---
title: "E-ABSA20K — paper by Sun et al. (2026) — summary and ideas"
description: "The authors also propose a two-stage extraction method. 77.4% precision. 14 references."
authors: ["Tong Sun", "Mingyang Ma", "Cheng Yu"]
year: 2026
venue: "Findings of the Association for Computational Linguistics: ACL 2026"
source: "https://aclanthology.org/2026.findings-acl.892/"
doi: "10.18653/v1/2026.findings-acl.892"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/sun-2026-e-absa20k-dataset-propose-verify/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/sun-2026-e-absa20k-dataset-propose-verify/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/28/) / E-ABSA20K: A Dataset and Propose-and-Verify for Aspect-Based Sentiment Analysis in Long E-commerce Reviews

# E-ABSA20K: A Dataset and Propose-and-Verify for Aspect-Based Sentiment Analysis in Long E-commerce Reviews

**Authors:** Tong Sun, Mingyang Ma and [Cheng Yu](https://md.research.elamur.ai/free/en/authors/cheng-yu/) · **Year:** 2026 · **Venue:** [Findings of the Association for Computational Linguistics: ACL 2026](https://md.research.elamur.ai/free/en/venues/findings/) ([Findings of the Association for Computational Linguistics 2026](https://md.research.elamur.ai/free/en/venues/findings/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.findings-acl.892/) · [PDF](https://aclanthology.org/2026.findings-acl.892.pdf) · [DOI 10.18653/v1/2026.findings-acl.892](https://doi.org/10.18653/v1/2026.findings-acl.892) · [OpenAlex](https://openalex.org/W7166861683)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
Most ABSA benchmarks consist of short texts from a few domains. Real e-commerce reviews are longer, mention many aspects at once, use informal language, and spread evidence across sentences. To address this gap, the authors built E-ABSA20K: 20,000 English reviews (about 0.8% code-mixed) from a Southeast Asian marketplace, covering WomenBags, Dresses, Cosmetics, and Furniture. Each review is annotated with (target, aspect, opinion, sentiment) quads. Labels come from a weak-supervision pipeline: a human-annotated seed set, four LLM annotators combined by majority voting, and a judge model that resolves disputed candidates.

The authors also propose a two-stage extraction method. First, it generates a broad, recall-oriented set of candidates that must follow the schema. Second, it checks each candidate for textual grounding, scope, and modality, then merges the results so each (target, aspect) pair appears once.

Across several Qwen3 sizes, this approach beats single-pass prompting (with or without chain-of-thought) and other extraction baselines. Gains are largest on difficult cases such as comparisons and conditionals. This makes smaller, cheaper models more practical for review analytics.

## Key points
- **Dataset scale:** E-ABSA20K averages 63.9 words and 6.0 quads per review, much denser than earlier ABSA resources.
- **Aspect inventories:** each category has its own list (WomenBags 34, Dresses 29, Cosmetics 42, Furniture 33). Targets and opinions must be verbatim spans, with at most 20 quads per review.
- **Annotation pipeline:**
  - GPT-4-turbo, Gemini-2.5-Pro, Deepseek-v3.1, and Deepseek-R1 label each review, and votes are aggregated.
  - Gemini-2.5-Pro judges low-consensus candidates, which occur in about 28% of reviews.
  - Against a 760-review human gold set, the labels reach micro P/R/F1 of 0.930/0.810/0.866.
- **Hard-case precision:** with Qwen3-8B, precision on conditionals rises from 0.567 to 0.659 and on comparisons from 0.549 to 0.651. Removing the consolidation step clearly lowers µF.
- **Human evaluation:** on 100 blind-rated reviews, the two-stage system reaches 77.4% precision versus 69.2% for single-stage. Gains persist after SFT, and the Stage 2 verifier can be distilled into a smaller model.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (1): [😊 Sentiment & emotion](https://md.research.elamur.ai/free/en/topics/sentiment-emotion/) · 972 works

## Cite

Cite as: Tong Sun, Mingyang Ma and Cheng Yu. 2026. E-ABSA20K: A Dataset and Propose-and-Verify for Aspect-Based Sentiment Analysis in Long E-commerce Reviews. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{sun2026e,
  title = {{E-ABSA20K: A Dataset and Propose-and-Verify for Aspect-Based Sentiment Analysis in Long E-commerce Reviews}},
  author = {Tong Sun, Mingyang Ma and Cheng Yu},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.892},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.892}
}
```

- [← previous D²Plan: Dual-Agent Dynamic Global Planning for Complex Retrieval-Augmented Reasoning](https://md.research.elamur.ai/free/en/papers/luo-2026-d-2-plan-dual-agent/)
- [next → E-CONAN (Entailment, CONtradition And Neutral) Benchmarks: Arabic Textual Entailment and Natural Inference Datasets](https://md.research.elamur.ai/free/en/papers/jallad-2026-e-conan-entailment-contradition-neutral-benchmarks/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/28/)

## References

14 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D. Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winte… [find](https://scholar.google.com/scholar?q=Tom+B%2E+Brown%2C+Benjamin+Mann%2C+Nick+Ryder%2C+Melanie+Subbiah%2C+Jared+D%2E+Kaplan%2C+Prafulla+Dhariwal%2C+Arvind+Neelakantan%2C+Pranav+Shyam%2C+Girish+Sastry%2C+Amanda+Askell%2C+Sandhini+Agarwal%2C+Ariel+Herbert%2DVoss%2C+Gret)
2. Hongjie Cai, Rui Xia, and Jianfei Yu. 2021. [Aspect-category-opinion-sentiment quadruple extraction with implicit aspects and opinions](https://md.research.elamur.ai/free/en/papers/cai-2021-aspect-category-opinion-sentiment-quadruple/). In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing…
3. Siva Uday Sampreeth Chebolu, Franck Dernoncourt, Nedim Lipka, and Thamar Solorio. 2024a. Oats: Opinion aspect target sentiment quadruple extraction dataset for aspect-based sentiment analysis. Preprint, arXiv:2309.13297. — [arXiv:2309.13297](https://arxiv.org/abs/2309.13297) [find](https://scholar.google.com/scholar?q=Siva+Uday+Sampreeth+Chebolu%2C+Franck+Dernoncourt%2C+Nedim+Lipka%2C+and+Thamar+Solorio%2E+2024a%2E+Oats%3A+Opinion+aspect+target+sentiment+quadruple+extraction+dataset+for+aspect%2Dbased+sentiment+analysis%2E+Preprin)
4. Siva Uday Sampreeth Chebolu, Franck Dernoncourt, Nedim Lipka, and Thamar Solorio. 2024b. Roast: Review-level opinion aspect sentiment target joint detection for absa. Preprint, arXiv:2405.20274. — [arXiv:2405.20274](https://arxiv.org/abs/2405.20274) [find](https://scholar.google.com/scholar?q=Siva+Uday+Sampreeth+Chebolu%2C+Franck+Dernoncourt%2C+Nedim+Lipka%2C+and+Thamar+Solorio%2E+2024b%2E+Roast%3A+Review%2Dlevel+opinion+aspect+sentiment+target+joint+detection+for+absa%2E+Preprint%2C+arXiv%3A2405%2E20274%2E+%E2%80%94+%5Bar)
5. Xuanwen Ding, Jie Zhou, Liang Dou, Qin Chen, Yuanbin Wu, Arlene Chen, and Liang He. 2024. Boosting large language models with continual learning for aspect-based sentiment analysis. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 4367–4377, Miami, Florida, USA. Associ… [find](https://scholar.google.com/scholar?q=Xuanwen+Ding%2C+Jie+Zhou%2C+Liang+Dou%2C+Qin+Chen%2C+Yuanbin+Wu%2C+Arlene+Chen%2C+and+Liang+He%2E+2024%2E+Boosting+large+language+models+with+continual+learning+for+aspect%2Dbased+sentiment+analysis%2E+In+Findings+of+the)
6. Joseph J Peper, Wenzhao Qiu, Ryan Bruggeman, Yi Han, Estefania Ciliotta Chehade, and Lu Wang. 2024. Shoes-ACOSI: A dataset for aspect-based sentiment analysis with implicit opinion extraction. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 15477–15490, Miami, Florida… [find](https://scholar.google.com/scholar?q=Joseph+J+Peper%2C+Wenzhao+Qiu%2C+Ryan+Bruggeman%2C+Yi+Han%2C+Estefania+Ciliotta+Chehade%2C+and+Lu+Wang%2E+2024%2E+Shoes%2DACOSI%3A+A+dataset+for+aspect%2Dbased+sentiment+analysis+with+implicit+opinion+extraction%2E+In+Find)
7. Maria Pontiki, Dimitris Galanis, Haris Papageorgiou, Ion Androutsopoulos, Suresh Manandhar, Mohammad AL-Smadi, Mahmoud Al-Ayyoub, Yanyan Zhao, Bing Qin, Orphée De Clercq, Véronique Hoste, Marianna Apidianaki, Xavier Tannier, Natalia Loukachevitch, Evgeniy Kotelnikov, Nuria Bel, Salud María Jiménez-Z… — [SemEval-2016 Task 5: Aspect Based Sentiment Analysis](https://md.research.elamur.ai/free/en/papers/pontiki-2016-semeval-2016-task-5-aspect/)
8. Maria Pontiki, Dimitris Galanis, John Pavlopoulos, Harris Papageorgiou, Ion Androutsopoulos, and Suresh Manandhar. 2014. [SemEval-2014 task 4: Aspect based sentiment analysis](https://md.research.elamur.ai/free/en/papers/pontiki-2014-semeval-2014-task-4-aspect/). In Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 27–35, Dublin, Ireland. Associa…
9. Paul F. Simmering and Paavo Huoviala. 2023. Large language models for aspect-based sentiment analysis. Preprint, arXiv:2310.18025. — [arXiv:2310.18025](https://arxiv.org/abs/2310.18025) [find](https://scholar.google.com/scholar?q=Paul+F%2E+Simmering+and+Paavo+Huoviala%2E+2023%2E+Large+language+models+for+aspect%2Dbased+sentiment+analysis%2E+Preprint%2C+arXiv%3A2310%2E18025%2E+%E2%80%94+%5BarXiv%3A2310%2E18025%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2310%2E18025%29)
10. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, Cristian Danescu-Niculescu-Mizil, Jim Devlin, Christopher Donahue, William Fedus, Alexey Fodor, Hubert Fort, Naman Goyal, Ankit Gupta, Abhishek… — [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://md.research.elamur.ai/free/en/papers/touvron-2023-llama-2-open-foundation-fine/)
11. Qianlong Wang, Keyang Ding, Hengxin Gao, Hui Wang, and Ruifeng Xu. 2025. Error comparison optimization for large language models on aspect-based sentiment analysis. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 18630– 18646,… [find](https://scholar.google.com/scholar?q=Qianlong+Wang%2C+Keyang+Ding%2C+Hengxin+Gao%2C+Hui+Wang%2C+and+Ruifeng+Xu%2E+2025%2E+Error+comparison+optimization+for+large+language+models+on+aspect%2Dbased+sentiment+analysis%2E+In+Proceedings+of+the+63rd+Annual+M)
12. Chengyan Wu, Bolei Ma, Zheyu Zhang, Ningyuan Deng, Yanqing He, and Yun Xue. 2025. Evaluating zero-shot multilingual aspect-based sentiment analysis with large language models. International Journal of Machine Learning and Cybernetics, 16(10):8079–8101. [find](https://scholar.google.com/scholar?q=Chengyan+Wu%2C+Bolei+Ma%2C+Zheyu+Zhang%2C+Ningyuan+Deng%2C+Yanqing+He%2C+and+Yun+Xue%2E+2025%2E+Evaluating+zero%2Dshot+multilingual+aspect%2Dbased+sentiment+analysis+with+large+language+models%2E+International+Journal+of)
13. Wenxuan Zhang, Yang Deng, Xin Li, Yifei Yuan, Lidong Bing, and Wai Lam. 2021. [Aspect sentiment quad prediction as paraphrase generation](https://md.research.elamur.ai/free/en/papers/zhang-2021-aspect-sentiment-quad-prediction-as/). In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 9209– 9219, Online and Punta Cana, Dominican Republic. Association…
14. Wenxuan Zhang, Xin Li, Yang Deng, Lidong Bing, and Wai Lam. 2022. A survey on aspect-based sentiment analysis: Tasks, methods, and challenges. Preprint, arXiv:2203.01054. [find](https://scholar.google.com/scholar?q=Wenxuan+Zhang%2C+Xin+Li%2C+Yang+Deng%2C+Lidong+Bing%2C+and+Wai+Lam%2E+2022%2E+A+survey+on+aspect%2Dbased+sentiment+analysis%3A+Tasks%2C+methods%2C+and+challenges%2E+Preprint%2C+arXiv%3A2203%2E01054%2E)

*Updated 1 November 2026*