Pith. sign in

REVIEW 4 major objections 5 minor 30 references

Rethinking Attribute Representation and Injection for Sentiment Classification

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read This paper claims that the standard bias-in-attention method is the least effective way to use user and product attributes for sentiment classification.

desk verdict CHIM is a genuinely new attribute gating method and the four-location comparison is valuable; the sweeping 'attention is worst' claim isn't supported by the paper's own numbers. read the letter →

arxiv 1908.09590 v1 pith:5KDBDCJ2 submitted 2019-08-26 cs.CL

classification cs.CL
keywords sentimentclassificationattributerepresentationinjectionuserandproductattributesattentionmechanismBiLSTMchunk-wiseimportancematrix
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to overturn a default practice in sentiment classification: encoding user and product information as extra bias terms inside the attention mechanism. The author argues that this standard method is the least effective way to use such attributes, and proposes instead to represent them as chunk-wise importance matrices (CHIM) that multiply into the model's weight matrices rather than adding to biases. With only a simple BiLSTM plus attention as the base model, the CHIM-equipped model is reported to beat published systems that use far more complex architectures on IMDB and Yelp benchmarks. The paper also claims that the attention mechanism is the worst of four injection locations for these attributes, and that learned attribute encodings transfer to product-category classification and review-headline generation. If right, the result would redirect how practitioners inject metadata and show that architectural complexity is not what drives these gains.

What carries the argument

Chunk-wise Importance Matrix (CHIM). For a target weight matrix $\mathbf{W}$ of shape $(D_1,D_2)$, the user and product embeddings are concatenated, linearly transformed, and reshaped into a chunk matrix $\mathbf{C}$ of shape $(D_1/C_1,D_2/C_2)$. This chunk is repeated $C_1 \times C_2$ times, concatenated, and passed through a sigmoid to become a gate matrix in $[0,1]^{D_1 \times D_2}$, which is broadcast-multiplied with $\mathbf{W}$ (the effective weight is $\mathbf{W}' \ast \mathbf{W}$). The chunking cuts the number of added parameters by the chunk-size factors, and the sigmoid gating keeps optimization stable by shrinking unimportant weights toward zero; the paper also frames the gate as a coarse-grained importance control over neurons. The same CHIM module is dropped into four locations—word embedding, BiLSTM encoder, attention, and classifier—to compare where attributes help most.

What would settle it

Run the same base model with CHIM at all four locations across ten random seeds on IMDB and Yelp 2013, and construct confidence intervals; if the attention-injected variant is not reproducibly the worst, the paper's headline conclusion fails.

Watch

Extended reading notes

Core claim

This paper's central claim is that attribute representation and injection matter more than model architecture: representing user and product attributes as chunk-wise importance matrices and injecting them into early or late parts of a simple BiLSTM classifier outperforms the field's default bias-in-attention method, and does so even against systems with much more elaborate architectures. Concretely, CHIM maps the concatenated user and product embeddings to a small matrix, tiles it to the shape of a target weight matrix $\mathbf{W}$, pushes it through a sigmoid, and broadcast-multiplies it into $\mathbf{W}$, so the text transformation itself becomes user- and product-dependent. In the reported results, CHIM injected into the embedding layer gives the best accuracy and CHIM into the classifier gives the best RMSE on the sentiment benchmarks, while the attention mechanism is consistently the worst CHIM injection location; the author concludes that attention is the worst location to inject attributes regardless of representation, contradicting earlier work.

Load-bearing premise

The paper assumes that single-run accuracy differences of around one point on held-out sets, without repeated seeds or significance tests, are enough to rank the four injection locations reliably; if those differences are noise, the conclusion that attention is the worst location collapses.

Editorial extensions

If this is right

  • Replacing bias-in-attention with a multiplicative CHIM gate at another location is a concrete, low-cost change practitioners can make to sentiment classifiers.
  • Reported gains over published systems imply that complex hierarchical or memory-network architectures are not necessary to exploit user and product information; attribute handling can do the work.
  • CHIM attribute encodings trained for sentiment also carry information about product category, so they can be transferred to other tasks instead of being trained from scratch.
  • Joint injection into attention plus another location can beat single-location injection, so the four locations are complementary rather than mutually exclusive.
  • The parameter savings of chunking make attribute injection feasible at the embedding and encoder layers, where full matrix-based representations would be too costly to optimize.

Reading between the lines

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

  • Because the paper's own results show the best location varies by metric and transfer task, a practical extension is to treat injection location as a tunable hyperparameter rather than a universal architectural rule.
  • CHIM's gating mechanism resembles an input-dependent regularizer, so one testable extension is to inspect gate values for individual users and check whether they correlate with known preference profiles.
  • The method is defined for user and product attributes but is attribute-agnostic; applying the same chunked importance matrix to location, time, topic, or aspect attributes would test how general the conclusion is.
  • The transfer results suggest attribute encodings pretrained on large review corpora could serve as reusable personalization features for other NLP tasks, although the paper only demonstrates two transfer tasks.
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

4 major / 5 minor

Summary. The paper proposes a new way to represent user and product attributes for sentiment classification: chunk-wise importance weight matrices (CHIM), where attribute embeddings are transformed into a small chunked matrix, repeated and sigmoid-gated, and then used to modulate the weight matrix of a non-linear layer. The paper studies four injection locations (embedding, encoder, attention, classifier) and three representations (bias, matrix, CHIM) on a common BiLSTM-with-attention base model. It claims that the standard bias-in-attention method is the least effective, that attention is the worst injection location regardless of representation, and that the CHIM representation lets a simple base model outperform prior state-of-the-art architectures on IMDB, Yelp 2013, and Yelp 2014. It also reports transfer experiments to product category classification and review headline generation. The paper releases its code and datasets.

Significance. If the CHIM results are replicable, the paper makes a useful empirical contribution: a parameter-efficient attribute injection method that works with a simple BiLSTM and reports accuracy gains over more complex published systems. The public release of code and datasets supports reproducibility. The transfer experiments are a constructive addition, and the comparison across representations and locations is a reasonable study design. However, the headline interpretive claim that attention is the worst injection location 'regardless of representation' is not supported by the paper's own data, and the 'significant improvement' wording lacks statistical support. The core method can stand, but the paper's central claims need to be substantially reworded or re-validated.

major comments (4)
  1. [§3.2, Figure 3 and §5] The universal claim that 'injection to the attention mechanism, no matter which representation method is used, garners the worst performance' is contradicted by the paper's own development-set results. The text in §3.2 states that matrix-based representations perform worst when injected to embeddings and encoder, while matrix-attention improves over bias-attention; Figure 3 shows matrix-attention above matrix-embedding and matrix-encoder. The conclusion should be narrowed to the CHIM representation or backed by per-representation location comparisons.
  2. [§3.2, Table 2] The claim that 'CHIM-attention consistently performs the worst among CHIM-based representations' is not supported on Yelp 2014, where CHIM-attention (68.5) slightly exceeds CHIM-encoder (68.4). In addition, the sentiment-classification results are single runs with no variance estimates or significance tests; differences of 0.1–1.0 accuracy points are within plausible noise. The paper should report multiple seeds and statistical tests before drawing a strong ranking conclusion from these numbers.
  3. [§2.2–§2.3] The design does not support the 'regardless of representation' conclusion because the bias-based representation is only evaluated at the attention location (bias-attention). The representation and injection-location factors are therefore confounded for the bias condition; bias at embedding, encoder, and classifier is untested. The experiments should either include bias at all four locations or the claims must be restricted to the combinations actually compared.
  4. [§3.2, Table 2] The abstract and §5 describe the improvements as 'significant' without reporting any significance test. Given that the differences against prior work are small in absolute terms and no variance is reported, 'significant' should be replaced by a quantitative claim backed by significance testing or by a more cautious phrase such as 'consistent improvements'.
minor comments (5)
  1. [§3.3] The word 'satistfy' should be 'satisfy' in the dataset description.
  2. [Figure 1] The attention weights in Figure 1b are displayed with a large '4' that is not explained in the caption; it obscures the illustrative point rather than clarifying it.
  3. [§2.2, Eqs. (13)–(16)] The dimension definitions are confusing: Eq. (13) gives a vector w' of size D1*D2 for the matrix-based method, while the CHIM method reshapes a vector c into a chunked matrix C of shape (D1/C1, D2/C2). The text should explicitly define the size of c and how it is produced in the CHIM case.
  4. [Table 3] Perplexity is reported without variance intervals, while accuracy is reported with ± intervals over 10 runs; please report the same repeated-run statistics for perplexity.
  5. [§2.3] The explanation that matrix-attention and matrix-classifier are easier to optimize because their weight matrices are 'relatively smaller' is stated as fact but is an untested speculation; it could be flagged as a hypothesis.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: CHIM's advantage and the attention-location finding are empirical results, not consequences of the method's definition or of self-citations.

full rationale

The paper's derivation chain is self-contained. The CHIM representation (Eqs. 13-17) is defined via attribute embeddings gated onto the base model's weight matrices; the improvement over bias-attention is obtained by comparing test-set accuracies against external baselines (Table 2) and by dev-set comparisons (Figure 3). Neither the definition of CHIM nor the four injection locations presupposes that attention is worst or that CHIM is best; those are empirical outcomes. The claim that attention is the worst location is an inductive generalization from experiments, not built into the method. Self-citations appear only as a baseline (HCSC, Amplayo et al. 2018a) and as optional inspiration for interpreting the gating mechanism (Amplayo et al. 2018b), neither of which is load-bearing: removing them would not change any equation or evaluation. Possible weaknesses in the strength of the attention-is-worst conclusion, such as single runs without significance tests and the observation that CHIM-attention slightly beats CHIM-encoder on Yelp 2014 in Table 2, are statistical-evidence concerns, not circularity. No equation reduces to its inputs and no fitted parameter is renamed as a prediction, so the paper does not exhibit the circularity patterns enumerated in the review rules.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the reliability of empirical comparisons, which depend on the modeling choices and dataset assumptions listed above. No new physical or mathematical entities are introduced; CHIM is an algorithmic construction, not a new object requiring independent evidence.

free parameters (2)
  • Chunk size factors C1 and C2 = 15
    Chosen by hand and fixed across all models; no sensitivity analysis is reported. These determine the parameter reduction in CHIM and directly affect the trade-off between representational capacity and trainability.
  • Attribute embedding dimension = 300
    Set to 300 for word, user, and product vectors; a hyperparameter not justified by experiments. It controls the size of the attribute lookup tables and the initial matrix transformations.
assumptions (4)
  • domain assumption User and product attributes are informative for predicting the sentiment of a review.
    The entire method relies on these categorical attributes improving sentiment classification relative to a text-only model. This is an established assumption in prior work, not tested here.
  • domain assumption A simple BiLSTM with attention and a logistic classifier is a sufficient base model for comparing attribute injection methods.
    Section 2.1 defines the base model, and the paper asserts that architectural complexity is orthogonal to attribute handling. This assumption justifies the claim that results transfer to more complex bases.
  • domain assumption The dataset splits and the twenty-core filtering from Tang et al. (2015) provide a fair and standard benchmark.
    Used without re-validation, and all comparisons to prior state-of-the-art results rely on the same splits. Any hidden flaw in the original preprocessing would affect all compared models.
  • domain assumption Pre-trained GloVe embeddings provide a suitable initialization for word vectors.
    The model initializes word vectors with GloVe; this is a standard tooling choice, but the paper does not test sensitivity to this initialization.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Rethinking Attribute Representation and Injection for Sentiment Classification." pith.science (2026). https://pith.science/paper/5KDBDCJ2

@misc{pith2026190809590,
  author       = {Pith},
  title        = {Pith review of: Rethinking Attribute Representation and Injection for Sentiment Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5KDBDCJ2}},
  note         = {Machine review of arXiv:1908.09590}
}
read the original abstract

Text attributes, such as user and product information in product reviews, have been used to improve the performance of sentiment classification models. The de facto standard method is to incorporate them as additional biases in the attention mechanism, and more performance gains are achieved by extending the model architecture. In this paper, we show that the above method is the least effective way to represent and inject attributes. To demonstrate this hypothesis, unlike previous models with complicated architectures, we limit our base model to a simple BiLSTM with attention classifier, and instead focus on how and where the attributes should be incorporated in the model. We propose to represent attributes as chunk-wise importance weight matrices and consider four locations in the model (i.e., embedding, encoding, attention, classifier) to inject attributes. Experiments show that our proposed method achieves significant improvements over the standard approach and that attention mechanism is the worst location to inject attributes, contradicting prior work. We also outperform the state-of-the-art despite our use of a simple base model. Finally, we show that these representations transfer well to other tasks. Model implementation and datasets are released here: https://github.com/rktamplayo/CHIM.

Figures

Figures reproduced from arXiv: 1908.09590 by the authors.

Figure 1
Figure 1. Illustrative examples of issues when repre [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. CHIM-based attribute representation and in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Accuracies (y-axis) of different attribute rep [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Heatmap of the accuracies of singly and jointly injected CHIM models. Values on each cell represents either the accuracy (for singly injected mod￾els) or the difference between the singly and doubly injected models per row. rized into two statements. Firstly, injecting…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 27 canonical work pages

  1. [1]

    Reinald Kim Amplayo, Jihyeok Kim, Sua Sung, and Seung-won Hwang. 2018a. Cold-start aware user and product attention for sentiment classification. In Proceedings of the 56th Annual Meeting of the As- sociation for Computational Linguistics, ACL 2018, Melbourne, Australia, July 15-20, 2018, Volume 1: Long Papers, pages 2535–2544. Reinald Kim Amplayo, Kyungja...

  2. [6]

    CoRR, abs/1707.02633

    Controlling linguistic style aspects in neural language genera- tion. CoRR, abs/1707.02633. Tomohiro Fukuhara, Hiroshi Nakagawa, and Toyoaki Nishida

  3. [9]

    In 2016 IEEE Conference on Computer Vi- sion and Pattern Recognition, CVPR 2016, Las Ve- gas, NV , USA, June 27-30, 2016, pages 770–778

    Deep residual learning for image recog- nition. In 2016 IEEE Conference on Computer Vi- sion and Pattern Recognition, CVPR 2016, Las Ve- gas, NV , USA, June 27-30, 2016, pages 770–778. Geoffrey E. Hinton, Nitish Srivastava, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhut- dinov

  4. [13]

    In IJCAI 2011, Proceedings of the 22nd International Joint Conference on Artificial In- telligence, Barcelona, Catalonia, Spain, July 16-22, 2011, pages 1820–1825

    Incorporat- ing reviewer and product information for review rat- ing prediction. In IJCAI 2011, Proceedings of the 22nd International Joint Conference on Artificial In- telligence, Barcelona, Catalonia, Spain, July 16-22, 2011, pages 1820–1825. Chenghua Lin and Yulan He

  5. [16]

    Extreme adap- tation for personalized neural machine translation. In Proceedings of the 56th Annual Meeting of the As- sociation for Computational Linguistics, ACL 2018, Melbourne, Australia, July 15-20, 2018, Volume 2: Short Papers, pages 312–318. Andrew Y . Ng

  6. [17]

    Personalized re- view generation by expanding phrases and attending on aspect-aware representations. In Proceedings of the 56th Annual Meeting of the Association for Com- putational Linguistics, ACL 2018, Melbourne, Aus- tralia, July 15-20, 2018, Volume 2: Short Papers , pages 706–711. Bo Pang and Lillian Lee

  7. [18]

    In Proceedings of the 38th International ACM SIGIR Conference on Re- search and Development in Information Retrieval, Santiago, Chile, August 9-13, 2015, pages 393–402

    Retrieval of relevant opinion sentences for new products. In Proceedings of the 38th International ACM SIGIR Conference on Re- search and Development in Information Retrieval, Santiago, Chile, August 9-13, 2015, pages 393–402. Jeffrey Pennington, Richard Socher, and Christo- pher D. Manning

  8. [19]

    Glove: Global vectors for word representation. In Proceedings of the 2014 Conference on Empirical Methods in Natural Lan- guage Processing, EMNLP 2014, October 25-29, 2014, Doha, Qatar, A meeting of SIGDAT, a Special Interest Group of the ACL, pages 1532–1543. Yikang Shen, Shawn Tan, Alessandro Sordoni, and Aaron C. Courville

Show all 30 references
  1. [21]

    Journal of Machine Learning Research, 15(1):1929–1958

    Dropout: a simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 15(1):1929–1958. Chenhao Tan, Lillian Lee, Jie Tang, Long Jiang, Ming Zhou, and Ping Li

  2. [22]

    In Proceedings of the 17th ACM SIGKDD International Confer- ence on Knowledge Discovery and Data Mining, San Diego, CA, USA, August 21-24, 2011 , pages 1397–1405

    User-level sentiment anal- ysis incorporating social networks. In Proceedings of the 17th ACM SIGKDD International Confer- ence on Knowledge Discovery and Data Mining, San Diego, CA, USA, August 21-24, 2011 , pages 1397–1405. Duyu Tang, Bing Qin, and Ting Liu

  3. [23]

    Learning semantic representations of users and products for document level sentiment classification. In Proceed- ings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th Interna- tional Joint Conference on Natural Language Pro- cessing of the...

  4. [24]

    In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, EMNLP 2015, Lisbon, Portu- gal, September 17-21, 2015, pages 601–611

    Sentiment flow - A general model of web review argumentation. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, EMNLP 2015, Lisbon, Portu- gal, September 17-21, 2015, pages 601–611. Zhen Wu, Xin-Yu Dai, Cunyan Yin, Shujian Huang, and Jiajun Chen

  5. [25]

    Improving review repre- sentations with user attention and product attention for sentiment classification. In Proceedings of the Thirty-Second AAAI Conference on Artificial Intelli- gence, (AAAI-18), the 30th innovative Applications of Artificial Intelligence (IAAI-18), and the 8...

  6. [26]

    In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, EMNLP 2017, Copenhagen, Denmark, September 9-11, 2017, pages 527–533

    Identifying and tracking sentiments and topics from social media texts during natural disasters. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, EMNLP 2017, Copenhagen, Denmark, September 9-11, 2017, pages 527–533. Min Yang, Qiang Qu,...

  7. [27]

    Hi- erarchical attention networks for document classifi- cation. In NAACL HLT 2016, The 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, San Diego California, USA, June 12- 17, 2016, pages 1480–1489. Ma...

  8. [28]

    CoRR, abs/1212.5701

    ADADELTA: an adaptive learning rate method. CoRR, abs/1212.5701. Weinan Zhang, Ting Liu, Yifa Wang, and Qingfu Zhu

  9. [29]

    CoRR, abs/1701.02073

    Neural personalized response generation as domain adaptation. CoRR, abs/1701.02073. Pengcheng Zhu and Yujiu Yang

  10. [30]

    In Proceedings of the Eighth International Sympo- sium on Information and Communication Technol- ogy, Nha Trang City, Viet Nam, December 7-8, 2017, pages 181–188

    Parallel multi- feature attention on neural sentiment classification. In Proceedings of the Eighth International Sympo- sium on Information and Communication Technol- ogy, Nha Trang City, Viet Nam, December 7-8, 2017, pages 181–188

  11. [2004]

    In COLING 2004, 20th International Conference on Computa- tional Linguistics, Proceedings of the Conference, 23-27 August 2004, Geneva, Switzerland

    Deter- mining the sentiment of opinions. In COLING 2004, 20th International Conference on Computa- tional Linguistics, Proceedings of the Conference, 23-27 August 2004, Geneva, Switzerland. Fangtao Li, Nathan Nan Liu, Hongwei Jin, Kai Zhao, Qiang Yang, and Xiaoyan Zhu

  12. [2007]

    In Proceedings of the First Inter- national Conference on Weblogs and Social Media, ICWSM 2007, Boulder, Colorado, USA, March 26- 28,

    Understanding sentiment of peo- ple from news articles: Temporal sentiment analysis of social events. In Proceedings of the First Inter- national Conference on Weblogs and Social Media, ICWSM 2007, Boulder, Colorado, USA, March 26- 28,

  13. [2009]

    In Pro- ceedings of the 18th ACM Conference on Infor- mation and Knowledge Management, CIKM 2009, Hong Kong, China, November 2-6, 2009 , pages 375–384

    Joint senti- ment/topic model for sentiment analysis. In Pro- ceedings of the 18th ACM Conference on Infor- mation and Knowledge Management, CIKM 2009, Hong Kong, China, November 2-6, 2009 , pages 375–384. Bing Liu

  14. [2011]

    InPro- ceedings of the Forth International Conference on Web Search and Web Data Mining, WSDM 2011, Hong Kong, China, February 9-12, 2011 , pages 815–824

    Aspect and sentiment unification model for online review analysis. InPro- ceedings of the Forth International Conference on Web Search and Web Data Mining, WSDM 2011, Hong Kong, China, February 9-12, 2011 , pages 815–824. Jihyeok Kim, Reinald Kim Amplayo, Kyungjae Lee, Sua Sung...

  15. [2012]

    CoRR, abs/1207.0580

    Improving neural networks by preventing co-adaptation of feature detectors. CoRR, abs/1207.0580. Sepp Hochreiter and J ¨urgen Schmidhuber

  16. [2013]

    In Sixth International Joint Conference on Natu- ral Language Processing, IJCNLP 2013, Nagoya, Japan, October 14-18, 2013, pages 1107–1111

    Modeling user leniency and product popularity for sentiment classification. In Sixth International Joint Conference on Natu- ral Language Processing, IJCNLP 2013, Nagoya, Japan, October 14-18, 2013, pages 1107–1111. Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun

  17. [2014]

    In Proceedings of the 52nd An- nual Meeting of the Association for Computational Linguistics, ACL 2014, June 22-27, 2014, Baltimore, MD, USA, System Demonstrations, pages 55–60

    The stanford corenlp natural language processing toolkit. In Proceedings of the 52nd An- nual Meeting of the Association for Computational Linguistics, ACL 2014, June 22-27, 2014, Baltimore, MD, USA, System Demonstrations, pages 55–60. Paul Michel and Graham Neubig

  18. [2015]

    In 3rd Inter- national Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings

    Neural machine translation by jointly learning to align and translate. In 3rd Inter- national Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings. Andrea Baruzzo, Antonina Dattolo, Nirmala Pudota, and Carlo Tasso

  19. [2016]

    In Pro- ceedings of the 2016 Conference on Empirical Meth- ods in Natural Language Processing, EMNLP 2016, Austin, Texas, USA, November 1-4, 2016 , pages 1650–1659

    Neural sentiment classi- fication with user and product attention. In Pro- ceedings of the 2016 Conference on Empirical Meth- ods in Natural Language Processing, EMNLP 2016, Austin, Texas, USA, November 1-4, 2016 , pages 1650–1659. Li Dong, Shaohan Huang, Furu Wei, Mirella Lapa...

  20. [2017]

    In Proceedings of the 2017 Conference on Empirical Methods in Natural Lan- guage Processing, pages 521–526

    Capturing user and product infor- mation for document level sentiment analysis with deep memory network. In Proceedings of the 2017 Conference on Empirical Methods in Natural Lan- guage Processing, pages 521–526. Association for Computational Linguistics. Jessica Ficler and Yo...

  21. [2018]

    In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, Brussels, Belgium, October 31 - November 4, 2018 , pages 3675–3686

    Summa- rizing opinions: Aspect extraction meets sentiment prediction and they are both weakly supervised. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, Brussels, Belgium, October 31 - November 4, 2018 , pages 3675–3686. Dzmitry Bahd...

  22. [2019]

    In 7th International Conference on Learn- ing Representations, ICLR 2019, New Orleans, LA, USA, May 6-9,

    Ordered neurons: In- tegrating tree structures into recurrent neural net- works. In 7th International Conference on Learn- ing Representations, ICLR 2019, New Orleans, LA, USA, May 6-9,

Pith tools

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