for Hate Speech Classification,

    S. Gite et al., “Textual Feature Extraction Using Ant Colony Optimization for Hate Speech Classification, ” Big Data and Cognitive Computing, vol. 7, no. 1, p. 45, Mar. 2023, doi: 10.3390/bdcc7010045

    work page doi:10.3390/bdcc7010045 2023

  7. [6]

    Rehman, O

    F. Rehman, O. Khalid, N. Haq, A. Khan, K. Bilal, and S. Madani, Diet-Right: A Smart Food Recommendation System, ksii transactions on Internet and information systems, 2017, vol. 11 (6)

    work page 2017

  8. [9]

    A clustering mining method for sports behavior characteristics of athletes based on the ant colony optimization,

    D. Yang, J. Wang, J. He, and C. Zhao, “A clustering mining method for sports behavior characteristics of athletes based on the ant colony optimization, ” Heliyon, vol. 10, no. 12, p. e33297, Jun. 2024, doi: https://doi.org/10.1016/j.heliyon.2024.e33297

    work page doi:10.1016/j.heliyon.2024.e33297 2024

  9. [7]

    A Hybrid Mechanism for Auto Text Categorization in Web Documents

    M. Yogi, M. Kalyan, and D. Aiswarya, "A Hybrid Mechanism for Auto Text Categorization in Web Documents", Journal of Soft Computing Paradigm, vol (4), pp. 272-282, doi: 10.36548/jscp.2022.4.006

    work page doi:10.36548/jscp.2022.4.006 2022

  10. [2]

    Bare bones particle swarms,

    J. F. Kennedy, “Bare bones particle swarms, ” Apr. 2003, doi: https://doi.org/10.1109/sis.2003.1202251

    arXiv 2003

Show all 23 references
  1. [5]

    Bezdek, R

    James C. Bezdek, R. Ehrlich, W. Full, FCM: The fuzzy c-means clustering algorithm, Computers & Geosciences, 1984, vol 10, pp. 191-203, doi:10.1016/0098-3004(84)90020-7

    1984 doi

  2. [8]

    Principal Component Analysis (PCA),

    T. Kurita, “Principal Component Analysis (PCA), ” Springer eBooks, pp. 1013–1016, Jan. 2021, doi: https://doi.org/10.1007/978-3-030-63416-2_649

    2021 doi

  3. [10]

    Natural Language Processing Recipes, 2019, https://link.springer.com/book/10.1007/978-1-4842-4267-4

    2019 doi

  4. [11]

    PDBI: A partitioning Davies-Bouldin index for clustering evaluation,

    F. Ros, R. Riad, and S. Guillaume, “PDBI: A partitioning Davies-Bouldin index for clustering evaluation, ” Neurocomputing, vol. 528, pp. 178–199, Apr. 2023, doi: https://doi.org/10.1016/j.neucom.2023.01.043

    2023 doi

  5. [12]

    A K-Means Clustering Algorithm

    J. A. Hartigan and M. A. Wong, "A K-Means Clustering Algorithm", Journal of the Royal Statistical Society. Series C (Applied Statistics), Vol. 28, No. 1, pp. 100-108, 1979, doi: 10.2307/2346830

    1979 doi

  6. [13]

    A STATISTICAL INTERPRETATION OF TERM SPECIFICITY AND ITS APPLICATION IN RETRIEVAL,

    K. SPARCK JONES, “A STATISTICAL INTERPRETATION OF TERM SPECIFICITY AND ITS APPLICATION IN RETRIEVAL, ” Journal of Documenta- tion, vol. 28, no. 1, pp. 11–21, Jan. 1972, doi: https://doi.org/10.1108/eb026526

    1972 doi

  7. [14]

    Is Cosine-Similarity of Embeddings Really About Similarity?,

    H. Steck, C. Ekanadham, and N. Kallus, “Is Cosine-Similarity of Embeddings Really About Similarity?, ” arXiv (Cornell University), Mar. 2024, doi: https://doi.org/10.1145/3589335.3651526

    2024

  8. [18]

    Feature selection using importance-based two-stage multi-modal multiobjective particle swarm optimization,

    Q. Ling, W. Liu, F. Han, J. Shi, A. A. Hussein, and B. S. Sayway, “Feature selection using importance-based two-stage multi-modal multiobjective particle swarm optimization, ” Cluster Computing, vol. 28, no. 2, Nov. 2024, doi: https://doi.org/10.1007/s10586-024-04807-7

    2024 doi

  9. [19]

    Grabowski, S., & Kowalski, T. M. (2021). Algorithms for all-pairs Hamming distance based similarity. Software: Practice and Experience, 51(7), 1580–1590. https://doi.org/10.1002/spe.2978

    2021 doi

  10. [20]

    DDoS Detection by Using Information Gain-Naïve Bayes,

    S. Mandala, A. I. Ramadhan, M. Rosalinda, W. M. S. Yafooz and R. H. Khohar, "DDoS Detection by Using Information Gain-Naïve Bayes, " 2022 2nd International Conference on Intelligent Cybernetics Technology & Applications (ICICyTA), Bandung, Indonesia, 2022, pp. 283-288, doi: 10...

    2022

  11. [21]

    Evaluation of feature selection using information gain and gain ratio on bank marketing classification using naïve bayes

    B. Prasetiyo1, Alamsyah, M. A. Muslim,N. Baroroh, "Evaluation of feature selection using information gain and gain ratio on bank marketing classification using naïve bayes", Journal of Physics: Conference Series, 2021, doi:10.1088/1742-6596/1918/4/042153

    2021 doi

  12. [22]

    Correlation-based feature selection using bio-inspired algorithms and optimized KELM classifier for glaucoma diagnosis,

    K. Balasubramanian and A. N.P., “Correlation-based feature selection using bio-inspired algorithms and optimized KELM classifier for glaucoma diagnosis, ” Applied Soft Computing, vol. 128, p. 109432, Oct. 2022, doi: https://doi.org/10.1016/j.asoc.2022.109432

    2022

  13. [23]

    A tutorial-based survey on feature selection: Recent advancements on feature selection,

    A. Moslemi, “A tutorial-based survey on feature selection: Recent advancements on feature selection, ” Engineering Applications of Artificial Intelligence, vol. 126, p. 107136, Nov. 2023, doi: https://doi.org/10.1016/j.engappai.2023.107136. Manuscript submitted to ACM

    2023

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.