Balanced Training Data Augmentation for Aspect-Based Sentiment Analysis

Junjie Liu; Yan Song; Yuanhe Tian

REVIEW 3 major objections 5 minor 1 cited by

Balanced Training Data Augmentation for Aspect-Based Sentiment Analysis

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Balanced, RL-optimized data augmentation improves aspect-based sentiment analysis.

desk verdict Useful incremental ABSA augmentation paper; the RL-DPO mechanism mostly works, but the balanced-label claim collapses on REST16 and the paper needs variance and reward-model details before the headline is credible. read the letter →

arxiv 2507.09485 v1 pith:DDR5TQJV submitted 2025-07-13 cs.CL

Junjie Liu , Yuanhe Tian , Yan Song This is my paper

classification cs.CL

keywords aspect-basedsentimentanalysisdataaugmentationreinforcementlearningdirectpreferenceoptimizationlargelanguagemodelsbalancedlabeldistributionLDAtopicrelevanceclassification

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

Aspect-based sentiment analysis decides whether a specific aspect term (for instance 'screen' in a laptop review) is positive, negative, or neutral, and real training sets for it are short, small, and skewed toward positive labels. The paper claims that the fix is in the data, not the model: prompt an LLM to rewrite each training sentence while keeping the aspect term and its sentiment, then train the ABSA model on a merged set whose label distribution has been balanced by oversampling the minority classes. It further claims that the quality of these rewrites can be steered by reinforcement learning, using direct preference optimization on pairs selected by two rewards: sentiment agreement with an untrained ABSA model and topic similarity measured by LDA. In the reported experiments this balanced, RL-tuned augmentation improves both accuracy and F1 over plain augmentation and over most prior systems on four English benchmarks.

What carries the argument

The load-bearing mechanism is a reinforcement-learning loop around the data augmentation LLM. For each original instance, the augmentation LLM samples several rewrites; reward 1 passes each rewrite through an untrained ABSA model and checks whether its predicted sentiment matches the gold label, and reward 2 measures cosine similarity between LDA topic distributions of the original and rewritten sentences. These two signals rank the rewrites into chosen and rejected pairs, with deliberately close pairs to force harder learning, and direct preference optimization (DPO) updates the augmentation LLM on those pairs. The ABSA model is then fine-tuned on the merged original-plus-augmented set, with a 'balanced' variant that first oversamples minority labels to equal class sizes before augmentation.

What would settle it

One concrete test: replace reward 1 with a random or sentiment-blind scoring function and keep everything else fixed; if accuracy and F1 do not drop on LAP14 and REST15, the claim that the sentiment-consistency reward drives the improvement is falsified. A second check is to compare the DPO-tuned generator against the unoptimized generator with test-time quality filtering of the same augmented corpus.

Watch

Extended reading notes

Core claim

The central claim is that a data augmentation LLM can be trained to produce augmented ABSA training instances that are both faithful and varied, and that an ABSA model trained on an augmented, label-balanced corpus generalizes better than one trained on the original unbalanced corpus or on unoptimized augmented data. The paper states this as its approach with reinforcement learning achieving better performance than the model that only uses data augmentation, with balanced label distributions adding a further gain, especially in F1. Its best configuration reports accuracy/F1 of 84.80/82.10 on LAP14, 87.95/81.89 on REST14, 92.54/89.96 on REST15, and 94.61/77.31 on REST16, and the paper says this outperforms most existing studies.

Load-bearing premise

The reward used to build preference pairs assumes that an 'untrained' ABSA model can tell whether a rewritten sentence still carries the original sentiment, but the paper neither defines 'untrained' nor validates that this proxy agrees with true sentiment preservation.

Editorial extensions

If this is right

Training with balanced augmented data raises F1 more than accuracy, so the largest gains should appear for minority sentiment classes rather than for overall correctness.
Because DPO lets the augmentation model generate high-quality data directly, the pipeline removes the need for a separate filtering stage after generation.
Using more than five sampled rewrites per instance for preference construction gives diminishing returns under top-k sampling, so a small sample budget suffices.
Augmented sentences that share transitional or contextual structure with test sentences improve predictions, suggesting the method helps the model learn context-sentiment correspondences.

Reading between the lines

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

A natural extension is to replace the untrained-model sentiment reward with a small trained classifier or an agreement score across several checkpoints; if that works equally well, the method no longer depends on an undefined 'untrained' proxy.
The same two-reward DPO loop could apply to other imbalanced fine-grained text tasks where an LLM rewrites instances, such as stance detection or aspect-level hate speech classification.
Balancing by duplication plus one augmentation per instance may partly conflate oversampling with augmentation quality; a direct test would compare the balanced pipeline against augmentation-only with the same final class sizes.
The case-study evidence suggests that gains come from lexical and structural overlap with test instances, so a follow-up could measure whether a retrieval-based selection of augmentations gives similar gains without reinforcement learning.

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.

Desk Editor's Note

read the letter

Quick take: this is an incremental but real ABSA contribution. The genuinely new piece is using DPO to train the augmentation LLM, with a two-part reward, rather than just prompting and filtering. The authors also test label-balanced augmentation. The RL mechanism mostly checks out: DA+RL beats only DA in F1 on all four datasets, and on LAP14/REST14/REST15 the balanced+RL variant is the best row in Table 4. The code is announced, benchmarks are standard, and ablations are sensible.

The soft spot the stress-test flags is real. On REST16, DA+RL* drops to 77.31 F1, below DA+RL's 81.59 and only DA*'s 81.94. Section 4.1 says balanced training improves F1; that is simply false for the RL-condition on REST16. And Table 5 still lists that 77.31 F1 run as 'Our Best Model' for REST16, so the claim of outperforming most prior work does not hold on that metric/dataset: A2SMvCL, IDG, and dotGCN all report higher F1. The balanced claim needs to be restated as data-dependent or the anomaly explained. It could be an interaction between oversampling and RL on a dataset with very few neutral examples, but the paper does not discuss it.

Second issue: no standard deviations or significance tests anywhere, and only one random 10% dev split. Some of the gaps are small enough, e.g. REST16 DA+RL versus only DA at 81.59 versus 81.30, that noise is a real alternative. This should be tightened with multiple seeds.

Third, reward 1 uses an 'untrained ABSA model' to score generated sentences. Untrained is never defined. If it means a randomly initialized or zero-shot model, the preference signal may just encode the model's prior bias. The authors need to say what the model is and show the signal correlates with human or gold-label correctness.

Minor: LDA topic count K is never given, and the number of DPO augmented instances appears only in a figure. These are fixable.

Bottom line: worth a serious referee, not a desk reject. It is an honest incremental method paper with code and a clear experimental scaffold, but the REST16 contradiction directly undercuts a headline claim. I would send it to review with a request for revision, and I would want the claims trimmed to match the data. For a reading group, it is a decent case study in why a single benchmark can invalidate a general claim.

Referee Report

3 major / 5 minor

Summary. The paper proposes an approach for aspect-based sentiment analysis (ABSA) in which an LLM (Llama-3-8B) rewrites each training sentence conditioned on an aspect term and its gold sentiment label, producing an augmented training set that is optionally balanced by duplicating minority-class instances. A reinforcement learning stage optimizes the augmentation LLM through direct preference optimization (DPO) using a reward composed of two terms: agreement of an 'untrained ABSA model' with the gold sentiment, and cosine similarity between LDA topic distributions of the original and rewritten texts. A Qwen-2.5 (1.5B) model is then fine-tuned on the original plus augmented data. Experiments on LAP14, REST14, REST15, and REST16 compare the baseline, data-augmentation-only, and RL-augmented variants, and report accuracy and F1 against prior work.

Significance. The paper is clearly written, releases code, and uses standard held-out evaluations on four benchmarks, which makes the results easy to check. The idea of using DPO to optimize a data-augmentation LLM with a task-specific reward is a useful step beyond one-shot prompted augmentation and could reduce manual filtering. However, the central empirical claims rest on single runs without significance tests, and one dataset (REST16) contradicts the headline claims. As a result, the current evidence supports a narrower conclusion than stated.

major comments (3)

[Section 4.1, Table 4] The claim that balanced label distributions improve performance, 'especially in the F1 score', is contradicted by REST16: the balanced RL variant DA+RL* reaches 77.31 F1, which is 4.28 points below the unbalanced DA+RL (81.59) and 4.63 points below the balanced only-DA* setting (81.94). On REST16, the balanced setting and the RL component each decrease F1 relative to their direct counterparts. Since no standard deviations or significance tests are reported, the drop cannot be dismissed as noise. Please provide multiple random-seed runs with variance and significance tests, and revise the claim accordingly.
[Section 2.2, reward 1] The 'untrained ABSA model' used to score sentiment consistency is never defined. If it is a randomly initialized model, its predictions are near chance and may be driven by token-level regularities rather than by sentiment; if it is a pretrained model before fine-tuning, its reliability on noisy rewrites is unknown. Because the DPO preference pairs are built directly from this scorer, any systematic bias in the untrained model is optimized into the generator. Please specify the exact model state (e.g., random weights, or pretrained Qwen with no fine-tuning), and validate the proxy, for example by computing agreement with gold labels on a sample of augmented sentences.
[Section 3.3, Table 3] The balanced setting duplicates minority-class original instances to the size of the largest class before augmentation. Consequently, the comparison between standard and balanced settings conflates the effect of adding duplicated original data with the effect of adding synthetic data. The REST16 results (only DA* F1 81.94 versus only DA F1 81.30) may reflect the duplicated originals rather than the balancing mechanism. Please add an ablation that separates duplicated original instances from generated instances, and report the contribution of each to the final model's performance.

minor comments (5)

[Section 4.2] The reward ablation text mislabels the rewards: it says 'only uses the sentiment prediction function (i.e., "reward 2")' later, while reward 1 is defined as sentiment prediction and reward 2 as LDA topic relevance. Please correct the labels so the ablation curves can be interpreted.
[Section 2.2] The paragraph describing the empty chosen or rejected sets contains 'the selected texttc', which appears to be a typesetting error for 'the selected text tc'. Please fix.
[Section 3.4] The random 10% development split is not seeded, and the tuned hyperparameter values are not reported. Please provide the seed and the final hyperparameter choices for reproducibility.
[Section 4.2, Figure 4] The caption and text do not fully define what the values 3, 5, and 8 represent; the text says they are the numbers of generated texts used to build the preference dataset, but the caption says 'different numbers of augmented data for preference dataset construction'. Please add a precise definition in the caption.
[Table 5] On REST16, the reported F1 of 77.31 is lower than several prior systems listed in the same table (e.g., dotGCN at 82.32, A2SMvCL at 83.80, and IDG at 82.83). The phrase 'outperforms most existing studies' is technically true, but the REST16 exception should be acknowledged in the text.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: evaluation is held out and the central comparisons are empirical; the REST16 F1 drop is an inconsistency, not a circular step.

full rationale

The paper's central claims (DA+RL beats only DA; balanced labels improve F1) are empirical comparisons on held-out test splits, not quantities derived from fitted parameters or from the paper's own definitions. Hyperparameters are tuned on a random 10% development subset of the training data and final models are trained on the full training data before evaluation, so the test numbers are not forced by construction. The DPO reward (Section 2.2) uses an 'untrained ABSA model' to score sentiment consistency and an LDA topic-relevance score, but these rewards only shape the generator's preference pairs; the final ABSA model is trained separately on the combined augmented data and evaluated on the official test sets, so there is no equation by which the test predictions equal the reward signal. Self-citations appear in the related work and among baselines, but no load-bearing claim is justified solely by the authors' prior papers, and no uniqueness/ansatz is imported from a self-citation. Therefore, no circular step can be exhibited under the hard rule. The closest issue is not circularity: Table 4 shows DA+RL* F1 77.31 on REST16 versus 81.94 for only DA*, contradicting the Section 4.1 claim that balanced training improves F1; this is an empirical robustness/consistency problem, not a circular derivation.

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

The central claim rests on three unproven modeling assumptions: LLM rewrites preserve labels, the two reward proxies measure quality, and balanced augmentation helps. The last is contradicted on REST16. The method introduces no invented entities and fits only hyperparameters, not the benchmark result.

free parameters (4)

Number of augmented texts per original instance for DPO preference construction = 5 (default; 3 and 8 also evaluated)
Section 4.2 and Figure 4 show 5 chosen because 3 underperforms the baseline; this is tuned on dev performance and directly affects DPO training.
LDA topic number K = not reported
Section 2.2 defines the topic vector over K topics but never states K; the topic similarity reward depends on this unspecified value.
LLM sampling hyperparameters for augmentation generation = not reported
Section 4.2 mentions top-k sampling, but temperature and k are not given; generation diversity and DPO preference pairs depend on them.
Random 10% dev split seed = not reported
Section 3.4 says 10% of training data is used as a dev set for hyperparameter tuning, but no seed is reported, so the selected hyperparameters may not be stable.

assumptions (4)

domain assumption A rewritten sentence generated by Llama-3 with the same aspect term and sentiment instruction can be labeled with the original gold label Y* and used as supervised training data.
Section 2.1 constructs new training instances {X', A, Y*} from LLM output; if rewrites shift meaning or sentiment, the labels are noisy.
domain assumption Topic similarity (LDA cosine) and sentiment-prediction agreement are sufficient reward signals for augmented-data quality.
Section 2.2 builds the full DPO reward from these two proxies; no human evaluation of augmented-data quality is provided.
domain assumption A balanced label distribution in the augmented training set improves ABSA model performance.
Section 4.1 makes this claim, but Table 4 shows REST16 F1 dropping from 81.59 to 77.31 under the balanced setting, so the assumption is not consistently supported.
domain assumption The 'untrained ABSA model' used for reward 1 is a meaningful sentiment predictor.
Section 2.2 first reward; no definition of the untrained state or evidence of its reliability is given.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Balanced Training Data Augmentation for Aspect-Based Sentiment Analysis." pith.science (2026). https://pith.science/paper/DDR5TQJV

@misc{pith2026250709485,
  author       = {Pith},
  title        = {Pith review of: Balanced Training Data Augmentation for Aspect-Based Sentiment Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DDR5TQJV}},
  note         = {Machine review of arXiv:2507.09485}
}

read the original abstract

Aspect-based sentiment analysis (ABSA) is a crucial fine-grained task in social media scenarios to identify the sentiment polarity of specific aspect terms in a sentence. Although many existing studies leverage large language models (LLMs) to perform ABSA due to their strong context understanding capabilities, they still face challenges to learn the context information in the running text because of the short text, as well as the small and unbalanced labeled training data, where most data are labeled with positive sentiment. Data augmentation (DA) is a feasible strategy for providing richer contextual information, especially when using LLMs to create synthetic training data, but faces challenges in ensuring a high quality of the augmented data.In this paper, we propose an LLM-based ABSA approach with training data augmentation.Specifically, an LLM is prompted to generate augmented training data based on the original training data, so as to construct a new training data with larger size and balanced label distributions to better train an ABSA model. Meanwhile, in order to improve the quality of the augmented data, we propose a reinforcement learning approach to optimize the data augmentation. LLM.Experiment results and further analyses on English benchmark datasets for ABSA demonstrate the effectiveness of our approach, where superior performance is observed over strong baselines and most existing studies.

Figures

Figures reproduced from arXiv: 2507.09485 by the authors.

**Figure 2.** Some examples of enhanced data generated [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 4.** The influence of using LLM to generate different numbers of augmented data for preference dataset construction in LAP14 and REST15 datasets [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗

Figures from the paper (1 more)

**Figure 5.** Figure 5: A case study on the data augmentation effect [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

Discussion (0). Sign in 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

Text Reinforcement for Multimodal Time Series Forecasting
cs.CL 2025-08 conditional novelty 6.0 of 10

Reinforcement learning trains an LLM to generate improved text from time series, improving multimodal forecasting on Time-MMD.

Reference graph

Works this paper leans on

63 extracted references · 48 canonical work pages · cited by 1 Pith paper

[1]

Kanwal Ahmed, Muhammad Imran Nadeem, Guanghui Wang, Fang Zuo, and Zhijie Han. 2025. Instruction-tuned absa with auxiliary sentences and knowledge-enhanced graphs for implicit aspect detection. Expert Systems with Applications, 289:128284

work page 2025
[2]

Jiahao Cao, Rui Liu, Huailiang Peng, Lei Jiang, and Xu Bai. 2022. Aspect Is Not You Need: No-aspect Differential Sentiment Framework for Aspect-based Sentiment Analysis . In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1599--1609, Seattle, United States

work page 2022
[3]

Heyan Chai, Ziyi Yao, Siyu Tang, Ye Wang, Liqiang Nie, Binxing Fang, and Qing Liao. 2023. Aspect-to-Scope Oriented Multi-view Contrastive Learning for Aspect-based Sentiment Analysis . In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 10902--10913, Singapore

work page 2023
[4]

Chenhua Chen, Zhiyang Teng, Zhongqing Wang, and Yue Zhang. 2022. Discrete opinion tree induction for aspect-based sentiment analysis. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2051--2064

work page 2022
[5]

Chenhua Chen, Zhiyang Teng, and Yue Zhang. 2020 a . Inducing target-specific latent structures for aspect sentiment classification. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 5596--5607

work page 2020
[6]

Guimin Chen, Yuanhe Tian, and Yan Song. 2020 b . Joint A spect E xtraction and S entiment A nalysis with D irectional G raph C onvolutional N etworks. In Proceedings of the 28th International Conference on Computational Linguistics, pages 272--279

work page 2020
[7]

Suchanek, and Ga \"e l Varoquaux

Lihu Chen, Alexandre Perez-Lebel, Fabian M. Suchanek, and Ga \"e l Varoquaux. 2024. Reconfidencing llms from the grouping loss perspective. ArXiv, abs/2402.04957

arXiv 2024
[8]

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT : P re-training of D eep B idirectional T ransformers for L anguage U nderstanding. In Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) , pages 4171--4...

work page 2019

Show all 63 references

[9]

Xuanwen Ding, Jie Zhou, Liang Dou, Qin Chen, Yuanbin Wu, Chengcai Chen, and Liang He. 2024. Boosting large language models with continual learning for aspect-based sentiment analysis. arXiv preprint arXiv:2405.05496

2024 arXiv
[10]

Angela Fan, Shruti Bhosale, Holger Schwenk, Zhiyi Ma, Ahmed El-Kishky, Siddharth Goyal, Mandeep Baines, Onur Celebi, Guillaume Wenzek, Vishrav Chaudhary, et al. 2021. Beyond english-centric multilingual machine translation. Journal of Machine Learning Research, 22(107):1--48

2021
[11]

Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, et al. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783

2024 arXiv
[12]

Jialong Han, Yan Song, Wayne Xin Zhao, Shuming Shi, and Haisong Zhang. 2018. hyperdoc2vec: Distributed representations of hypertext documents. arXiv preprint arXiv:1805.03793

2018 arXiv
[13]

Nils Constantin Hellwig, Jakob Fehle, and Christian Wolff. 2025. Exploring large language models for the generation of synthetic training samples for aspect-based sentiment analysis in low resource settings. Expert Systems with Applications, 261:125514

2025
[14]

Qingnan Jiang, Lei Chen, Ruifeng Xu, Xiang Ao, and Min Yang. 2019. A C hallenge D ataset and E ffective M odels for A spect- B ased S entiment A nalysis. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint C...

2019
[15]

Kipf and Max Welling

Thomas N. Kipf and Max Welling. 2016. Semi-supervised classification with graph convolutional networks. CoRR, abs/1609.02907

2016 arXiv
[17]

Shichen Li, Zhongqing Wang, Zheyu Zhao, Yue Zhang, and Peifeng Li. 2025. Exploring model editing for llm-based aspect-based sentiment classification. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 24467--24475

2025
[18]

Xin Li, Lidong Bing, Wai Lam, and Bei Shi. 2018. T ransformation N etworks for T arget- O riented S entiment C lassification. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 946--956

2018
[19]

Shuo Liang, Wei Wei, Xian-Ling Mao, Fei Wang, and Zhiyong He. 2022. B i S yn- GAT +: Bi-Syntax Aware Graph Attention Network for Aspect-based Sentiment Analysis . In Findings of the Association for Computational Linguistics: ACL 2022, pages 1835--1848, Dublin, Ireland

2022
[20]

Heng-yang Lu, Tian-ci Liu, Rui Cong, Jun Yang, Qiang Gan, Wei Fang, and Xiao-jun Wu. 2025. Qaie: Llm-based quantity augmentation and information enhancement for few-shot aspect-based sentiment analysis. Information Processing & Management, 62(1):103917

2025
[21]

Meng Luo, Hao Fei, Bobo Li, Shengqiong Wu, Qian Liu, Soujanya Poria, Erik Cambria, Mong-Li Lee, and Wynne Hsu. 2024. Panosent: A panoptic sextuple extraction benchmark for multimodal conversational aspect-based sentiment analysis. In Proceedings of the 32nd ACM International C...

2024
[22]

Yu, and Lijie Wen

Fukun Ma, Xuming Hu, Aiwei Liu, Yawen Yang, Shuang Li, Philip S. Yu, and Lijie Wen. 2023. AMR -based Network for Aspect-based Sentiment Analysis . In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 322--337...

2023
[23]

Qianren Mao, Jianxin Li, Senzhang Wang, Yuanning Zhang, Hao Peng, Min He, and Lihong Wang. 2019. A spect- B ased S entiment C lassification with A ttentive N eural T uring M achines. In Proceedings of the 28th International Joint Conference on Artificial Intelligence, pages 5139--5145

2019
[24]

Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781

2013 arXiv
[25]

Gaurav Negi, Rajdeep Sarkar, Omnia Zayed, and Paul Buitelaar. 2024. A hybrid approach to aspect based sentiment analysis using transfer learning. arXiv preprint arXiv:2403.17254

2024 arXiv
[26]

Maria Pontiki, Dimitrios Galanis, Haris Papageorgiou, Ion Androutsopoulos, Suresh Manandhar, Mohammad Al-Smadi, Mahmoud Al-Ayyoub, Yanyan Zhao, Bing Qin, Orph \'e e De Clercq, et al. 2016. Semeval-2016 task 5: Aspect based sentiment analysis. In International workshop on seman...

2016
[27]

Maria Pontiki, Dimitrios Galanis, Harris Papageorgiou, Suresh Manandhar, and Ion Androutsopoulos. 2015. S emeval-2015 task 12: A spect based sentiment analysis. In Proceedings of the 9th international workshop on semantic evaluation (SemEval 2015), pages 486--495

2015
[28]

Maria Pontiki, Dimitris Galanis, John Pavlopoulos, Harris Papageorgiou, Ion Androutsopoulos, and Suresh Manandhar. 2014. S em E val-2014 task 4: Aspect based sentiment analysis. In Proceedings of the 8th International Workshop on Semantic Evaluation ( S em E val 2014) , pages ...

2014
[29]

Han Qin, Guimin Chen, Yuanhe Tian, and Yan Song. 2021. Improving F ederated L earning for A spect-based S entiment A nalysis via T opic M emories. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 3942--3954, Online and Punta Cana...

2021
[30]

Han Qin, Yuanhe Tian, Fei Xia, and Yan Song. 2022. Complementary L earning of A spect T erms for A spect-based S entiment A nalysis. In Proceedings of the 13th Language Resources and Evaluation Conference

2022
[31]

Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2023. Direct preference optimization: Your language model is secretly a reward model. In Advances in Neural Information Processing Systems, volume 36, pages 53728--53741. Cur...

2023
[32]

Paul F Simmering and Paavo Huoviala. 2023. Large language models for aspect-based sentiment analysis. arXiv preprint arXiv:2310.18025

2023 arXiv
[33]

Yan Song, Chia-Jung Lee, and Fei Xia. 2017. L earning W ord R epresentations with R egularization from P rior K nowledge. In Proceedings of the 21st Conference on Computational Natural Language Learning ( C o NLL 2017) , pages 143--152

2017
[34]

Yan Song and Shuming Shi. 2018. Complementary L earning of W ord E mbeddings. In Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence, IJCAI-18 , pages 4368--4374

2018
[35]

Yan Song, Tong Zhang, Yonggang Wang, and Kai-Fu Lee. 2021. ZEN 2.0: C ontinue T raining and A daption for N -gram E nhanced T ext E ncoders. arXiv preprint arXiv:2105.01279

2021 arXiv
[36]

Hao Tang, Donghong Ji, Chenliang Li, and Qiji Zhou. 2020. Dependency graph enhanced dual-transformer structure for aspect-based sentiment classification. In Proceedings of the 58th annual meeting of the association for computational linguistics, pages 6578--6588

2020
[37]

Siyu Tang, Heyan Chai, Ziyi Yao, Ye Ding, Cuiyun Gao, Binxing Fang, and Qing Liao. 2022. Affective Knowledge Enhanced Multiple-Graph Fusion Networks for Aspect-based Sentiment Analysis . In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing,...

2022
[38]

Yuanhe Tian, Guimin Chen, and Yan Song. 2021. Aspect-based sentiment analysis with type-aware graph convolutional networks and layer ensemble. In Proceedings of the 2021 conference of the North American chapter of the association for computational linguistics: human language t...

2021
[39]

Yuanhe Tian, Weidong Chen, Bo Hu, Yan Song, and Fei Xia. 2023. E nd-to-end A spect-based S entiment A nalysis with C ombinatory C ategorial G rammar. In Findings of the Association for Computational Linguistics: ACL 2023, pages 13597--13609, Toronto, Canada

2023
[40]

Yuanhe Tian, Xu Li, Wei Wang, Guoqing Jin, Pengsen Cheng, and Yan Song. 2025. Large language models enhanced by plug and play syntactic knowledge for aspect-based sentiment analysis. arXiv preprint arXiv:2506.12991

2025 arXiv
[41]

Petar Velickovic, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio’, and Yoshua Bengio. 2017. Graph attention networks. ArXiv, abs/1710.10903

2017 arXiv
[42]

Amir Pouran Ben Veyseh, Nasim Nouri, Franck Dernoncourt, Quan Hung Tran, Dejing Dou, and Thien Huu Nguyen. 2020. I mproving A spect-based S entiment A nalysis with G ated G raph C onvolutional N etworks and S yntax-based R egulation. In Proceedings of the 2020 Conference on Em...

2020
[43]

Kai Wang, Weizhou Shen, Yunyi Yang, Xiaojun Quan, and Rui Wang. 2020. Relational graph attention network for aspect-based sentiment analysis. arXiv preprint arXiv:2004.12362

2020 arXiv
[44]

Qianlong Wang, Keyang Ding, Bin Liang, Min Yang, and Ruifeng Xu. 2023 a . Reducing spurious correlations in aspect-based sentiment analysis with explanation from large language models. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 2930--2941

2023
[45]

Qianlong Wang, Keyang Ding, Bin Liang, Min Yang, and Ruifeng Xu. 2023 b . Reducing Spurious Correlations in Aspect-based Sentiment Analysis with Explanation from Large Language Models . In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 2930--2941,...

2023
[46]

Jason Wei and Kai Zou. 2019. EDA : Easy data augmentation techniques for boosting performance on text classification tasks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language ...

2019
[47]

Xing Wu, Shangwen Lv, Liangjun Zang, Jizhong Han, and Songlin Hu. 2019. Conditional bert contextual augmentation. In Computational Science--ICCS 2019: 19th International Conference, Faro, Portugal, June 12--14, 2019, Proceedings, Part IV 19, pages 84--95. Springer

2019
[48]

Qizhe Xie, Zihang Dai, Eduard Hovy, Minh-Thang Luong, and Quoc V. Le. 2020. Unsupervised data augmentation for consistency training

2020
[49]

Wang, Jiwei Li, Daniel Lévy, Aiming Nie, Dan Jurafsky, and Andrew Y

Ziang Xie, Sida I. Wang, Jiwei Li, Daniel Lévy, Aiming Nie, Dan Jurafsky, and Andrew Y. Ng. 2017. Data noising as smoothing in neural network language models

2017
[50]

Hu Xu, Bing Liu, Lei Shu, and S Yu Philip. 2019. BERT P ost- T raining for R eview R eading C omprehension and A spect-based S entiment A nalysis. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Langua...

2019
[51]

Lu Xu, Lidong Bing, Wei Lu, and Fei Huang. 2020. Aspect sentiment classification with aspect-specific opinion spans. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 3561--3567

2020
[52]

Wei Xue and Tao Li. 2018. A spect B ased S entiment A nalysis with G ated C onvolutional N etworks. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2514--2523

2018
[53]

An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, et al. 2024. Qwen2. 5 technical report. arXiv preprint arXiv:2412.15115

2024 arXiv
[54]

Adams Wei Yu, David Dohan, Quoc Le, Thang Luong, Rui Zhao, and Kai Chen. 2018. Fast and accurate reading comprehension by combining self-attention and convolution. In International conference on learning representations, volume 2

2018
[55]

Chen Zhang, Qiuchi Li, and Dawei Song. 2019. Aspect-based sentiment classification with aspect-specific graph convolutional networks. arXiv preprint arXiv:1909.03477

2019 arXiv
[56]

Mao Zhang, Yongxin Zhu, Zhen Liu, Zhimin Bao, Yunfei Wu, Xing Sun, and Linli Xu. 2023 a . Span-level Aspect-based Sentiment Analysis via Table Filling . In ACL-2023, pages 9273--9284, Toronto, Canada

2023
[57]

Mi Zhang and Tieyun Qian. 2020. Convolution over hierarchical syntactic and lexical graphs for aspect level sentiment analysis. In Proceedings of the 2020 conference on empirical methods in natural language processing (EMNLP), pages 3540--3549

2020
[58]

Xiang Zhang, Junbo Zhao, and Yann LeCun. 2016. Character-level convolutional networks for text classification

2016
[59]

Yice Zhang, Yifan Yang, Bin Liang, Shiwei Chen, Bing Qin, and Ruifeng Xu. 2023 b . An Empirical Study of Sentiment-Enhanced Pre-Training for Aspect-Based Sentiment Analysis . In Findings of the Association for Computational Linguistics: ACL 2023, pages 9633--9651, Toronto, Canada

2023
[60]

Zheng Zhang, Zili Zhou, and Yanna Wang. 2022. SSEGCN : Syntactic and Semantic Enhanced Graph Convolutional Network for Aspect-based Sentiment Analysis . In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human ...

2022
[61]

Qihuang Zhong, Haiyun Li, Luyao Zhuang, Juhua Liu, and Bo Du. 2024. Iterative data generation with large language models for aspect-based sentiment analysis. arXiv preprint arXiv:2407.00341

2024 arXiv
[62]

Changzhi Zhou, Dandan Song, Yuhang Tian, Zhijing Wu, Hao Wang, Xinyu Zhang, Jun Yang, Ziyi Yang, and Shuhao Zhang. 2024 a . A comprehensive evaluation of large language models on aspect-based sentiment analysis. arXiv preprint arXiv:2412.02279

2024 arXiv
[63]

Ting Zhou, Ying Shen, and Yinghui Li. 2024 b . Gcnet: Global-and-context collaborative learning for aspect-based sentiment analysis. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pa...

2024
[64]

Senbin Zhu, Hanjie Zhao, Xingren Wang, Shanhong Liu, Yuxiang Jia, and Hongying Zan. 2024. Zzu-nlp at sighan-2024 dimabsa task: Aspect-based sentiment analysis with coarse-to-fine in-context learning. arXiv preprint arXiv:2407.15341

2024 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Kanwal Ahmed, Muhammad Imran Nadeem, Guanghui Wang, Fang Zuo, and Zhijie Han. 2025. Instruction-tuned absa with auxiliary sentences and knowledge-enhanced graphs for implicit aspect detection. Expert Systems with Applications, 289:128284

work page 2025

[2] [2]

Jiahao Cao, Rui Liu, Huailiang Peng, Lei Jiang, and Xu Bai. 2022. Aspect Is Not You Need: No-aspect Differential Sentiment Framework for Aspect-based Sentiment Analysis . In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1599--1609, Seattle, United States

work page 2022

[3] [3]

Heyan Chai, Ziyi Yao, Siyu Tang, Ye Wang, Liqiang Nie, Binxing Fang, and Qing Liao. 2023. Aspect-to-Scope Oriented Multi-view Contrastive Learning for Aspect-based Sentiment Analysis . In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 10902--10913, Singapore

work page 2023

[4] [4]

Chenhua Chen, Zhiyang Teng, Zhongqing Wang, and Yue Zhang. 2022. Discrete opinion tree induction for aspect-based sentiment analysis. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2051--2064

work page 2022

[5] [5]

Chenhua Chen, Zhiyang Teng, and Yue Zhang. 2020 a . Inducing target-specific latent structures for aspect sentiment classification. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 5596--5607

work page 2020

[6] [6]

Guimin Chen, Yuanhe Tian, and Yan Song. 2020 b . Joint A spect E xtraction and S entiment A nalysis with D irectional G raph C onvolutional N etworks. In Proceedings of the 28th International Conference on Computational Linguistics, pages 272--279

work page 2020

[7] [7]

Suchanek, and Ga \"e l Varoquaux

Lihu Chen, Alexandre Perez-Lebel, Fabian M. Suchanek, and Ga \"e l Varoquaux. 2024. Reconfidencing llms from the grouping loss perspective. ArXiv, abs/2402.04957

arXiv 2024

[8] [8]

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT : P re-training of D eep B idirectional T ransformers for L anguage U nderstanding. In Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) , pages 4171--4...

work page 2019