Pith. sign in

REVIEW 3 major objections 6 minor 51 references

Enhancing Topic Interpretability for Neural Topic Modeling through Topic-wise Contrastive Learning

T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A topic-wise contrastive regularizer makes neural topic models produce more coherent, more distinct topics.

desk verdict A genuinely new topic-wise contrastive regularizer that likely improves topic coherence and diversity, though the headline cross-corpus interpretability claim outruns the evidence. read the letter →

arxiv 2412.17338 v1 pith:VESRXA6U submitted 2024-12-23 cs.AI

Xin Gao , Yang Lin , Ruiqing Li , Yasha Wang , Xu Chu , Xinyu Ma , Hailong Yu This is my paper · ORCID
classification cs.AI
keywords neuraltopicmodelsinterpretabilitycontrastivelearningcoherencediversityNPMIGumbel-Softmaxregularization
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

Neural topic models are typically trained to maximize the likelihood of observed text, but that objective does not guarantee that the discovered topics make sense to a person. This paper argues that topic interpretability should be optimized directly, and proposes ContraTopic, a regularizer that measures interpretability during training. The regularizer treats words drawn from the same topic as positive samples and words from different topics as negative samples, with word similarity scored by precomputed NPMI. Training with this term alongside the usual reconstruction and KL terms yields topics that the paper reports are more coherent and more diverse than those from ten baseline models across 20NG, Yahoo, and NYTimes. A human word intrusion study on 20NG gives ContraTopic the highest score, supporting the claim that the regularizer improves interpretability as people perceive it.

What carries the argument

The load-bearing component is the topic-wise contrastive regularizer Lcon defined by Eq. (2): a supervised contrastive loss whose samples are words, not documents. Positive pairs are words sampled from the same topic; negative pairs are words from different topics; and the similarity kernel K(·) is the precomputed NPMI matrix over the vocabulary. To keep the sampling differentiable, the paper uses a Gumbel-Softmax reparameterization with a relaxed subset-sampling procedure (Eqs. (3)-(5)) that draws top-v words per topic without replacement. This mechanism lets gradients flow from an interpretability-style objective directly into the topic-word distribution, which is what allows the model to trade likelihood for more human-readable topics.

What would settle it

Run the paper's word intrusion study on Yahoo or NYTimes rather than only 20NG. If ContraTopic's word intrusion scores do not exceed the baselines on those corpora, or if a corpus is constructed where high-NPMI word sets are not semantically coherent to annotators, the claim that optimizing NPMI via topic-wise contrastive learning robustly improves human interpretability would be refuted.

Watch

Extended reading notes

Core claim

ContraTopic's central claim is that topic interpretability can be decomposed into internal coherence and external distinctiveness and steered directly by a contrastive loss on words, not documents. For each topic, the model samples its top-v words using a relaxed Gumbel-Softmax subset sampler, then applies a supervised contrastive loss where words from the same topic are positive pairs and words from different topics are negative pairs. The similarity function K(·) is the precomputed NPMI between word pairs, which the paper adopts because NPMI is a standard coherence metric and because maximizing in-topic NPMI while minimizing cross-topic NPMI is equivalent to maximizing mutual information within topics and minimizing it between topics. The full objective is Ltr = Lrec + Lkl + λLcon, and the paper reports that this regularizer consistently improves topic coherence and topic diversity over ETM, ProdLDA, WLDA, NSTM, WeTe, NTM-R, VTMRL, and CLNTM on three datasets.

Load-bearing premise

The argument rests on treating NPMI scores computed on the training corpus as a reliable, corpus-independent proxy for what humans find coherent; if NPMI and human judgment diverge on a given corpus, optimizing it could raise reported coherence without improving real interpretability.

Editorial extensions

If this is right

  • Adding the ContraTopic regularizer to an NTM is a plug-in way to improve topic coherence and diversity without changing the backbone architecture; the paper shows it works with ETM, WLDA, and WeTe as backbones.
  • Because the regularizer couples to word similarity via NPMI, replacing the similarity kernel (e.g., with embeddings or LLM-based scores) is a direct lever for steering topic quality toward different notions of interpretability.
  • The ablation results imply that both positive and negative pairs contribute: dropping negative pairs costs about 5% and dropping positive pairs about 12% on the reported metrics, so the method's benefit depends on the full topic-wise contrastive structure.

Reading between the lines

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

  • If NPMI-based coherence is a bottleneck, one could swap in a learned or LLM-based similarity function that still admits the same contrastive loss, potentially improving alignment with human judgments on corpora where NPMI is weak.
  • The topic-wise regularizer could be combined with document-wise contrastive objectives in a unified multi-level framework, possibly improving document representation quality without sacrificing topic interpretability.
  • The metric-overfitting danger suggests testing on held-out corpora with a different distribution than the training set; the paper already evaluates coherence on unseen test data, but a cross-corpus transfer test would sharpen the claim.
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 / 6 minor

Summary. The paper proposes ContraTopic, a regularizer for neural topic models that augments the ELBO of an ETM-style backbone with a topic-wise contrastive loss. The regularizer samples v words per topic using a relaxed Gumbel subset sampler, treats words from the same topic as positive pairs and words from different topics as negative pairs, and uses precomputed NPMI scores as the similarity function K(·). The authors report improvements in average NPMI topic coherence and topic diversity over nine baselines on 20NG, Yahoo, and NYTimes, supplemented by ablations, sensitivity analyses, backbone-substitution experiments, a word-intrusion study on 20NG, and qualitative topic lists.

Significance. If the reported improvements are genuine and independently confirmed, ContraTopic would be a simple and useful plug-in: it adds one regularizer to an existing neural topic model, requires no human labels, and improves both coherence and diversity in the reported settings. The paper has real strengths: Table II ablates the positive/negative terms and the similarity function, Figure 6 tests the regularizer across different backbone models, the sensitivity analysis is informative, and Section VI candidly discusses limitations. The main reservations concern the alignment between the optimized objective and the evaluation metric, and the statistical support for the headline claims.

major comments (3)
  1. [Section IV.A, Eq. (2), and Section V.B] The regularizer's similarity function K(·) is the precomputed NPMI, and the main topic-coherence evaluation is also mean NPMI on the test set. The coherence gains in Figure 2 are therefore, to a substantial degree, the expected consequence of directly optimizing a smoothed surrogate of the evaluation statistic; they do not by themselves establish improved interpretability. The word-intrusion study in Section V.J is the only independent evidence, and it covers only 20NG. Section VI further concedes that precomputed NPMI is only a proxy for human interpretability. I would ask for either (i) human or otherwise independent topic-quality evaluations on at least Yahoo and NYTimes, or a clear reframing of the claim as 'higher NPMI and higher unique-word diversity' rather than 'higher interpretability'; and (ii) an analysis of how much of the NPMI gain persists when K(·) is computed on a held-out corpus or with a different coherence estimator.
  2. [Section V.F, Figure 2, and Section V.J, Table III] The abstract and Section VII claim that ContraTopic 'consistently produces topics with significantly higher levels of interpretability,' but Figure 2 omits error bars 'for brevity' and Table III reports no variance or significance test. Three random-seed runs with only means reported cannot support the word 'significantly.' The word-intrusion result (ContraTopic 0.80 vs. the best baseline 0.68) needs at least a confidence interval or a paired significance test (e.g., over topics or over participants) before it can be used as evidence for the central claim.
  3. [Section V.C and Section V.D] Hyperparameter selection is not fully controlled. ContraTopic's λ, v, τβ, and τg are tuned on a validation split, while baselines are said to use 'best reported settings' or official implementations. Because λ is dataset-specific (40, 40, and 300 for 20NG, Yahoo, and NYTimes), it is unclear how much of the reported improvement comes from the regularizer itself rather than from additional tuning on the target datasets. Reporting validation-tuned results for the baselines, or at least a sensitivity range for key baselines, would make the comparison fairer.
minor comments (6)
  1. [Section IV.B, Eq. (3)] The text repeatedly says 'non-differential' where 'non-differentiable' is meant; please correct this throughout the methodology.
  2. [Section V.D] The sentence listing tuned hyperparameters says 'λ, v, τg, and τg'; the second τg should be τβ.
  3. [Section V.G] The heading 'The sensitive analysis' should read 'The sensitivity analysis.'
  4. [Section IV.A and elsewhere] The spelling 'Contratopic' and 'ContraTopic' is inconsistent; please unify to one form.
  5. [Section V.J] The questionnaire description should state whether all 300 questions were answered by each participant in a fixed order or a randomized order, and should report per-participant variability in addition to the mean WIS.
  6. [Section V.E and Eq. (2)] The similarity function K(·) in Eq. (2) has no temperature or scaling parameter; since K is an NPMI value, please clarify whether it is used directly or scaled, because the scale interacts with the choice of λ.

Circularity Check

1 steps flagged · score 6.0 of 10

ContraTopic's main coherence result is a directly optimized NPMI objective evaluated out-of-sample; human validation covers only 20NG, so the cross-corpus interpretability claim is only partially independent.

  1. fitted input called prediction [Section IV.A (Eq. 2); Section V.B; Section V.D]
    "By employing the pre-computed NPMI in the corpus for K(·), the positive samples in our regularizer serve to optimize topic coherence directly during training... Topic coherence measures the average NPMI over the topic KT Cwords of the selected topics. ... we implement the function K(·) with the pre-computed NPMI scores on the training set. Therefore, we evaluate the topic coherence on the unseen test data to make fair comparisons."

    The central interpretability claim is primarily evidenced by topic coherence, defined as average NPMI. The regularizer K(·) is the same precomputed NPMI statistic: positive pairs are rewarded for high within-topic NPMI and negative pairs for low cross-topic NPMI. Consequently, the reported test-set NPMI improvement is the expected out-of-sample consequence of directly optimizing a differentiable relaxation of the evaluation metric, rather than an independent confirmation of human interpretability. The word-intrusion study (Table III) is a genuine external check, but it is limited to 20NG with 20 participants and no significance tests or error bars; the Yahoo and NYTimes interpretability claims rest on the optimized NPMI/diversity metrics.

full rationale

Scores 6 rather than 8 because the paper does include one independent human word-intrusion evaluation on 20NG, and the diversity metric is not directly optimized under the same definition. However, the paper's strongest claim—consistent, significant interpretability gains on all three datasets—is underwritten by NPMI topic coherence, and the regularizer explicitly uses NPMI as its similarity function, so the main quantitative result reduces largely to optimizing the metric being reported. The cited correlation study [40] and the word-intrusion study are external evidence, but neither covers all three corpora at the strength needed to make the NPMI-only results independent. The self-citation [37] appears only in related-work context for seed-word topic models and is not load-bearing. No uniqueness-theorem or ansatz-smuggling circularity is present.

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

The central claim depends on a few tuned hyperparameters and on the assumption that NPMI computed on the training corpus is a meaningful coherence surrogate. The relaxed sampling trick is inherited from prior work. No new entities are postulated. The most consequential choice is lambda, which the sensitivity analysis shows to be dataset-dependent.

free parameters (5)
  • regularizer weight lambda = 40 (20NG), 40 (Yahoo), 300 (NYTimes)
    Chosen by grid search on a validation split (Section V.D); sensitivity analysis shows lambda is dataset-sensitive.
  • sampled words per topic v = 10
    Grid searched and set to 10 for all datasets (Section V.D).
  • Gumbel-Softmax temperature tau_g = 0.5
    Set by grid search (Section V.D).
  • softmax temperature tau_beta = 0.1
    Set by grid search (Section V.D).
  • NPMI estimation settings (window size, smoothing, thresholds) = unspecified
    The similarity matrix K is precomputed from the corpus without specifying estimation details, so the central regularizer is not exactly reproducible from the text.
assumptions (4)
  • domain assumption Precomputed NPMI on the training corpus is a valid proxy for human topic interpretability
    Invoked in Section IV.A to justify K(·)=NPMI; the paper acknowledges debate and relies on one correlation study [40].
  • standard math Gumbel-Softmax and relaxed subset sampling give differentiable and sufficiently low-variance gradients for sampling top words without replacement
    Used in Section IV.B via [18] and [41]; no bias or variance verification is reported in this setting.
  • domain assumption Frozen pretrained GloVe embeddings and the ETM decoder define meaningful topic-word distributions
    Borrowed from the ETM backbone (Sections III.B and V.D); shared by baseline comparisons.
  • domain assumption Co-occurrence statistics in the training corpus are stable enough that optimizing training NPMI transfers to test-set NPMI
    The regularizer uses training-set NPMI while the reported coherence is computed on unseen test data (Sections IV.A and V.B).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing Topic Interpretability for Neural Topic Modeling through Topic-wise Contrastive Learning." pith.science (2026). https://pith.science/paper/VESRXA6U

@misc{pith2026241217338,
  author       = {Pith},
  title        = {Pith review of: Enhancing Topic Interpretability for Neural Topic Modeling through Topic-wise Contrastive Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VESRXA6U}},
  note         = {Machine review of arXiv:2412.17338}
}
read the original abstract

Data mining and knowledge discovery are essential aspects of extracting valuable insights from vast datasets. Neural topic models (NTMs) have emerged as a valuable unsupervised tool in this field. However, the predominant objective in NTMs, which aims to discover topics maximizing data likelihood, often lacks alignment with the central goals of data mining and knowledge discovery which is to reveal interpretable insights from large data repositories. Overemphasizing likelihood maximization without incorporating topic regularization can lead to an overly expansive latent space for topic modeling. In this paper, we present an innovative approach to NTMs that addresses this misalignment by introducing contrastive learning measures to assess topic interpretability. We propose a novel NTM framework, named ContraTopic, that integrates a differentiable regularizer capable of evaluating multiple facets of topic interpretability throughout the training process. Our regularizer adopts a unique topic-wise contrastive methodology, fostering both internal coherence within topics and clear external distinctions among them. Comprehensive experiments conducted on three diverse datasets demonstrate that our approach consistently produces topics with superior interpretability compared to state-of-the-art NTMs.

Figures

Figures reproduced from arXiv: 2412.17338 by the authors.

Figure 1
Figure 1. The fundamental insight of ContraTopic. During the training of NTMs, words are sampled from each topic for the evaluation of topic coherence [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The results of topic interpretability evaluation. The first row shows the topic coherence of all the datasets in the test set, respectively. The second row [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. The results of document representation evaluation. (a) The two subfigures show the km-Purity scores on 20NG (left) and Yahoo(right). (b) The two [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: The sensitivity analysis results of λ and v on 20NG and Yahoo. Topic Coherence Topic Diversity Topic Coherence Topic Diversity λ λ 𝑣 𝑣 [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: The sensitivity analysis results of λ and v on NYTimes. ContraTopic outperforms almost every baseline in terms of topic coherence. Though NSTM has competitive results on 20NG, its topic diversity still has a certain gap with ContraTopic. Some baselines have shown an ob…
Figure 6
Figure 6. Figure 6: The results of backbone model substitution for 20NG in the first row and Yahoo in the second row. The blue lines correspond to various backbone [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: An Example of the questionnaire used in our human evaluation. [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 45 canonical work pages

  1. [1]

    Latent dirichlet allocation,

    D. M. Blei, A. Y . Ng, and M. I. Jordan, “Latent dirichlet allocation,” the Journal of machine Learning research , vol. 3, pp. 993–1022, 2003

    work page 2003

  2. [2]

    Survival topic models for predict- ing outcomes for trauma patients,

    Y . Zhang, R. Jiang, and L. Petzold, “Survival topic models for predict- ing outcomes for trauma patients,” in 2017 IEEE 33rd International Conference on Data Engineering (ICDE) . IEEE, 2017, pp. 1497–1504

    work page 2017

  3. [3]

    Source-lda: Enhancing probabilistic topic models using prior knowledge sources,

    J. Wood, P. Tan, W. Wang, and C. Arnold, “Source-lda: Enhancing probabilistic topic models using prior knowledge sources,” in 2017 IEEE 33rd International Conference on Data Engineering (ICDE) . IEEE, 2017, pp. 411–422

    work page 2017

  4. [4]

    Is automated topic model evaluation broken? the incoherence of coherence,

    A. Hoyle, P. Goel, A. Hian-Cheong, D. Peskov, J. Boyd-Graber, and P. Resnik, “Is automated topic model evaluation broken? the incoherence of coherence,” Advances in Neural Information Processing Systems , vol. 34, 2021

    work page 2021

  5. [5]

    Latent dirichlet allocation (lda) and topic modeling: models, applications, a survey,

    H. Jelodar, Y . Wang, C. Yuan, X. Feng, X. Jiang, Y . Li, and L. Zhao, “Latent dirichlet allocation (lda) and topic modeling: models, applications, a survey,” Multimedia Tools and Applications, vol. 78, no. 11, pp. 15 169– 15 211, 2019

    work page 2019

  6. [6]

    Tensor topic models with graphs and applications on individ- ualized travel patterns,

    L. Ziyue, “Tensor topic models with graphs and applications on individ- ualized travel patterns,” in 2021 IEEE 37th International Conference on Data Engineering (ICDE) . IEEE, 2021, pp. 2756–2761

    work page 2021

  7. [7]

    Intent mining: A social and semantic enhanced topic model for operation-friendly digital marketing,

    W. Wang, X. Cheng, Z. Liu, Y . Lin, Y . Shen, B. Hu, Z. Zhang, X. Zeng, J. Zhou, J. Gu et al., “Intent mining: A social and semantic enhanced topic model for operation-friendly digital marketing,” in 2022 IEEE 38th International Conference on Data Engineering (ICDE) . IEEE, 2022, pp. 3254–3267

    work page 2022

  8. [8]

    Neural variational inference for text processing,

    Y . Miao, L. Yu, and P. Blunsom, “Neural variational inference for text processing,” in International conference on machine learning . PMLR, 2016, pp. 1727–1736

    work page 2016

Show all 51 references
  1. [9]

    Autoencoding variational inference for topic models,

    A. Srivastava and C. Sutton, “Autoencoding variational inference for topic models,” in 5th International Conference on Learning Representations , 2017

    2017

  2. [10]

    Auto-encoding variational bayes,

    D. P. Kingma and M. Welling, “Auto-encoding variational bayes,” arXiv preprint arXiv:1312.6114, 2013

    2013 arXiv

  3. [11]

    Topic modeling in embedding spaces,

    A. B. Dieng, F. J. Ruiz, and D. M. Blei, “Topic modeling in embedding spaces,” Transactions of the Association for Computational Linguistics , vol. 8, pp. 439–453, 2020

    2020

  4. [12]

    Neural topic model via optimal transport,

    H. Zhao, D. Phung, V . Huynh, T. Le, and W. Buntine, “Neural topic model via optimal transport,” arXiv preprint arXiv:2008.13537 , 2020

    2008 arXiv

  5. [13]

    Coherence-aware neural topic modeling,

    R. Ding, R. Nallapati, and B. Xiang, “Coherence-aware neural topic modeling,” in Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing , 2018, pp. 830–836

    2018

  6. [14]

    Topic model or topic twaddle? re- evaluating semantic interpretability measures,

    C. Doogan and W. Buntine, “Topic model or topic twaddle? re- evaluating semantic interpretability measures,” in Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies , 2021, pp. 3824–3848

    2021

  7. [15]

    Topic modeling with wasserstein autoencoders,

    F. Nan, R. Ding, R. Nallapati, and B. Xiang, “Topic modeling with wasserstein autoencoders,” in Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics , 2019, pp. 6345–6381

    2019

  8. [16]

    Reading tea leaves: How humans interpret topic models,

    J. Chang, S. Gerrish, C. Wang, J. L. Boyd-Graber, and D. M. Blei, “Reading tea leaves: How humans interpret topic models,” in Advances in neural information processing systems , 2009, pp. 288–296

    2009

  9. [17]

    Discriminative topic mining via category-name guided text embedding,

    Y . Meng, J. Huang, G. Wang, Z. Wang, C. Zhang, Y . Zhang, and J. Han, “Discriminative topic mining via category-name guided text embedding,” in Proceedings of The Web Conference 2020 , 2020, pp. 2121–2132

    2020

  10. [18]

    Categorical reparameterization with gumbel-softmax,

    E. Jang, S. Gu, and B. Poole, “Categorical reparameterization with gumbel-softmax,” arXiv preprint arXiv:1611.01144 , 2016

    2016 arXiv

  11. [19]

    Whai: Weibull hybrid autoencoding inference for deep topic modeling,

    H. Zhang, B. Chen, D. Guo, and M. Zhou, “Whai: Weibull hybrid autoencoding inference for deep topic modeling,” in International Conference on Learning Representations , 2018

    2018

  12. [20]

    Decoupling sparsity and smoothness in the dirichlet variational autoencoder topic model

    S. Burkhardt and S. Kramer, “Decoupling sparsity and smoothness in the dirichlet variational autoencoder topic model.” J. Mach. Learn. Res. , vol. 20, no. 131, pp. 1–27, 2019

    2019

  13. [21]

    Atm: Adversarial-neural topic model,

    R. Wang, D. Zhou, and Y . He, “Atm: Adversarial-neural topic model,” Information Processing & Management , vol. 56, no. 6, p. 102098, 2019

    2019

  14. [22]

    Neural topic model with attention for supervised learning,

    X. Wang and Y . Yang, “Neural topic model with attention for supervised learning,” in International Conference on Artificial Intelligence and Statistics. PMLR, 2020, pp. 1147–1156

    2020

  15. [23]

    Graph attention topic modeling network,

    L. Yang, F. Wu, J. Gu, C. Wang, X. Cao, D. Jin, and Y . Guo, “Graph attention topic modeling network,” in Proceedings of The Web Conference 2020, 2020, pp. 144–154

    2020

  16. [24]

    Graph topic neural network for document representation,

    Q. Xie, J. Huang, P. Du, M. Peng, and J.-Y . Nie, “Graph topic neural network for document representation,” in Proceedings of the Web Conference 2021, 2021, pp. 3055–3065

    2021

  17. [25]

    Representing mixtures of word embeddings with mixtures of topic embeddings,

    D. Wang, D. Guo, H. Zhao, H. Zheng, K. Tanwisuth, B. Chen, and M. Zhou, “Representing mixtures of word embeddings with mixtures of topic embeddings,” arXiv preprint arXiv:2203.01570 , 2022

    2022 arXiv

  18. [26]

    Effective neural topic modeling with embedding clustering regularization,

    X. Wu, X. Dong, T. T. Nguyen, and A. T. Luu, “Effective neural topic modeling with embedding clustering regularization,” in International Conference on Machine Learning . PMLR, 2023, pp. 37 335–37 357

    2023

  19. [27]

    Learning deep representations by mutual information estimation and maximization,

    R. D. Hjelm, A. Fedorov, S. Lavoie-Marchildon, K. Grewal, P. Bach- man, A. Trischler, and Y . Bengio, “Learning deep representations by mutual information estimation and maximization,” arXiv preprint arXiv:1808.06670, 2018

    2018 arXiv

  20. [28]

    Supervised contrastive learning,

    P. Khosla, P. Teterwak, C. Wang, A. Sarna, Y . Tian, P. Isola, A. Maschinot, C. Liu, and D. Krishnan, “Supervised contrastive learning,” Advances in Neural Information Processing Systems , vol. 33, pp. 18 661–18 673, 2020

    2020

  21. [29]

    Detco: Unsupervised contrastive learning for object detection,

    E. Xie, J. Ding, W. Wang, X. Zhan, H. Xu, P. Sun, Z. Li, and P. Luo, “Detco: Unsupervised contrastive learning for object detection,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 8392–8401

    2021

  22. [30]

    Fsce: Few-shot object detection via contrastive proposal encoding,

    B. Sun, B. Li, S. Cai, Y . Yuan, and C. Zhang, “Fsce: Few-shot object detection via contrastive proposal encoding,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 7352–7362

    2021

  23. [31]

    Virtual adversarial training: a regularization method for supervised and semi-supervised learning,

    T. Miyato, S.-i. Maeda, M. Koyama, and S. Ishii, “Virtual adversarial training: a regularization method for supervised and semi-supervised learning,” IEEE transactions on pattern analysis and machine intelligence, vol. 41, no. 8, pp. 1979–1993, 2018

    1979

  24. [32]

    An efficient framework for learning sentence representations,

    L. Logeswaran and H. Lee, “An efficient framework for learning sentence representations,” in International Conference on Learning Representations, 2018

    2018

  25. [33]

    Improving topic disentanglement via contrastive learning,

    X. Zhou, J. Bu, S. Zhou, Z. Yu, J. Zhao, and X. Yan, “Improving topic disentanglement via contrastive learning,” Information Processing & Management, vol. 60, no. 2, p. 103164, 2023

    2023

  26. [34]

    Contrastive learning for neural topic model,

    T. Nguyen and A. T. Luu, “Contrastive learning for neural topic model,” Advances in Neural Information Processing Systems , vol. 34, 2021

    2021

  27. [35]

    Mitigating data sparsity for short text topic modeling by topic-semantic contrastive learning,

    X. Wu, A. T. Luu, and X. Dong, “Mitigating data sparsity for short text topic modeling by topic-semantic contrastive learning,” arXiv preprint arXiv:2211.12878, 2022

    2022 arXiv

  28. [36]

    Keyword assisted em- bedded topic model,

    B. Harandizadeh, J. H. Priniski, and F. Morstatter, “Keyword assisted em- bedded topic model,” in Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining , 2022, pp. 372–380

    2022

  29. [37]

    Enhancing neural topic model with multi-level supervisions from seed words,

    Y . Lin, X. Gao, X. Chu, Y . Wang, J. Zhao, and C. Chen, “Enhancing neural topic model with multi-level supervisions from seed words,” in Findings of the Association for Computational Linguistics: ACL 2023 , 2023, pp. 13 361–13 377

    2023

  30. [38]

    Neural topic model with reinforcement learning,

    L. Gui, J. Leng, G. Pergola, Y . Zhou, R. Xu, and Y . He, “Neural topic model with reinforcement learning,” in Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNL...

    2019

  31. [39]

    Are neural topic models broken?

    A. Hoyle, P. Goel, R. Sarkar, and P. Resnik, “Are neural topic models broken?” arXiv preprint arXiv:2210.16162 , 2022

    2022 arXiv

  32. [40]

    Large-scale correlation analysis of automated metrics for topic models,

    J. P. Lim and H. Lauw, “Large-scale correlation analysis of automated metrics for topic models,” in Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , 2023, pp. 13 874–13 898

    2023

  33. [41]

    Reparameterizable subset sampling via continuous relaxations,

    S. M. Xie and S. Ermon, “Reparameterizable subset sampling via continuous relaxations,” arXiv preprint arXiv:1901.10517 , 2019

    1901 arXiv

  34. [42]

    Stochastic optimization of areas under precision-recall curves with provable convergence,

    Q. Qi, Y . Luo, Z. Xu, S. Ji, and T. Yang, “Stochastic optimization of areas under precision-recall curves with provable convergence,” Advances in neural information processing systems , vol. 34, pp. 1752–1765, 2021

    2021

  35. [43]

    Stochastic auc maximization with deep neural networks,

    M. Liu, Z. Yuan, Y . Ying, and T. Yang, “Stochastic auc maximization with deep neural networks,” arXiv preprint arXiv:1908.10831 , 2019

    1908 arXiv

  36. [44]

    Mutual information neural estimation,

    M. I. Belghazi, A. Baratin, S. Rajeshwar, S. Ozair, Y . Bengio, A. Courville, and D. Hjelm, “Mutual information neural estimation,” in International conference on machine learning . PMLR, 2018, pp. 531–540

    2018

  37. [45]

    Newsweeder: Learning to filter netnews,

    K. Lang, “Newsweeder: Learning to filter netnews,” in Machine Learning Proceedings 1995. Elsevier, 1995, pp. 331–339

    1995

  38. [46]

    Importance of semantic representation: Dataless classification

    M.-W. Chang, L.-A. Ratinov, D. Roth, and V . Srikumar, “Importance of semantic representation: Dataless classification.” in AAAI, vol. 2, 2008, pp. 830–835

    2008

  39. [47]

    Short text topic modeling with topic distribution quantization and negative sampling decoder,

    X. Wu, C. Li, Y . Zhu, and Y . Miao, “Short text topic modeling with topic distribution quantization and negative sampling decoder,” in Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2020, pp. 1772–1782

    2020

  40. [48]

    Keyword assisted embedded topic model,

    B. Harandizadeh, J. H. Priniski, and F. Morstatter, “Keyword assisted embedded topic model,” arXiv preprint arXiv:2112.03101 , 2021

    2021 arXiv

  41. [49]

    Re-visiting automated topic model evaluation with large language models,

    D. Stammbach, V . Zouhar, A. Hoyle, M. Sachan, and E. Ash, “Re-visiting automated topic model evaluation with large language models,” arXiv preprint arXiv:2305.12152, 2023

    2023 arXiv

  42. [50]

    On-line lda: Adaptive topic models for mining text streams with applications to topic detection and tracking,

    L. AlSumait, D. Barbará, and C. Domeniconi, “On-line lda: Adaptive topic models for mining text streams with applications to topic detection and tracking,” in 2008 eighth IEEE international conference on data mining. IEEE, 2008, pp. 3–12

    2008

  43. [51]

    On-line trend analysis with topic models:# twitter trends detection topic model online,

    J. H. Lau, N. Collier, and T. Baldwin, “On-line trend analysis with topic models:# twitter trends detection topic model online,” in Proceedings of COLING 2012, 2012, pp. 1519–1534

    2012

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.