Pith. sign in

REVIEW 5 major objections 6 minor 45 references

Learning Credible Deep Neural Networks with Rationale Regularization

T0 review · 5 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Adding a rationale-alignment penalty during training makes deep neural networks generalize better to new data.

desk verdict CREX is a reasonable rationale-regularization method whose OOD generalization claim is worth testing, but the credibility metric is partly self-referential and the omission explanation's faithfulness is unverified. read the letter →

arxiv 1908.05601 v1 pith:PXABIWBF submitted 2019-08-13 cs.LG cs.IRstat.ML

Mengnan Du , Ninghao Liu , Fan Yang , Xia Hu This is my paper · ORCID
classification cs.LGcs.IRstat.ML
keywords credibledeepneuralnetworksrationaleregularizationlocalexplanationstextclassificationout-of-distributiongeneralizationexplanationalignmentsparseCREX
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

The paper sets out to make deep neural networks credible in a concrete sense: the words a model relies on for a prediction should match the evidence a human expert would cite. To achieve this, it adds a training loss called CREX that penalizes models whose local explanations disagree with expert rationales, plus a sparsity penalty for training instances that have no rationale. The central empirical claim is that this regularization does not reliably improve accuracy on the held-out test set, but it does increase accuracy on new, previously unseen datasets drawn from related tasks. The authors argue this matters because test-set accuracy can hide overfitting to dataset-specific bias, and a model that uses correct evidence is more likely to transfer and to earn trust.

What carries the argument

The mechanism is a differentiable, architecture-agnostic explanation score. The contribution of word $x_n^{(t)}$ to class $c$ is computed by deleting phrases containing that word and averaging the drop in prediction score, $$$s_n^{{(t,c)}}$ = \frac{1}{m}\sum_{j=1}^{m} [f_c(x_n)-f_c($x_n^{{(1:t-1-m+j)}}$ \oplus $x_n^{{(t+j:T)}}$)].$$ CREX minimizes a combined loss: a confident-explanation term that penalizes explanation mass on features outside the rationale, an uncertain-explanation term that feeds the model the rationale-suppressed input $x_n' = x_n \odot (1-r_n)$ and asks for lower confidence plus a more uniform explanation, and a sparsity term for instances without rationales. Expert knowledge flows into the network parameters through backpropagation, and the same loss applies to any differentiable text classifier.

What would settle it

Take a CREX-trained model and a vanilla model, delete exactly the rationale tokens from held-out test instances, and compare the resulting prediction changes. If the CREX model is not substantially more sensitive to the removal of rationale tokens, then its improved explanation scores are not evidence that its decisions now depend on those tokens.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a model's local explanation—the per-word contribution scores produced by deleting phrases from the input—can be used as a training signal, and steering those scores toward expert rationales makes the model rely on the evidence experts consider important. Across CNN, LSTM, and self-attention architectures on two text classification datasets, CREX-trained models show lower symmetric KL divergence between their explanations and expert rationales, and their accuracy on out-of-distribution datasets (Kaggle movie reviews, the sentence polarity dataset, and an adversarially filtered beer review set) rises by up to roughly four percentage points relative to vanilla training while held-out test accuracy remains comparable. The paper presents this as evidence that credible models have better real-world generalization, not merely better explanation scores.

Load-bearing premise

The method assumes that deleting phrases from the input and measuring the drop in prediction score faithfully reveals which words the model actually uses; if that attribution is inaccurate, matching it to expert rationales may not change what the model truly relies on.

Editorial extensions

If this is right

  • If the paper's claim is right, test-set accuracy alone is a poor yardstick for text classifiers meant to be deployed; out-of-distribution accuracy should be part of evaluation.
  • Because the benefit appears across CNN, LSTM, and self-attention models, the regularization is not tied to one network family and can be applied to other differentiable classifiers.
  • Sparse explanations alone, without any expert rationales, already beat vanilla training on both credibility and out-of-distribution accuracy, so the approach is usable when annotation is not available.
  • A few hundred rationales give nearly the full benefit of the complete rationale set, which keeps annotation cost down.
  • CREX adds no inference-time cost; the same trained model answers as quickly as a vanilla model, so the credibility gain does not slow deployment.

Reading between the lines

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

  • Editorial inference: if faithful explanation is the active ingredient, the same recipe should transfer to image classification, where rationales could be object masks or bounding boxes and deletion-based attribution is already standard.
  • Editorial inference: the observed correlation between credibility and out-of-distribution accuracy suggests that explanation-to-rationale divergence could serve as a model-selection signal when a target distribution is unavailable.
  • Editorial inference: since the regularizer forces explanations to be sparse, a natural next test is whether CREX-trained models are also more stable under adversarial edits that remove rationale words; the paper does not report this.
  • Editorial inference: the conclusions are tied to one omission-based attribution method, so a direct check would be to repeat the experiments with gradient-based attribution to see whether the out-of-distribution gain is caused by explanation alignment generally rather than by the specific deletion procedure.
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

5 major / 6 minor

Summary. The paper proposes CREX, a training-time regularizer for text-classification DNNs that aligns omission-based local explanations (contribution scores over input words) with expert rationales. When rationales are absent, CREX applies an L1 sparsity penalty to the explanation. The authors instantiate the method with CNN, LSTM, and self-attention architectures and evaluate on movie-review and beer-review datasets. They report that CREX improves the credibility score (symmetric KL between normalized explanation and rationale) on the test set, preserves test accuracy, and improves accuracy on external or adversarially constructed datasets not used in training or hyperparameter tuning. The paper also includes ablations, hyperparameter sensitivity analyses, and robustness checks for rationale quantity and quality.

Significance. If the out-of-distribution generalization result is reliable, the paper makes a practically useful contribution: it suggests a simple, architecture-agnostic way to use sparse human rationales to improve DNN robustness beyond the i.i.d. test set. The external evaluation is a genuine strength: the Kaggle and Polarity datasets are not used for fitting or tuning, and the improvement appears across three different architectures. The sparse self-guidance variant without rationales is also a useful extension. However, the central mechanistic claim—that aligning explanations with rationales makes the model actually rely on those rationales—is not established, because the faithfulness of the omission-based explanation is asserted but never verified, and the credibility metric is essentially the training objective. The statistical evidence for the OOD claim is also thin (three runs, no error bars).

major comments (5)
  1. [Section IV-A and Eq. (2)] The explanation method in Eq. (2) is the foundation of both the regularization losses and the credibility metric, but the manuscript never verifies that these omission-based contribution scores faithfully reflect the features the trained model actually uses. The paper lists faithfulness as a desideratum and cites [29], but no empirical faithfulness check is provided for the trained vanilla or CREX models. If the scores are unfaithful, then aligning them with rationales does not by itself imply the model relies on rationales, and the reported OOD gains could stem from generic sparsity or implicit regularization rather than the stated credibility mechanism. Please add a faithfulness evaluation, such as deletion/insertion fidelity curves for top-ranked words or a comparison with an independent attribution method (e.g., gradient-based or Integrated Gradients), for both vanilla and CREX models.
  2. [Section IV-B1 and Eq. (3)] The confident-explanation loss only penalizes the L1 magnitude of contributions on non-rationale words; it contains no term that increases the magnitude of contributions on rationale words. The text claims that this 'implicitly' encourages dense explanation scores on relevant features, but that is not guaranteed by the loss form: with the supervised loss fixed, the model could satisfy Eq. (3) by becoming generally insensitive to all words rather than by focusing on the rationale. Please either add an explicit positive-attraction term for rationale words or provide direct evidence (e.g., attribution mass on rationale versus non-rationale words before and after training) that the intended focusing behavior actually occurs.
  3. [Section IV-B2 and Eq. (6)] The uncertain-explanation loss is not aligned with the stated intuition. Eq. (6) contains the absolute value |f_yn(x_n) - f_yn(x'_n)|, so the loss rewards any change in the ground-truth probability after deleting rationale words, including an increase; the directional claim in the same section that the probability for the label 'should be much smaller' is not enforced by the objective. In addition, Eq. (4) constructs the negative input by zeroing the embeddings of rationale words, which is a different perturbation from the token deletion used in Eq. (2) and in the text's example. Please correct the loss to a directional form (e.g., max(0, f_yn(x_n) - f_yn(x'_n))) and align the negative-input construction with the perturbation used by the explanation operator.
  4. [Section V-B1 and Eq. (11)] The credibility metric is symmetric KL divergence between the normalized absolute explanation and the normalized rationale, which is essentially the same quantity minimized by the training losses in Eqs. (3) and (6). Therefore, the improved credibility scores in Table II are partly expected by construction and do not provide independent evidence that the trained models are more credible. This issue does not invalidate the external OOD accuracy results, but it does weaken the paper's interpretation of Table II. Please either report an independently defined credibility measure that is not directly optimized (e.g., agreement with held-out human rationales on a separate set, or consistency with an alternative attribution method) or explicitly acknowledge the circularity and rely on the OOD results for empirical support.
  5. [Section V-C, Tables IV and V] The central empirical claim that CREX 'significantly increases' accuracy on data beyond the test set is not supported by the reported statistics. All results are averages over three runs with no standard deviations, confidence intervals, or significance tests, and several improvements are small in magnitude (e.g., 0.5 percentage points on Kaggle for the attention model; 0.2 to 0.8 percentage points on the PR adversarial dataset). Please report run-level results with error bars and perform appropriate significance tests, and clarify whether the same hyperparameters were used across runs.
minor comments (6)
  1. [Section III and Eq. (2)] The phrase length m used in Eq. (2) is not defined in the notation section; please introduce it explicitly before first use.
  2. [Section IV-B2 and Eq. (5)] The normalization in Eq. (5) divides by the sum of absolute explanation scores, which is undefined if all scores are zero; please specify a small epsilon or a fallback for degenerate cases.
  3. [Section V-A3] The back-translation data augmentation uses the Google Translate API, which is non-deterministic and not reproducible; please specify translation settings or note the reproducibility limitation.
  4. [Section V-C1] The text reports 'relative accuracy improvement of 4.1%, 2.1%, 0.5%' for Kaggle, but the numbers in Table IV are percentage-point differences (e.g., 78.4 - 74.3 = 4.1 percentage points); use 'percentage points' or compute relative improvements correctly.
  5. [Section V-A4] Hyperparameters are tuned on the development set using both accuracy and credibility; because the credibility metric is the same as the training objective, this selection procedure may further inflate the test credibility results. Please clarify whether the reported test credibility was used in any form during model selection.
  6. [Fig. 3] The two qualitative examples are selected cases and cannot by themselves support the general credibility claim; consider adding a quantitative aggregate of explanation quality on a random sample of test instances.

Circularity Check

0 steps flagged · score 2.0 of 10

No load-bearing circularity: the headline out-of-distribution generalization is measured on external datasets untouched by training or tuning, so the central claim does not reduce to its inputs.

full rationale

The paper's claimed derivation chain is an empirical training/evaluation loop rather than a formal derivation from first principles. CREX optimizes L = L_supv + lambda1 * L_rationale + lambda2 * L_sparse (Eq. 10), where L_rationale combines a confident-explanation loss (Eq. 3) and an uncertain-explanation loss (Eq. 6). The credibility metric in Eq. (11) measures the same general quantity, namely agreement between the explanation s_n and the expert rationale r_n, so Tab. II partly reflects the objective that CREX was trained to optimize. However, this is not a circular prediction in the load-bearing sense: the credibility scores are computed on the held-out test split, whose rationales were not used to form the training loss, and the paper's headline claim is accuracy on Kaggle and Polarity, with the explicit statement that 'none of the data from these two datasets is utilized to train DNN models or tune hyperparameters.' Hyperparameters were selected on the development set rather than on the external test sets. The faithfulness assumption for the omission-based explanation in Eq. (2) is an external validity assumption, not a circular step, and the self-citations in the related work are background context rather than load-bearing evidence. Therefore the central generalization claim is independently testable and does not reduce to its inputs.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

CREX introduces no new physical entities. The central method relies on several free hyperparameters tuned on the dev set and on assumptions about the faithfulness of the explanation method and the reliability of expert rationales.

free parameters (6)
  • alpha (α) = 0.2, 0.5, 0.3, 0.5, 0.3, 0.5 for CNN/LSTM/Atten on MR/PR
    Balance term in uncertain explanation loss (Eq. 6), tuned per model and dataset on dev set.
  • lambda_1 (λ1) = 5e-2, 1e-3, 2e-4, 1e-4, 2e-4, 1e-4
    Weight of rationale loss (Eq. 10), tuned per model and dataset on dev set.
  • beta (β) = 1
    Weight combining confident and uncertain explanation losses (Eq. 7), fixed by hand.
  • lambda_2 (λ2) = 1e-5
    Weight of sparse explanation loss (Eq. 10), fixed by hand.
  • learning_rate = tuned over {1e-4, 1e-3, 1e-2, 1e-1}; final value not reported
    Section V-A4 reports the search range but not the selected values.
  • phrase_length_m = not reported
    Length of n-grams omitted in Eq. (2); an important detail for computing explanations is omitted.
assumptions (5)
  • domain assumption Omission-based explanation scores (Eq. 2) faithfully reflect the contribution of each word to the DNN's prediction.
    Section IV-A. The paper cites [29] for increased faithfulness of phrase omission, but the central regularization and evaluation depend on this without verification on the trained models.
  • domain assumption Expert rationales indicate features that are truly responsible for the prediction task and that generalize beyond the training set.
    Section III and experiments. The method's benefit relies on rationales being correct evidence; sensitivity to noisy rationales is acknowledged in Section V-E2.
  • domain assumption Sparse explanations are simpler and therefore more credible.
    Section IV-C cites [30]-[32]. Used to justify the sparse loss for instances without rationales.
  • ad hoc to paper Removing the rationale words should make the model uncertain and should spread explanation uniformly.
    Section IV-B2, Eqs. (4)-(6). This behavioral expectation is a design choice, not independently grounded, and the loss in Eq. (6) uses an absolute value that does not enforce the direction of the change.
  • ad hoc to paper The total explanation loss over all classes C, instead of only the label class, helps performance.
    Section IV-B1 states 'summing all categories could yield better results' with no supporting experiment; this design choice affects the loss.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning Credible Deep Neural Networks with Rationale Regularization." pith.science (2026). https://pith.science/paper/PXABIWBF

@misc{pith2026190805601,
  author       = {Pith},
  title        = {Pith review of: Learning Credible Deep Neural Networks with Rationale Regularization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PXABIWBF}},
  note         = {Machine review of arXiv:1908.05601}
}
read the original abstract

Recent explainability related studies have shown that state-of-the-art DNNs do not always adopt correct evidences to make decisions. It not only hampers their generalization but also makes them less likely to be trusted by end-users. In pursuit of developing more credible DNNs, in this paper we propose CREX, which encourages DNN models to focus more on evidences that actually matter for the task at hand, and to avoid overfitting to data-dependent bias and artifacts. Specifically, CREX regularizes the training process of DNNs with rationales, i.e., a subset of features highlighted by domain experts as justifications for predictions, to enforce DNNs to generate local explanations that conform with expert rationales. Even when rationales are not available, CREX still could be useful by requiring the generated explanations to be sparse. Experimental results on two text classification datasets demonstrate the increased credibility of DNNs trained with CREX. Comprehensive analysis further shows that while CREX does not always improve prediction accuracy on the held-out test set, it significantly increases DNN accuracy on new and previously unseen data beyond test set, highlighting the advantage of the increased credibility.

Figures

Figures reproduced from arXiv: 1908.05601 by the authors.

Figure 1
Figure 1. Two examples of expert rationale: words marked with purple [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Schematic of CREX. Black solid lines denote forward pass. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Sentence-level explanation heatmap comparison between [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: CNN performance under different values of parameter [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: CNN performance under different numbers of rationale. (a) [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Rationale quality analysis using CNN generalization accuracy [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 37 canonical work pages

  1. [29]

    Representation of linguistic form and function in recurrent neural networks,

    A. K ´ad´ar, G. Chrupała, and A. Alishahi, “Representation of linguistic form and function in recurrent neural networks,” Computational Lin- guistics, pp. 761–780, 2017

    work page 2017

  2. [1]

    Methods for interpreting and understanding deep neural networks,

    G. Montavon, W. Samek, and K.-R. M ¨uller, “Methods for interpreting and understanding deep neural networks,” Digital Signal Processing (DSP), 2018

    work page 2018

  3. [2]

    Techniques for interpretable machine learning,

    M. Du, N. Liu, and X. Hu, “Techniques for interpretable machine learning,” Communications of the ACM (CACM) , 2019

    work page 2019

  4. [3]

    Towards explanation of dnn-based prediction with guided feature inversion,

    M. Du, N. Liu, Q. Song, and X. Hu, “Towards explanation of dnn-based prediction with guided feature inversion,” Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD), 2018

    work page 2018

  5. [4]

    On attribution of recurrent neural network predictions via additive decomposition,

    M. Du, N. Liu, F. Yang, S. Ji, and X. Hu, “On attribution of recurrent neural network predictions via additive decomposition,” in The World Wide Web Conference (WWW), 2019

    work page 2019

  6. [5]

    Learning credible models,

    J. Wang, J. Oh, H. Wang, and J. Wiens, “Learning credible models,” in Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD) , 2018

    work page 2018

  7. [6]

    Did the model understand the question?

    P. K. Mudrakarta, A. Taly, M. Sundararajan, and K. Dhamdhere, “Did the model understand the question?” 56th Annual Meeting of the Association for Computational Linguistics (ACL) , 2018

    work page 2018

  8. [7]

    Does it care what you asked? understanding importance of verbs in deep learning qa system,

    B. Rychalska, D. Basaj, P. Biecek, and A. Wroblewska, “Does it care what you asked? understanding importance of verbs in deep learning qa system,” EMNLP workshop, 2018

    work page 2018

Show all 45 references
  1. [8]

    Why should i trust you?: Explaining the predictions of any classifier,

    M. T. Ribeiro, S. Singh, and C. Guestrin, “Why should i trust you?: Explaining the predictions of any classifier,” in Proceedings of the 22th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD) , 2016

    2016

  2. [9]

    Annotation artifacts in natural language inference data,

    S. Gururangan, S. Swayamdipta, O. Levy, R. Schwartz, S. R. Bowman, and N. A. Smith, “Annotation artifacts in natural language inference data,” North American Chapter of the Association for Computational Linguistics (NAACL), 2018

    2018

  3. [10]

    Man is to computer programmer as woman is to homemaker? debiasing word embeddings,

    T. Bolukbasi, K.-W. Chang, J. Y . Zou, V . Saligrama, and A. T. Kalai, “Man is to computer programmer as woman is to homemaker? debiasing word embeddings,” in Thirtieth Conference on Neural Information Processing Systems (NIPS) , 2016

    2016

  4. [11]

    Swag: A large-scale adversarial dataset for grounded commonsense inference,

    R. Zellers, Y . Bisk, R. Schwartz, and Y . Choi, “Swag: A large-scale adversarial dataset for grounded commonsense inference,” Empirical Methods in Natural Language Processing (EMNLP) , 2018

    2018

  5. [12]

    Harnessing deep neural networks with logic rules,

    Z. Hu, X. Ma, Z. Liu, E. Hovy, and E. Xing, “Harnessing deep neural networks with logic rules,” 54th Annual Meeting of the Association for Computational Linguistics (ACL) , 2016

    2016

  6. [13]

    Knowledgeable reader: Enhancing cloze- style reading comprehension with external commonsense knowledge,

    T. Mihaylov and A. Frank, “Knowledgeable reader: Enhancing cloze- style reading comprehension with external commonsense knowledge,” 56th Annual Meeting of the Association for Computational Linguistics (ACL), 2018

    2018

  7. [14]

    Rationale-augmented convo- lutional neural networks for text classification,

    Y . Zhang, I. Marshall, and B. C. Wallace, “Rationale-augmented convo- lutional neural networks for text classification,” in Empirical Methods in Natural Language Processing (EMNLP) , 2016

    2016

  8. [15]

    Using annotator rationales to improve machine learning for text categorization,

    O. Zaidan, J. Eisner, and C. Piatko, “Using annotator rationales to improve machine learning for text categorization,” in North American Chapter of the Association for Computational Linguistics (NAACL) , 2007

    2007

  9. [16]

    Rationalizing neural predictions,

    T. Lei, R. Barzilay, and T. Jaakkola, “Rationalizing neural predictions,” Empirical Methods in Natural Language Processing (EMNLP) , 2016

    2016

  10. [17]

    Annotator rationales for visual recogni- tion,

    J. Donahue and K. Grauman, “Annotator rationales for visual recogni- tion,” International Conference on Computer Vision (ICCV) , 2011

    2011

  11. [18]

    Why is that relevant? collecting annotator rationales for relevance judgments,

    T. McDonnell, M. Lease, M. Kutlu, and T. Elsayed, “Why is that relevant? collecting annotator rationales for relevance judgments,” in Fourth AAAI Conference on Human Computation and Crowdsourcing , 2016

    2016

  12. [19]

    Towards a rigorous science of interpretable machine learning,

    F. Doshi-Velez and B. Kim, “Towards a rigorous science of interpretable machine learning,” arXiv preprint arXiv:1702.08608 , 2017

    2017 arXiv

  13. [20]

    Evaluating explanation without ground truth in interpretable machine learning,

    F. Yang, M. Du, and X. Hu, “Evaluating explanation without ground truth in interpretable machine learning,” arXiv preprint arXiv:1907.06831, 2019

    1907 arXiv

  14. [21]

    Representation interpretation with spatial encoding and multimodal analytics,

    N. Liu, M. Du, and X. Hu, “Representation interpretation with spatial encoding and multimodal analytics,” in Proceedings of the Twelfth ACM International Conference on Web Search and Data Mining (WSDM) , 2019

    2019

  15. [22]

    Don’t just assume; look and answer: Overcoming priors for visual question answering,

    A. Agrawal, D. Batra, D. Parikh, and A. Kembhavi, “Don’t just assume; look and answer: Overcoming priors for visual question answering,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018

    2018

  16. [23]

    Women also snowboard: Overcoming bias in captioning models,

    L. A. Hendricks, K. Burns, K. Saenko, T. Darrell, and A. Rohrbach, “Women also snowboard: Overcoming bias in captioning models,” in 15th European Conference on Computer Vision (ECCV) , 2018

    2018

  17. [24]

    Overcoming language priors in visual question answering with adversarial regularization,

    S. Ramakrishnan, A. Agrawal, and S. Lee, “Overcoming language priors in visual question answering with adversarial regularization,” in Advances in Neural Information Processing Systems (NeurIPS) , 2018

    2018

  18. [25]

    Know what you don’t know: Unan- swerable questions for squad,

    P. Rajpurkar, R. Jia, and P. Liang, “Know what you don’t know: Unan- swerable questions for squad,” 56th Annual Meeting of the Association for Computational Linguistics (ACL) , 2018

    2018

  19. [26]

    Sequence classification with human attention,

    M. Barrett, J. Bingel, N. Hollenstein, M. Rei, and A. Søgaard, “Sequence classification with human attention,” in Proceedings of the 22nd Con- ference on Computational Natural Language Learning (CoNLL) , 2018, pp. 302–312

    2018

  20. [27]

    Deriving machine attention from human rationales,

    Y . Bao, S. Chang, M. Yu, and R. Barzilay, “Deriving machine attention from human rationales,” 2018 Conference on Empirical Methods in Natural Language Processing (EMNLP) , 2018

    2018

  21. [28]

    Understanding neural networks through representation erasure,

    J. Li, W. Monroe, and D. Jurafsky, “Understanding neural networks through representation erasure,” arXiv preprint arXiv:1612.08220, 2016

    2016 arXiv

  22. [30]

    Interpretable structure induc- tion via sparse attention,

    B. Peters, V . Niculae, and A. F. Martins, “Interpretable structure induc- tion via sparse attention,” in EMNLP Workshop, 2018

    2018

  23. [31]

    Sparse and constrained attention for neural machine translation,

    C. Malaviya, P. Ferreira, and A. F. Martins, “Sparse and constrained attention for neural machine translation,” 56th Annual Meeting of the Association for Computational Linguistics (ACL) , 2018

    2018

  24. [32]

    The mythos of model interpretability,

    Z. C. Lipton, “The mythos of model interpretability,” arXiv preprint arXiv:1606.03490, 2016

    2016 arXiv

  25. [33]

    Convolutional neural networks for sentence classification,

    Y . Kim, “Convolutional neural networks for sentence classification,” Empirical Methods in Natural Language Processing (EMNLP) , 2014

    2014

  26. [34]

    Long short-term memory,

    S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural computation, 1997

    1997

  27. [35]

    A structured self-attentive sentence embedding,

    Z. Lin, M. Feng, C. N. d. Santos, M. Yu, B. Xiang, B. Zhou, and Y . Bengio, “A structured self-attentive sentence embedding,” Interna- tional Conference on Learning Representations (ICLR) , 2017

    2017

  28. [36]

    Distilling the knowledge in a neural network,

    G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,” arXiv preprint arXiv:1503.02531 , 2015

    2015 arXiv

  29. [37]

    A sentimental education: Sentiment analysis using subjectivity summarization based on minimum cuts,

    B. Pang and L. Lee, “A sentimental education: Sentiment analysis using subjectivity summarization based on minimum cuts,” in Proceedings of the 42nd annual meeting on Association for Computational Linguistics (ACL), 2004

    2004

  30. [38]

    Learning attitudes and attributes from multi-aspect reviews,

    J. McAuley, J. Leskovec, and D. Jurafsky, “Learning attitudes and attributes from multi-aspect reviews,” in International Conference on Data Mining (ICDM) . IEEE, 2012

    2012

  31. [39]

    Improving neural machine translation models with monolingual data,

    R. Sennrich, B. Haddow, and A. Birch, “Improving neural machine translation models with monolingual data,” 54th Annual Meeting of the Association for Computational Linguistics (ACL) , 2016

    2016

  32. [40]

    Investigating backtranslation in neural machine translation,

    A. Poncelas, D. Shterionov, A. Way, G. M. d. B. Wenniger, and P. Passban, “Investigating backtranslation in neural machine translation,” arXiv preprint arXiv:1804.06189 , 2018

    2018 arXiv

  33. [41]

    Distributed representations of words and phrases and their composition- ality,

    T. Mikolov, I. Sutskever, K. Chen, G. S. Corrado, and J. Dean, “Distributed representations of words and phrases and their composition- ality,” in Conference on Neural Information Processing Systems (NIPS) , 2013

    2013

  34. [42]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

  35. [43]

    Dropout: a simple way to prevent neural networks from overfit- ting,

    N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhut- dinov, “Dropout: a simple way to prevent neural networks from overfit- ting,” The Journal of Machine Learning Research , 2014

    2014

  36. [44]

    Adversarially regularising neural nli models to integrate logical background knowledge,

    P. Minervini and S. Riedel, “Adversarially regularising neural nli models to integrate logical background knowledge,” The SIGNLL Conference on Computational Natural Language Learning (CoNLL) , 2018

    2018

  37. [45]

    Seeing stars: Exploiting class relationships for sentiment categorization with respect to rating scales,

    B. Pang and L. Lee, “Seeing stars: Exploiting class relationships for sentiment categorization with respect to rating scales,” inAnnual meeting on association for computational linguistics (ACL) , 2005

    2005

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.