Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Domain Adaptive Text Style Transfer

T0 review · 3 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 Borrowed source-domain data substantially improves text style transfer when target-domain data is scarce, the paper argues.

desk verdict Genuinely useful low-resource text style transfer results, but the novelty claim overreaches and the style-classifier training setup is a load-bearing ambiguity. read the letter →

arxiv 1908.09395 v1 pith:WHIPA3R6 submitted 2019-08-25 cs.CL

Dianqi Li , Yizhe Zhang , Zhe Gan , Yu Cheng , Chris Brockett , Ming-Ting Sun , Bill Dolan This is my paper · ORCID
classification cs.CL
keywords textstyletransferdomainadaptationlow-resourcenaturallanguageprocessingsentimentformalitynon-paralleldatavectorscontentpreservation
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 takes up text style transfer — rewriting a sentence's sentiment or formality while keeping its meaning — in the hard case where the target domain supplies very little non-parallel data. It argues that data from a related source domain, even when abundant, cannot simply be mixed in because the domains are distributionally mismatched. The proposed DAST models share an encoder-decoder across domains, add learnable domain vectors, and deploy domain-specific style classifiers to keep style fitting the target while borrowing generic content and style knowledge from the source. The central empirical claim is that with 1% of the target training data, DAST markedly outperforms models trained on the target alone and a simple fine-tuning baseline, on Yelp and Amazon sentiment transfer and on Enron formality transfer.

What carries the argument

The mechanism is a shared encoder-decoder trained jointly on source and target non-parallel data, with three components doing distinct work. Learnable domain vectors ($d_S$, $d_T$) condition the decoder so generations carry the target's lexical character; domain-specific style classifiers ($C_S$, $C_T$) are pretrained separately on each domain and used in the style-regularization loss to force transferred sentences to carry the requested style without letting one domain's style bleed into the other; and the auto-encoding reconstruction losses on both domains ($L^{S,T}_{ae}$) let the model learn generic content representations from the massive source corpus. A simplified variant, DAST-C, drops the source style labels and only borrows content knowledge; DAST uses both content and style from the source. The classifiers are kept fixed after pretraining, and the discrete-token sampling is made differentiable via the hard-sampling trick.

What would settle it

Train DAST from scratch with its style classifier pretrained only on the same 1% (or 0.1%) target training split used for the generator, keeping everything else identical. If the reported style accuracy and G-score drop substantially relative to the paper's numbers, then the low-resource gains depend on an information leak from the full target split rather than on domain adaptation alone; if the scores hold, the transfer claim is confirmed under the stricter regime.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a text style transfer system trained mostly on an out-of-domain source corpus can transfer useful content and style knowledge into a target domain without destroying domain-specific lexical character, and this becomes decisive when target data is scarce. In the experiments, with 1% of Yelp training data DAST reaches human-reference BLEU (hBLEU) 20.1 and G-score 23.1, against 3.7 and 8.6 for the best target-only model (ControlGen) and 13.9 and 18.5 for Finetune; comparable advantages hold on Amazon, and on Yahoo and Enron. Both DAST variants also keep high style accuracy and domain accuracy, and the advantage persists down to 0.1% of the target data (about 400 sentences), where target-only training degenerates into nonsense. The paper also finds, through ablations, that domain vectors are what keeps generations target-like, while domain-specific style classifiers prevent the two domains' style signals from being smeared together.

Load-bearing premise

The low-resource results hinge on an unstated detail in the training setup: the paper does not say whether the style classifier used inside the training objective is trained on the full target split or on the same small split as the generator, so part of the measured advantage could come from a classifier that already knows the full target distribution.

Editorial extensions

If this is right

  • When target data is very sparse, training only in the target domain is not viable; borrowing a sizable source corpus is what keeps hBLEU and G-score at usable levels.
  • Source data with unknown style labels still helps: DAST-C improves content preservation substantially, which matters for corpora like large unlabeled question logs.
  • When source and target share style labels, DAST additionally transfers generic style words while domain vectors keep the output target-appropriate, avoiding generations like 'The pizza is dramatic.'
  • The recipe extends to formality transfer where the source is parallel: adding a sequence-to-sequence loss on the source helps the non-parallel target, improving style accuracy and hBLEU on Enron.
  • Ablations show that removing domain-specific components lowers domain accuracy and style accuracy, so both the domain vector and the domain-specific classifier matter.

Reading between the lines

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

  • An implication the paper leaves implicit: the value of the source corpus should scale with how much its style vocabulary overlaps the target's, so choosing a source closer in genre or register to the target should further narrow the gap to full-data performance.
  • A testable extension is to stack multiple source domains in one model, with one domain vector per source, to see whether the per-domain classifiers keep the source signals mutually disentangled.
  • The reported 'limited target data' regime could be stricter: if the pretrained style classifier sees the full target split while the generator sees only 1%, the setup measures transfer plus a privileged classifier; rerunning with the classifier trained on the same small split would place the low-resource claim on firmer ground.
  • One could push below the 0.1% floor, asking whether domain vectors alone plus an unrelated source suffice, since DAST-C already suggests content transfer is the dominant benefit.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes two domain-adaptive text style transfer models, DAST-C and DAST, which use a large source-domain corpus to improve style transfer when the target domain has very little non-parallel data. DAST-C leverages source data with an unknown style label to improve content preservation, while DAST assumes the source and target share style labels and introduces learnable domain vectors plus separate domain-specific style classifiers to keep transferred text domain-appropriate. The models are evaluated on sentiment transfer (Yelp, Amazon, Yahoo) and formality transfer (Enron) using automatic metrics (hBLEU, S-acc, D-acc, G-score) and pairwise human evaluation. The central empirical claim is that with 1% or 0.1% of the target training data, DAST substantially improves content preservation and overall G-score over target-only baselines and over a source-pretrained then target-finetuned baseline.

Significance. If the results hold, the paper addresses a practical data-scarcity problem and appears to be the first systematic study of domain adaptation for text style transfer. The paper is strengthened by multiple target domains, an ablation study, human evaluation, and a released implementation. The low-resource gains on Yelp, Amazon, Yahoo, and Enron are large and consistent, and the human judgments support the headline comparison against ControlGen. However, the central quantitative claims depend on an unstated detail about how the style classifiers used in training are trained in the low-resource settings, and on the independence of the automatic style metric from the training objective. These issues need to be resolved before the results can be fully trusted.

major comments (3)
  1. [§5.2, §5.3, Eq. (3), Eq. (8)] The paper does not state which training split is used to pretrain the domain-specific style classifiers CT and CS in the low-resource experiments. In the 1% and 0.1% settings reported in Table 2 and Figure 2, the training objective includes a style-classifier regularization term; if CT is trained on the full target split (444k Yelp reviews or 554k Amazon reviews), then the style signal in training is not low-resource, and the comparison against target-only baselines is not apples-to-apples. Because the hard-sampling trick back-propagates through CT (Section 5.3), the decoder could in that case acquire full-target-domain lexical knowledge. Please state the exact training data used for CT and CS in every setting, and if CT uses the full split, rerun the low-resource experiments with CT trained only on the corresponding subsample, or otherwise justify the design.
  2. [§5.2, §5.3, Table 2, Figure 2] The metric S-acc is computed with a TextCNN style classifier 'pre-trained on the target domain,' and the training regularizers in Eq. (3) and Eq. (8) are also TextCNN classifiers 'pre-trained on corresponding domains' with fixed parameters. If the evaluation classifier and the training classifier are the same model, then G-score measures how well the generator satisfies the very discriminator it was trained against, not its ability to control style in general. Please state explicitly whether the S-acc classifier is distinct from CT; if it is not, report style accuracy with an independently trained or externally provided classifier in addition to the human evaluation already provided.
  3. [Table 2, Figure 2] All quantitative results are reported as single runs, with no standard deviations or significance tests. This is particularly important for the 0.1% setting (approximately 400 samples), where Figure 2 shows the claim of 'reasonable text style transfer' without a numerical table or variance estimate. Please report results over multiple random seeds with standard deviations, and assess whether the headline differences, especially in G-score, are statistically distinguishable.
minor comments (6)
  1. [Table 1] The column layout is hard to parse; for instance, the row 'YELP 444k 4k 1k' should make explicit which number is the target training set, and similarly for AMAZON and YAHOO.
  2. [Table 2] Several models have missing D-acc entries (for example, target-only baselines in the 100% block except ControlGen and Finetune); please state whether these were measured or are not applicable.
  3. [Table 3] In the Enron style-control row, the value '6%' should be formatted as '6.0%' for consistency.
  4. [§5.1] For Yahoo, the text says '7k question samples' and the table lists 4k/2k/1k splits; this is consistent but should be stated in one place to avoid confusion.
  5. [Appendix A.2] The conclusion that 'the number of source training data significantly affects performance' is confounded by the fact that the compared source datasets differ in genre and style-label availability; this should be acknowledged.
  6. [Table 6] The 'w/o both' row is described as 'simply increasing the number of training data,' but the ablation also removes domain vectors and domain-specific classifiers, so the interpretation should be phrased more carefully.

Circularity Check

1 steps flagged · score 4.0 of 10

The automatic style metric S-acc is self-referential (same target-domain TextCNN family as the training critic CT), but the central low-resource content-preservation claim is independently grounded by hBLEU and human evaluation.

  1. fitted input called prediction [Section 5.2, 'Style Control'; Section 3, Eq. (3); Section 5.3, Experimental Setup]
    "Style Control. We generate samples from the model and measure the style accuracy with a style classifier that is pre-trained on the target domain. ... where CT is the style classifier pretrained on the target domain. ... TextCNN (Kim, 2014) is employed for the domain-specific style classifiers pre-trained on corresponding domains."

    The classifier used to measure S-acc is a target-domain TextCNN, the same architecture and training regime as the CT whose log-probability is maximized in the training loss L_T_style (Eq. 3) and L_S,T_style (Eq. 8). A generator trained to satisfy CT is then scored by the same kind of critic, so S-acc partly measures fit to the training signal rather than style transfer against an external standard. The reported G-score is the geometric mean of S-acc and hBLEU, so the headline G-score gain inherits this self-referential component. The central content-preservation improvement is still independently supported by hBLEU against human references and by human evaluations, so the circularity is partial, not total.

full rationale

The core derivation chain is not circular: DAST trains an encoder-decoder with source-domain reconstruction, target-domain reconstruction, and fixed domain-specific style classifiers, then evaluates on held-out target sets against external baselines. The main low-resource content-preservation result (hBLEU 20.1 vs 3.7/13.9 on Yelp 1%) is measured against human references, and human preference judgments independently prefer DAST for content, style, and fluency. The one partial circularity is the automatic style metric: S-acc uses a target-domain TextCNN classifier of the same kind as the CT used in the training objective, so the style score is not a fully independent probe. The paper also leaves unspecified whether CT (and the evaluation classifier) in the 1% and 0.1% target-data experiments is trained on the full target split or on the same low-resource subsample; if the full split is used, the low-resource claim is weakened, but this is an experimental-reporting gap, not a circular reduction. No load-bearing self-citation chain is present: the self-citations (e.g., Zhang et al. 2017 for TextCNN implementation) are implementation details, and the style-classifier technique is attributed to external prior work (Hu et al. 2017). Overall, the central claim is empirically grounded outside the fitted values, with only the S-acc/G-score component partially self-referential.

Assumptions & free parameters 2 free parameters · 3 assumptions · 2 invented entities

The central claim rests on empirical assumptions rather than derivations: that source-domain autoencoding improves target content preservation, that domain vectors and per-domain classifiers factor out domain shift, and that the style classifier provides a reliable training signal. The paper introduces no physical or geometric entities; the domain vectors and unknown-style label are internal model components. The free parameters are standard embedding and architecture sizes chosen by hand.

free parameters (2)
  • Domain vector dimension = 50
    Chosen by hand in Section 5.3; a standard architectural choice for conditioning the decoder, not fitted to the evaluation metric.
  • Style label vector dimension = 150, or 200 when no domain vectors are used
    Chosen by hand in Section 5.3; standard embedding size, not fitted to the data.
assumptions (3)
  • domain assumption A shared encoder-decoder can learn generic content from source and target data without explicit domain alignment.
    The method's success relies on source data improving target content preservation, yet no adversarial or invariant feature constraint is imposed. This enters in Sections 4.2 and 4.3.
  • domain assumption Domain vectors plus per-domain style classifiers are sufficient to prevent domain mixing.
    The ablation study in Table 6 shows removing these components hurts performance, but the premise that they fully encode domain-specific style is assumed rather than proven.
  • domain assumption Style classifier gradients obtained via the hard-sampling trick are a reliable training signal for discrete text generation.
    The model samples discrete tokens and backpropagates classifier loss through those samples, an approximation. This is described in Section 5.3.
invented entities (2)
  • Domain vectors dS and dT
    purpose: Condition the decoder to generate domain-appropriate text
    Learned vectors introduced in Section 4.3; they are internal model parameters with no falsifiable handle outside the paper.
  • Unknown-style label lu
    purpose: Allow source-domain autoencoding when source style labels are unavailable
    An abstract label introduced for DAST-C in Section 4.2; it has no external testable prediction.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Domain Adaptive Text Style Transfer." pith.science (2026). https://pith.science/paper/WHIPA3R6

@misc{pith2026190809395,
  author       = {Pith},
  title        = {Pith review of: Domain Adaptive Text Style Transfer},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WHIPA3R6}},
  note         = {Machine review of arXiv:1908.09395}
}
read the original abstract

Text style transfer without parallel data has achieved some practical success. However, in the scenario where less data is available, these methods may yield poor performance. In this paper, we examine domain adaptation for text style transfer to leverage massively available data from other domains. These data may demonstrate domain shift, which impedes the benefits of utilizing such data for training. To address this challenge, we propose simple yet effective domain adaptive text style transfer models, enabling domain-adaptive information exchange. The proposed models presumably learn from the source domain to: (i) distinguish stylized information and generic content information; (ii) maximally preserve content information; and (iii) adaptively transfer the styles in a domain-aware manner. We evaluate the proposed models on two style transfer tasks (sentiment and formality) over multiple target domains where only limited non-parallel data is available. Extensive experiments demonstrate the effectiveness of the proposed model compared to the baselines.

Figures

Figures reproduced from arXiv: 1908.09395 by the authors.

Figure 1
Figure 1. Illustration of the proposed DAST-C (left) and DAST (right) model. DAST-C learns the generic content information through L S ae on massive source domain data with unknown style l u . For DAST, d T , dS and C T , CS denote domain vectors and domain-specific style classifiers, respectively. Better looked in color. This can be perceived as combining the source do￾main data with the target domain data to train a better … view at source ↗
Figure 2
Figure 2. Results on Yelp test set in terms of different percentage of target domain data. [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Questionnaire used to elicit pairwise judgments from crowd-sourced annotators. Candidate responses [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. SynGen-Vision: Synthetic Data Generation for training industrial vision models

    cs.CV 2025-09 conditional novelty 4.0 of 10

    A synthetic data pipeline combining Stable Diffusion textures, style transfer, noise filtering, and Blender rendering trains a rust detector that reports mAP50 0.87 on real images.

Reference graph

Works this paper leans on

43 extracted references · 31 canonical work pages · cited by 1 Pith paper

  1. [1]

    URL: " 'urlintro :=

    ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprinturl eprintpr...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Reina Akama, Kazuaki Inada, Naoya Inoue, Sosuke Kobayashi, and Kentaro Inui. 2017. Generating stylistically consistent dialog responses with transfer learning. In IJCNLP

    work page 2017

  4. [4]

    Amittai Axelrod, Xiaodong He, and Jianfeng Gao. 2011. Domain adaptation via pseudo in-domain data selection. In EMNLP

    work page 2011

  5. [5]

    Kyunghyun Cho, Bart van Merrienboer, Caglar Gulcehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. 2014. Learning phrase representations using rnn encoder--decoder for statistical machine translation. In EMNLP

    work page 2014

  6. [6]

    Chenhui Chu, Raj Dabre, and Sadao Kurohashi. 2017. An empirical comparison of simple domain adaptation methods for neural machine translation. arXiv preprint arXiv:1701.03214

    arXiv 2017

  7. [7]

    Qiming Diao, Minghui Qiu, Chao-Yuan Wu, Alexander J Smola, Jing Jiang, and Chong Wang. 2014. Jointly modeling aspects, ratings and sentiments for movie recommendation (jmars). In SIGKDD

    work page 2014

  8. [8]

    Zhenxin Fu, Xiaoye Tan, Nanyun Peng, Dongyan Zhao, and Rui Yan. 2018. Style transfer in text: Exploration and evaluation. In AAAI

    work page 2018

Show all 43 references
  1. [9]

    Chuang Gan, Zhe Gan, Xiaodong He, Jianfeng Gao, and Li Deng. 2017. Stylenet: Generating attractive visual captions with styles. In CVPR

    2017

  2. [10]

    Hongyu Gong, Suma Bhat, Lingfei Wu, Jinjun Xiong, and Wen-mei Hwu. 2019. Reinforcement learning based text style transfer without parallel training corpus. arXiv preprint arXiv:1903.10671

    2019 arXiv

  3. [11]

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. 2014. Generative adversarial nets. In NeurIPS

    2014

  4. [12]

    Ian J Goodfellow, Mehdi Mirza, Da Xiao, Aaron Courville, and Yoshua Bengio. 2013. An empirical investigation of catastrophic forgetting in gradient-based neural networks. arXiv preprint arXiv:1312.6211

    2013 arXiv

  5. [13]

    Smola, Jiayuan Huang, Marcel Schmittfull, Karsten Borgwardt, and Bernhard Sch \"o lkopf

    Arthur Gretton, AJ. Smola, Jiayuan Huang, Marcel Schmittfull, Karsten Borgwardt, and Bernhard Sch \"o lkopf. 2009. Covariate shift and local learning by distribution matching. MIT Press

    2009

  6. [14]

    Zhiting Hu, Haoran Shi, Zichao Yang, Bowen Tan, Tiancheng Zhao, Junxian He, Wentao Wang, Lianhui Qin, Di Wang, et al. 2018. Texar: A modularized, versatile, and extensible toolkit for text generation. arXiv preprint arXiv:1809.00794

    2018 arXiv

  7. [15]

    Zhiting Hu, Zichao Yang, Xiaodan Liang, Ruslan Salakhutdinov, and Eric P Xing. 2017. Toward controlled generation of text. In ICML

    2017

  8. [16]

    Xinyu Hua and Lu Wang. 2017. A pilot study of domain adaptation effect for neural abstractive summarization. In Proceedings of the Workshop on New Frontiers in Summarization

    2017

  9. [17]

    Harsh Jhamtani, Varun Gangal, Eduard Hovy, and Eric Nyberg. 2017. Shakespearizing modern language using copy-enriched sequence to sequence models. In Proceedings of the Workshop on Stylistic Variation

    2017

  10. [18]

    Yoon Kim. 2014. Convolutional neural networks for sentence classification. In EMNLP

    2014

  11. [19]

    Diederik P Kingma and Max Welling. 2013. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114

    2013 arXiv

  12. [20]

    Philipp Koehn and Josh Schroeder. 2007. Experiments in domain adaptation for statistical machine translation. In Proceedings of the second workshop on statistical machine translation

    2007

  13. [21]

    Juncen Li, Robin Jia, He He, and Percy Liang. 2018. Delete, retrieve, generate: a simple approach to sentiment and style transfer. In NAACL

    2018

  14. [22]

    Kevin Lin, Dianqi Li, Xiaodong He, Zhengyou Zhang, and Ming-Ting Sun. 2017. Adversarial ranking for language generation. In NeurIPS

    2017

  15. [23]

    Lajanugen Logeswaran, Honglak Lee, and Samy Bengio. 2018. Content preserving text generation with attribute controls. In NeurIPS

    2018

  16. [24]

    Paul Michel and Graham Neubig. 2018. Extreme adaptation for personalized neural machine translation. In ACL

    2018

  17. [25]

    Remi Mir, Bjarke Felbo, Nick Obradovich, and Iyad Rahwan. 2019. Evaluating style transfer for text. arXiv preprint arXiv:1904.02295

    2019 arXiv

  18. [26]

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In ACL

    2002

  19. [27]

    Shrimai Prabhumoye, Yulia Tsvetkov, Ruslan Salakhutdinov, and Alan W Black. 2018. Style transfer through back-translation. In ACL

    2018

  20. [28]

    Xiaoye Qu, Zhikang Zou, Yu Cheng, Yang Yang, and Pan Zhou. 2019. Adversarial category alignment network for cross-domain sentiment classification. In NAACL

    2019

  21. [29]

    Sudha Rao and Joel Tetreault. 2018. Dear sir or madam, may i introduce the gyafc dataset: Corpus, benchmarks and metrics for formality style transfer. In NAACL

    2018

  22. [30]

    Cicero Nogueira dos Santos, Igor Melnyk, and Inkit Padhi. 2018. Fighting offensive language on social media with unsupervised text style transfer. In ACL

    2018

  23. [31]

    Rico Sennrich, Barry Haddow, and Alexandra Birch. 2016 a . Controlling politeness in neural machine translation via side constraints. In NAACL

    2016

  24. [32]

    Rico Sennrich, Barry Haddow, and Alexandra Birch. 2016 b . Improving neural machine translation models with monolingual data. In ACL

    2016

  25. [33]

    Tianxiao Shen, Tao Lei, Regina Barzilay, and Tommi Jaakkola. 2017. Style transfer from non-parallel text by cross-alignment. In NeurIPS

    2017

  26. [34]

    Sandeep Subramanian, Guillaume Lample, Eric Michael Smith, Ludovic Denoyer, Marc'Aurelio Ranzato, and Y-Lan Boureau. 2018. Multiple-attribute text style transfer. arXiv preprint arXiv:1811.00552

    2018 arXiv

  27. [35]

    Tsung-Hsien Wen, Milica Ga s i \'c , Nikola Mrk s i \'c , Lina M Rojas-Barahona, Pei-Hao Su, David Vandyke, and Steve Young. 2016. Multi-domain neural network language generation for spoken dialogue systems. In NAACL

    2016

  28. [36]

    Jingjing Xu, Sun Xu, Qi Zeng, Xiaodong Zhang, Xuancheng Ren, Houfeng Wang, and Wenjie Li. 2018. Unpaired sentiment-to-sentiment translation: A cycled reinforcement learning approach. In ACL

    2018

  29. [37]

    Zichao Yang, Zhiting Hu, Chris Dyer, Eric P Xing, and Taylor Berg-Kirkpatrick. 2018. Unsupervised text style transfer using language models as discriminators. In NeurIPS

    2018

  30. [38]

    Saizheng Zhang, Emily Dinan, Jack Urbanek, Arthur Szlam, Douwe Kiela, and Jason Weston. 2018 a . Personalizing dialogue agents: I have a dog, do you have pets too? In ACL

    2018

  31. [39]

    Xiang Zhang, Junbo Zhao, and Yann LeCun. 2015. Character-level convolutional networks for text classification. In NeurIPS

    2015

  32. [40]

    Ye Zhang, Nan Ding, and Radu Soricut. 2018 b . Shaped: Shared-private encoder-decoder for text style adaptation. In NAACL

    2018

  33. [41]

    Yi Zhang, Jingjing Xu, Pengcheng Yang, and Xu Sun. 2018 c . Learning sentiment memories for sentiment modification without parallel data. In EMNLP

    2018

  34. [42]

    Yizhe Zhang, Xiang Gao, Sungjin Lee, Chris Brockett, Michel Galley, Jianfeng Gao, and Bill Dolan. 2019. Consistent dialogue generation with self-supervised feature learning. arXiv preprint arXiv:1903.05759

    2019 arXiv

  35. [43]

    Yizhe Zhang, Dinghan Shen, Guoyin Wang, Zhe Gan, Ricardo Henao, and Lawrence Carin. 2017. Deconvolutional paragraph representation learning. In NeurIPS

    2017

Pith tools

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