Elamur Research⌕
☰
home / papers / 2026 / Label-Consistent Data Generation for Aspect-Based Sentiment Analysis Using LLM Agents

Label-Consistent Data Generation for Aspect-Based Sentiment Analysis Using LLM Agents

Paper · 2026 · Mohammad Hossein Akbari Monfared, Lucie Flek and Akbar Karimi

Summary

Aspect-Based Sentiment Analysis (ABSA) depends on fine-grained aspect-level annotation, which is expensive and often sparse for rare aspect–sentiment combinations. LLM-based augmentation could help, but single-prompt generation frequently produces sentences whose content does not match the intended labels. The authors propose an agentic pipeline built on Qwen2.5-14B in a ReAct-style setup. A generator agent samples aspect–polarity targets from SemEval training data, derives a style policy from real reviews, and writes a candidate sentence. An evaluator agent then checks that every required aspect term appears exactly and that each term carries the correct polarity. Rejected candidates are discarded and regenerated. A matched single-prompt baseline, using the same model and nearly identical instructions, isolates the effect of the agentic structure. Under the InstructABSA framework, the agentic data showed better label consistency, especially for subtasks involving aspect terms. Mixed with real data, it beat prompt-generated data and let T5-Base approach or exceed the Tk-Instruct baseline. This suggests careful augmentation can partly substitute for heavy instruction tuning.

Key points

  • Tasks, data and framework: The evaluation covers three subtasks (ATE, ATSC, ASPE) on four SemEval datasets: Laptop14, Rest14, Rest15 and Rest16. All runs use the InstructABSA pipeline with 20 fine-tuning epochs and scores averaged over three runs.
  • Models fine-tuned: T5-Base and Tk-Instruct-base. They share an architecture, but Tk-Instruct is instruction-tuned on more than 1,600 tasks.
  • Tool-based workflow: The generator uses get policy (samples 1–4 aspect terms with polarities and extracts style) and generate sentences. The evaluator uses label inclusion (exact-term check) and label verifier (an OK/NOT OK polarity judgment).
  • Gains by task and model: Improvements are largest for ATSC and for T5-Base. ASPE benefits little because of its structural complexity, and Tk-Instruct gains less overall.
  • Mixing ratio: Augmentation helps only in moderation. A 1:1 real-to-synthetic mix improves results, while larger synthetic shares reduce performance. Synthetic data alone cannot replace human annotations.

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

 Institutions

🎓University of BonnBonn · Germany10 works · 5,751 citations🔬Lamarr Institute for Machine Learning and Artificial IntelligenceSankt Augustin · Germany6 works · 121 citations🔬Bonn Aachen International Center for Information TechnologyBonn · Germany2 works · 13 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 1Research subjects and languages the work is about.Sentiment & emotion · 972 works

Cite

Cite as Mohammad Hossein Akbari Monfared, Lucie Flek and Akbar Karimi. 2026. Label-Consistent Data Generation for Aspect-Based Sentiment Analysis Using LLM Agents. The Proceedings for the 15th Workshop on Computational Approaches to Subjectivity, Sentiment Social Media Analysis (WASSA 2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.wassa-1.17

@inproceedings{monfared2026label,
  title = {{Label-Consistent Data Generation for Aspect-Based Sentiment Analysis Using LLM Agents}},
  author = {Mohammad Hossein Akbari Monfared, Lucie Flek and Akbar Karimi},
  booktitle = {The Proceedings for the 15th Workshop on Computational Approaches to Subjectivity, Sentiment Social Media Analysis (WASSA 2026)},
  year = {2026},
  doi = {10.18653/v1/2026.wassa-1.17},
  url = {https://doi.org/10.18653/v1/2026.wassa-1.17}
}

References

16 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Jinze Bai, Shuai Bai, Yunfei Chu, Zeyu Cui, Kai Dang, Xiaodong Deng, Yang Fan, Wenbin Ge, Yu Han, Fei Huang, and 1 others. 2023. Qwen technical report. arXiv preprint arXiv:2309.16609.
  2. David Z Chen, Adam Faulkner, and Sahil Badyal. 2022. Unsupervised data augmentation for aspect based sentiment analysis. In Proceedings of the 29th international conference on computational linguistics, pages 6746–6751. find
  3. Chance DeSmet and Diane Cook. 2024. Hydragan: A cooperative agent model for multi-objective data generation. ACM transactions on intelligent systems and technology, 15(3):1–21. find
  4. Ting-Wei Hsu, Chung-Chi Chen, Hen-Hsen Huang, and Hsin-Hsi Chen. 2021a. Semantics-preserved data augmentation for aspect-based sentiment analysis. In Proceedings of the 2021 conference on empirical methods in natural language processing, pages 4417– 4422. find
  5. Ting-Wei Hsu, Chung-Chi Chen, Hen-Hsen Huang, and Hsin-Hsi Chen. 2021b. Semantics-preserved data augmentation for aspect-based sentiment analysis. In Proceedings of the 2021 conference on empirical methods in natural language processing, pages 4417– 4422. find
  6. Akbar Karimi, Leonardo Rossi, and Andrea Prati. 2021a. Adversarial training for aspect-based sentiment analysis with bert. In 2020 25th International conference on pattern recognition (ICPR), pages 8797–8803. IEEE. find
  7. Akbar Karimi, Leonardo Rossi, and Andrea Prati. 2021b. Aeda: An easier data augmentation technique for text classification. arXiv preprint arXiv:2108.13230. find
  8. Z. Li, H. Zhu, Z. Lu, M. Yin, R. Bonatti, S. Abdali, J. Wagle, and K. Koishida. 2023. Synthetic data generation with large language models for text classification: Potential and limitations. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP).
  9. Dancheng Liu, Amir Nassereldine, Ziming Yang, Chenhui Xu, Yuting Hu, Jiajie Li, Utkarsh Kumar, Changjae Lee, Ruiyang Qin, Yiyu Shi, and 1 others. 2024. Large language models have intrinsic self-correction ability. arXiv preprint arXiv:2406.15673. — [arXiv:2406.15673](https://arxiv.org/abs/2406.15673) find
  10. Yu Meng, Jiaxin Huang, Yu Zhang, and Jiawei Han. 2022. Generating training data with language models: Towards zero-shot language understanding. Advances in Neural Information Processing Systems, 35:462–477.
  11. 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
  12. Maria Pontiki, Dimitris Galanis, Haris Papageorgiou, Suresh Manandhar, and Ion Androutsopoulos. 2015. SemEval-2015 Task 12: Aspect Based Sentiment Analysis. In Proceedings of the 9th International Workshop on Semantic Evaluation (SemEval 2015), pages 486–495, Denver, Colorado.
  13. Maria Pontiki, Dimitris Galanis, John Pavlopoulos, Harris Papageorgiou, Ion Androutsopoulos, and Suresh Manandhar. 2014. Semeval-2014 task 4: Aspect based sentiment analysis. In Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 27–35, Dublin, Ireland.
  14. Akshara Prabhakar, Zuxin Liu, Ming Zhu, Jianguo Zhang, Tulika Awalgaonkar, Shiyu Wang, Zhiwei Liu, Haolin Chen, Thai Hoang, Juan Carlos Niebles, and 1 others. 2025. Apigen-mt: Agentic pipeline for multi-turn data generation via simulated agent-human interplay. arXiv preprint arXiv:2504.03601.
  15. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of machine learning research, 21(140):1–67. feedback. In Findings of the assoc…
  16. 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. — [arXiv:2407.00341](https://arxiv.org/abs/2407.00341) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/monfared-2026-label-consistent-data-generation-aspect/