Pith. sign in

REVIEW 2 major objections 4 minor 48 references

Oversampling a rare text class should grow its vocabulary, and a Markov chain that borrows word transitions from the majority class tops balanced accuracy on severely imbalanced text data.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

EMCO oversamples minority text by estimating word-transition probabilities from both minority and majority documents, expanding the synthetic minority vocabulary.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection EMCO is a real, reproducible oversampling method, but the paper's 'every row' claim needs significance testing before it can be taken at face value. the 2 major comments →

arxiv 2509.02332 v1 pith:EYYQ5HGS submitted 2025-09-02 cs.LG

Extrapolated Markov Chain Oversampling Method for Imbalanced Text Classification

classification cs.LG
keywords imbalanced text classificationsynthetic oversamplingMarkov chainfeature space growthHeaps' lawbalanced accuracyminority classbag-of-words
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

The paper's central claim is that synthetic oversampling of a rare text category should not be confined to the words that happen to appear in the minority training documents. Because real text vocabularies grow with sample size (Heaps' law), a synthetic minority sample should also grow; the paper proposes EMCO, a first-order Markov chain over words whose transition probabilities are estimated from the minority class plus, with weight gamma, from majority documents at words that also occur in the minority class. This lets generated minority documents contain words never seen in the minority training set. In experiments with linear SVM on Reuters, HuffPost, and 20 Newsgroups, EMCO with gamma=1 reports the best balanced accuracy in very low frequency categories on every table row except one, where gamma=0.1 is best, and it also improves recall and F2 scores. A practical virtue is that EMCO requires no external resources such as pretrained embeddings or synonym dictionaries, and the code is released.

Core claim

EMCO models each minority document as a path in a finite Markov chain whose states are the training vocabulary plus a stop token. The transition matrix has a block structure: from minority-vocabulary states, transition weights are the minority transition counts plus gamma times majority counts for the same word pair; from majority-only states, the chain returns to minority words according to the minority marginal word distribution; and starts and ends are taken from minority documents. Self-transitions are zeroed. Sampling a synthetic document means drawing a length, then drawing words row by row from this matrix. Because majority transitions are included only for pairs whose first word is i

What carries the argument

The central object is the gamma-weighted word-transition matrix with three blocks: minority-to-all transitions (minority counts plus gamma-weighted majority counts), majority-only-to-minority returns (minority marginal distribution), and stop-token start/end rows from minority documents. The mechanism that carries the argument is treating topic and sequence as partly independent: any word common to both classes can act as a bridge, so majority-only words that follow that bridge word in majority documents are treated as plausible next words in synthetic minority documents. The scalar gamma controls how much extrapolation the sampler is allowed and, in the experiments, trades recall against pr

Load-bearing premise

The load-bearing premise is that word-order patterns are similar enough across topics that a majority-only word which follows a shared word in majority documents is a plausible continuation of a minority document; if this fails, EMCO injects off-topic words and degrades precision at high gamma.

What would settle it

Build a binary text corpus in which the minority and majority classes share only function words but use disjoint topical nouns, oversample the minority with EMCO, and count how many introduced majority-only words belong to the majority's topic lexicon. If removing those off-topic words does not change balanced accuracy, the topic-sequence independence premise is not doing the work; if the synthetic minority vocabulary contains words that no genuinely new minority document would contain, the extrapolation premise fails.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • On severely imbalanced multiclass text tasks in the tested document-embedding setting, EMCO can be the best available oversampling choice when the evaluation target is balanced accuracy or recall-heavy F2.
  • The gamma hyperparameter gives a direct, interpretable control over the recall-precision trade-off; users who need precision can lower gamma, while users who need recall or balanced accuracy can raise it to about one.
  • The method works without pretrained word vectors, synonym dictionaries, or LLM generation, making it usable for low-resource languages and private or small datasets.
  • Feature-space growth under EMCO tracks Heaps' law better than EDA or DECOM in the reported Reuters example, suggesting the method reproduces a natural vocabulary-growth dynamic rather than arbitrary expansion.
  • Because the sampling happens before vectorization, EMCO can be combined with any downstream classifier; the paper also reports improved balanced accuracy with a bidirectional LSTM over no oversampling for very low frequency Reuters headline categories.

Where Pith is reading between the lines

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

  • A testable extension is to replace the first-order chain with an n-gram or learned language model that still restricts extrapolation to majority transitions from shared words; the same bridge idea should transfer.
  • The gamma choice could be set by comparing synthetic minority vocabulary growth against Heaps' law fitted on the training corpus rather than by grid search; in one reported example gamma=0.1 matched Heaps' law closely.
  • If topic and sequence are not sufficiently independent—for example, if the majority-only words following a shared word are strongly tied to majority topics—EMCO will inject off-topic vocabulary and its balanced-accuracy advantage should shrink; a held-out study comparing human ratings of synthetic documents across domains would expose the boundary.
  • The comparison relies on tf-idf bag-of-words and linear SVM; whether the balanced-accuracy ranking transfers to transformer-based classifiers trained from scratch is untested by the paper.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper proposes EMCO, a Markov-chain oversampling method for imbalanced text classification. The transition matrix is estimated from minority-class documents and augmented with transitions observed in majority-class documents from minority-vocabulary words to majority-only words, with a weight parameter γ. Synthetic minority documents are sampled from the resulting chain, allowing the minority feature space to expand beyond the convex hull of the minority training sample. The method is compared with ROS, SMOTE, ADASYN, DRO, DECOM, EDA, MCO (EMCO with γ=0), and a no-oversampling SVM baseline on Reuters-21578, HuffPost, and 20 Newsgroups data, with an additional LSTM experiment on Reuters headlines. The headline empirical claim is that EMCO with γ=1 achieves the best balanced accuracy in very low frequency categories on essentially every row of Table 3, at the cost of lower precision than several competitors. Section 7 reports vocabulary-growth analyses, including a comparison of synthetic vocabulary growth with Heaps' law.

Significance. If the empirical claim holds, EMCO is a practically useful, language-independent oversampling method for severely imbalanced text classification in small-data regimes. The method is clearly specified, the code is released, and the construction is not circular: γ is fixed before evaluation and the Heaps' law material is an external reference, not an input to the method. The paper also honestly identifies the recall-precision tradeoff controlled by γ and the scalability limitations of a vocabulary-squared transition matrix. However, the central empirical claim currently rests on point estimates without error bars or significance tests, and the extrapolation assumption is only indirectly validated. The method is therefore promising, but the evidence as reported does not yet support the 'every row' claim at the strength asserted.

major comments (2)
  1. [Section 6, Table 3] The headline claim that 'EMCO with γ=1 outperforms the other approaches in very low frequency categories on every row, except for one' is not supported by the evidence as reported. Reuters and 20 Newsgroups results are averages of five repetitions per category; HuffPost Titles rows without an asterisk come from a single random split; no standard deviations, confidence intervals, or paired significance tests are given anywhere. Several decisive margins are small (e.g., HuffPost Titles* at 10%: .638 vs .630; Reuters Full at 10%: .801 vs .791; several rows differ by .001–.002), and macro-averaged balanced accuracy on very low frequency categories is high-variance because positive test sets are tiny. The deterministic reading of point estimates is therefore not secure. Please add per-category variances or confidence intervals, use repeated or nested splits for all data sets, and report paire
  2. [Section 4.1, with Section 7 and Appendix B] The central extrapolation assumption—that majority-only words following minority vocabulary words in majority documents are reasonable additions to the synthetic minority vocabulary—is only indirectly validated. Section 7's binary classification of synthetic vocabulary is a useful start, but it reports only macro-averaged recall, TNR, and balanced accuracy, not precision per category, so the degree of off-topic injection is not directly quantified. Appendix B shows that at γ=1 precision drops sharply (e.g., Reuters Titles 10%: EMCO .212 vs SMOTE .508) while recall rises; the balanced-accuracy advantage is thus partly a re-labeling of the recall-precision tradeoff. A direct diagnostic of the generated vocabulary—for example, precision for predicting held-out minority test words per category, or a topic-coherence sample of generated words—would substantiate the assumption. As the paper its
minor comments (4)
  1. [Section 7, Figure 3] The Heaps' law comparison is qualitative: parameters k≈63 and θ≈0.378 are fitted on the same training data, with no fitting procedure, fit diagnostics, or uncertainty reported. Please state how the parameters were estimated and treat the match as illustrative rather than as a formal validation.
  2. [Section 5] For reproducibility, please report random seeds for the five repetitions and for the random train-test splits. The code release is a strength, but seed information is necessary for exact replication and for others to compute standard errors.
  3. [Throughout] There are several typographical errors and inconsistencies, e.g., 'exaclty' in Section 5, 'also form the majority class' in Section 8, and inconsistent use of the † and ∗ table symbols. These should be cleaned up.
  4. [Algorithm 1] Please clarify in the text or pseudocode that the loop keeps drawing tokens until the document contains the drawn number of non-<stop> tokens, and that <stop> is only a control token not included in the final document. The current pseudocode can be misread as allowing documents shorter than the drawn length.

Circularity Check

0 steps flagged

No circularity: EMCO is defined from co-occurrence counts, not from the target metric; the only self-citation is peripheral.

full rationale

The EMCO construction is self-contained. Section 4.2 defines the transition matrix from word co-occurrence counts in minority and majority documents and Algorithm 1 samples from it; balanced accuracy, F-scores, and classifier outputs never appear in the estimator. Gamma is a fixed hyperparameter (0.1 or 1, plus 0 for MCO) and is not fitted to the test results or to the balanced-accuracy ranking. The Heaps' law comparison in Section 7 is a descriptive reference: the law is fitted to the full training corpus and used only to interpret vocabulary-growth curves; it is not an input to EMCO and not the basis for the classification claims. The paper's topic/sequence independence assumption is an explicit modeling assumption and could be empirically wrong, but a false assumption is not circularity. The absence of error bars or significance tests in Tables 3-5 is a statistical-robustness concern, not a circularity concern. The only self-citation (Avela 2024, p. 12) supports the choice of balanced accuracy as an evaluation metric and is accompanied by an independent survey (Henning et al. 2023); it does not carry the derivation. No load-bearing step reduces to its own input, so no circularity is found.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 0 invented entities

The paper introduces no new physical or theoretical entities; the proposed constructs (minority vocabulary, majority-only vocabulary, gamma-weighted transition matrix) are procedural definitions over the training data. The main free parameters are gamma and the Heaps' law reference constants.

free parameters (3)
  • gamma (γ) = 1, 0.1, 0.01 (tested, not tuned per dataset)
    Weight for majority-class transitions. The paper tests a few values and shows the effect on recall/precision but gives no selection rule, so it is a hand-chosen hyperparameter used in all experiments.
  • Heaps' law k and theta = k ≈ 63, theta ≈ 0.378 (Reuters training set)
    Fitted to the full Reuters training data in Section 7 for vocabulary-growth validation. They are reference constants, not inputs to EMCO, but they are fitted numbers that the validation narrative depends on.
  • LSTM epochs, dropout, hidden size = 3 epochs, 0.4 dropout, 32 hidden
    Fixed hyperparameters for the secondary word-embedding experiment; not central to EMCO.
axioms (4)
  • ad hoc to paper Sequential structure of text is sufficiently independent of topic so that majority-only words that follow minority words in majority documents are reasonable additions to synthetic minority documents.
    Section 4.1. This is the load-bearing premise for EMCO's extrapolation. If false, generated documents drift off-topic and precision drops, which the paper observes at high gamma.
  • domain assumption An order-1 Markov chain adequately captures word transitions for oversampling purposes.
    Section 4.2 and Algorithm 1. Long-range dependencies are ignored; the authors acknowledge this in the conclusion.
  • domain assumption Heaps' law describes the growth of vocabulary with sample size in the evaluated corpora.
    Sections 4.1 and 7. Used as motivation and as a reference to validate EMCO's vocabulary growth.
  • domain assumption Bag-of-words with tf-idf is a suitable representation for the classification tasks.
    Section 5. The main experiments use linear SVM on tf-idf vectors; EMCO generates documents that are subsequently embedded in this space.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Extrapolated Markov Chain Oversampling Method for Imbalanced Text Classification." pith.science (2026). https://pith.science/paper/EYYQ5HGS

@misc{pith2026250902332,
  author       = {Pith},
  title        = {Pith review of: Extrapolated Markov Chain Oversampling Method for Imbalanced Text Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EYYQ5HGS}},
  note         = {Machine review of arXiv:2509.02332}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Text classification is the task of automatically assigning text documents correct labels from a predefined set of categories. In real-life (text) classification tasks, observations and misclassification costs are often unevenly distributed between the classes - known as the problem of imbalanced data. Synthetic oversampling is a popular approach to imbalanced classification. The idea is to generate synthetic observations in the minority class to balance the classes in the training set. Many general-purpose oversampling methods can be applied to text data; however, imbalanced text data poses a number of distinctive difficulties that stem from the unique nature of text compared to other domains. One such factor is that when the sample size of text increases, the sample vocabulary (i.e., feature space) is likely to grow as well. We introduce a novel Markov chain based text oversampling method. The transition probabilities are estimated from the minority class but also partly from the majority class, thus allowing the minority feature space to expand in oversampling. We evaluate our approach against prominent oversampling methods and show that our approach is able to produce highly competitive results against the other methods in several real data examples, especially when the imbalance is severe.

Figures

Figures reproduced from arXiv: 2509.02332 by Aleksi Avela, Pauliina Ilmonen.

Figure 1
Figure 1. Figure 1: The feature space of the “trade” category headline [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The effect of EMCO’s γ-hyperparameter on recall, tnr, and precision with Reuters data set based on averages of five repetitions per category with a sampling ratio of 20%. Note that the values on the horizontal axis are not distributed linearly. Even small values of γ seem to improve the classification performance with respect to recall and balanced accuracy compared to γ = 0. The optimal choice of γ depend… view at source ↗
Figure 3
Figure 3. Figure 3: Synthetic vocabulary growth with different sampl [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

48 extracted references · 44 canonical work pages · 2 internal anchors

  1. [1]

    and Elmaghraby, A

    Abdelwahab, O. and Elmaghraby, A. (2018). Deep learning bas ed vs. Markov chain based text generation for cross domain adaptation for sentiment classification. In 2018 IEEE International Conference on Information Reuse an d Integration (IRI), pages 252–255

  2. [2]

    Avela, A. (2024). On Fβ-score and cost-consistency in evaluation of imbalanced cl assification. In Proceedings of the 2024 European Symposium on Artificial Neural Networks, C omputational Intelligence and Machine Learning , ESANN, pages 245—-250

  3. [3]

    A., and Reuter, C

    Bayer, M., Kaufhold, M. A., and Reuter, C. (2022). A survey on data augmentation for text classification. ACM Computing Surveys, 55(7):146

  4. [4]

    Bird, S., Klein, E., and Loper, E. (2009). Natural Language Processing with Python . O’Reilly Media, Inc

  5. [5]

    Bishop, C. M. (2006). Pattern Recognition and Machine Learning . Springer New Y ork, NY

  6. [6]

    M., Ng, A

    Blei, D. M., Ng, A. Y ., and Jordan, M. I. (2003). Latent Dirich let Allocation. Journal of Machine Learning Research , 3:993–1022

  7. [7]

    R., Vilnis, L., Vinyals, O., Dai, A

    Bowman, S. R., Vilnis, L., Vinyals, O., Dai, A. M., Jozefowic z, R., and Bengio, S. (2015). Generating sentences from a continuous space. arXiv preprint arXiv:1511.06349

  8. [8]

    V ., Bowyer, K

    Chawla, N. V ., Bowyer, K. W ., Hall, L. O., and Kegelmeyer, W . P. (2002). SMOTE: Synthetic minority over-sampling technique. Journal of Artificial Intelligence Research , 16:321–357

  9. [9]

    Chen, E., Lin, Y ., Xiong, H., Luo, Q., and Ma, H. (2011). Explo iting probabilistic topic models to improve text categorization under class imbalance. Information Processing & Management , 47(2):202–214

  10. [10]

    Cloutier, N. A. and Japkowicz, N. (2023). Fine-tuned genera tive LLM oversampling can improve performance over traditional techniques on multiclass imbalanced text clas sification. In 2023 IEEE International Conference on Big Data (BigData), pages 5181–5186

  11. [11]

    C., and Cook, D

    Das, B., Krishnan, N. C., and Cook, D. J. (2015). RACOG and wRA COG: Two probabilistic oversampling techniques. IEEE Transactions on Knowledge and Data Engineering , 27(1):222–234

  12. [12]

    Deng, Y ., Wu, M., and Ma, Y . (2024). AGO-FT: An adaptive guide d oversampling based on fast space division and trustworthy sampling space for imbalanced noisy datasets. In 2024 IEEE International Conference on Big Data (BigData), pages 529–538

  13. [13]

    B., Wang, C., Gao, J., and Paisley, J

    Dieng, A. B., Wang, C., Gao, J., and Paisley, J. (2016). Topic RNN: A recurrent neural network with long-range semantic dependency. arXiv preprint arXiv:1611.01702

  14. [14]

    and Bacao, F

    Douzas, G. and Bacao, F. (2019). Geometric SMOTE a geometric ally enhanced drop-in replacement for SMOTE. Information Sciences, 501:118–135

  15. [15]

    Egghe, L. (2007). Untangling Herdan’s law and Heaps’ law: Ma thematical and informetric arguments. Journal of the American Society for Information Science and T echnology, 58(5):702—-709

  16. [16]

    Elkan, C. (2001). The foundations of cost-sensitive learning. In Proceedings of the 17th International Joint Conference on Artificial Intelligence , IJCAI’01, pages 973–978

  17. [17]

    E., Chang, K

    Fan, R. E., Chang, K. W ., Hsieh, C. J., Wang, X. R., and Lin, C. J . (2008). LIBLINEAR: A library for large linear classification. Journal of Machine Learning Research , 9:1871–1874

  18. [18]

    Haixiang, G., Yijing, L., Shang, J., Mingyun, G., Y uanyue, H ., and Bing, G. (2017). Learning from class-imbalanced data: Review of methods and applications. Expert Systems with Applications , 73:220–239

  19. [19]

    Y ., and Mao, B

    Han, H., Wang, W . Y ., and Mao, B. H. (2005). Borderline-SMOTE : a new over-sampling method in imbalanced data sets learning. In Advances in Intelligent Computing , ICIC’05, pages 878—-887

  20. [20]

    Harris, Z. S. (1954). Distributional structure. WORD, 10(2–3):146–162

  21. [21]

    A., and Li, S

    He, H., Bai, Y ., Garcia, E. A., and Li, S. (2008). ADASYN: Adap tive synthetic sampling approach for imbalanced learning. In 2008 IEEE International Joint Conference on Neural Network s (IEEE W orld Congress on Computa- tional Intelligence), pages 1322–1328

  22. [22]

    and Garcia, E

    He, H. and Garcia, E. A. (2009). Learning from imbalanced dat a. IEEE Transactions on Knowledge and Data Engineering, 21(9):1263–1284

  23. [23]

    Heaps, H. S. (1978). Information Retrieval: Computational and Theoretical Asp ects. Academic Press, Inc

  24. [24]

    Henning, S., Beluch, W ., Fraser, A., and Friedrich, A. (2023 ). A survey of methods for addressing class imbalance in deep-learning based natural language processing. arXiv preprint arXiv:2210.04675 . 18

  25. [25]

    Herdan, G. (1964). Quantitative Linguistics. London: Butterworths

  26. [26]

    and Schmidhuber, J

    Hochreiter, S. and Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8):1735—-1780

  27. [27]

    Japkowicz, N. (2000). The class imbalance problem: Signific ance and strategies. In Proceedings of the 2000 Interna- tional Conference on Artificial Intelligence , ICAI, pages 111–117

  28. [28]

    and Japkowicz, N

    Jo, T. and Japkowicz, N. (2004). Class imbalances versus sma ll disjuncts. ACM SIGKDD Explorations Newsletter , 6(1):40–49

  29. [29]

    Joachims, T. (2002). Learning to Classify T ext Using Support V ector Machines. Springer New Y ork, NY . Köknar-Tezel, S. and Latecki, L. J. (2009). Improving SVM cl assification on imbalanced data sets in distance spaces. In 2009 9th IEEE International Conference on Data Mining , pages 259–267. Lemaître, G., Nogueira, F., and Aridas, C. K. (2017). Imbala ...

  30. [30]

    Luo, Y ., Feng, H., Weng, X., Huang, K., and Zheng, H. (2019). A novel oversampling method based on SeqGAN for imbalanced text classification. In Proceedings of 2019 IEEE International Conference on Big Da ta, pages 2891–2894. López, V ., Fernández, A., Moreno-Torres, J. G., and Herrera , F. (2012). Analysis of preprocessing vs. cost-sensitive learning for...

  31. [31]

    Mikolov, T., Grave, E., Bojanowski, P ., Puhrsch, C., and Joulin, A. (2018). Advances in pre-training distributed word representations. In Proceedings of the International Conference on Language Resources and Evaluation, LREC’18

  32. [32]

    Misra, R. (2022). News category dataset. arXiv preprint arXiv:2209.11429

  33. [33]

    Moreo, A., Esuli, A., and Sebastiani, F. (2016). Distributi onal random oversampling for imbalanced text classification. In Proceedings of the 39th International ACM SIGIR Conference on Research and Development in Information

  34. [34]

    Nakada, R., Xu, Y ., Li, L., and Zhang, L. (2024). Synthetic ov ersampling: Theory and a practical approach using LLMs to address data imbalance. arXiv preprint arXiv:2406.03628

  35. [35]

    K., Thrun, S., and Mitchell, T

    Nigam, K., McCallum, A. K., Thrun, S., and Mitchell, T. (2000 ). Text classification from labeled and unlabeled documents using EM. Machine Learning, 39:103–134

  36. [36]

    , Brucher, M., Perrot, M., and Duchesnay, E

    Pedregosa, F., V aroquaux, G., Gramfort, A., Michel, V ., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P ., Weiss, R., Dubourg, V ., V anderplas, J., Passos, A., Cournapeau, D. , Brucher, M., Perrot, M., and Duchesnay, E. (2011). Scikit-learn: Machine learning in Python. Journal of Machine Learning Research , 12:2825–2830

  37. [37]

    Piyadasa, T. D. and Gunawardana, K. (2023). A review on overs ampling techniques for solving the data imbalance problem in classification. The International Journal on Advances in ICT for Emerging Re gions, 16(1)

  38. [38]

    Porter, M. F. (1980). An algorithm for suffix stripping. Program: electronic library and information systems , 14(3):130–137

  39. [39]

    D., Shih, L., Teevan, J., and Karger, D

    Rennie, J. D., Shih, L., Teevan, J., and Karger, D. R. (2003). Tackling the poor assumptions of naive Bayes text classifiers. In Proceedings of the 20th International Conference on Machin e Learning, ICML’03, pages 616–623

  40. [40]

    and Buckley, C

    Salton, G. and Buckley, C. (1988). Term-weighting approach es in automatic text retrieval. Information Processing & Management, 24(5):513–523

  41. [41]

    Sano, Y ., Takayasu, H., and Takayasu, M. (2012). Zipf’s law a nd Heaps’ law can predict the size of potential words. Progress of Theoretical Physics Supplement , 194:202—-209

  42. [42]

    M., and Furht, B

    Shorten, C., Khoshgoftaar, T. M., and Furht, B. (2021). Text data augmentation for deep learning. Journal of Big Data, 8:101. V eropoulos, K., Campbell, C., and Cristianini, N. (1999). C ontrolling the sensitivity of support vector machines. In Proceedings of the International Joint Conference on AI , pages 55–60

  43. [43]

    Wang, W ., Gan, Z., Xu, H., Zhang, R., Wang, G., Shen, D., Chen, C., and Carin, L. (2019). Topic-guided variational autoencoders for text generation. arXiv preprint arXiv:1903.07137

  44. [44]

    and Zou, K

    Wei, J. and Zou, K. (2019). EDA: Easy data augmentation techn iques for boosting performance on text classification tasks. In Proceedings of the 2019 Conference on Empirical Methods in N atural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 6383–6389. Association for Computational Linguistics. 19

  45. [45]

    Weiss, G. M. (2004). Mining with rarity: a unifying framewor k. ACM SIGKDD Explorations Newsletter, 6(1):7–19

  46. [46]

    M., McCarthy, K., and Zabar, B

    Weiss, G. M., McCarthy, K., and Zabar, B. (2007). Cost-sensi tive learning vs. sampling: Which is best for handling unbalanced classes with unequal error costs? In Proceedings of 2007 International Conference on Data Minin g, DMIN’07

  47. [47]

    Yin, J., Gan, C., Zhao, K., Lin, X., Quan, Z., and Wang, Z.-J. (2020). A novel model for imbalanced data classification. In Proceedings of the AAAI Conference on Artificial Intelligen ce, 34(04), pages 6680–6687

  48. [48]

    and Elkan, C

    Zadrozny, B. and Elkan, C. (2001). Obtaining calibrated pro bability estimates from decision trees and naive Bayesian classifiers. In Proceedings of the 18th International Conference on Machin e Learning, ICML’01, pages 609–616. 20

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.