Elamur Research⌕
☰
home / papers / 2020 / Relation-Aware Collaborative Learning for Unified Aspect-Based Sentiment Analysis

Relation-Aware Collaborative Learning for Unified Aspect-Based Sentiment Analysis

Paper · 2020 · Zhuang Chen and Tieyun Qian

Summary

Aspect-based sentiment analysis (ABSA) is typically split into three subtasks: identifying the aspect terms (AE), identifying the opinion terms (OE), and classifying sentiment at the aspect level (SC). Earlier systems either solved these separately and chained them into a pipeline, or used unified models that still ignored most of the ways the subtasks inform each other. Chen and Qian describe four such interactions (R1–R4). For example, an opinion word like "delicious" limits which word can be the aspect it describes.

Their framework, RACL, treats each subtask as word-level sequence labeling and works as follows:

  • A shared fully-connected layer learns features common to all three subtasks.
  • Separate CNN encoders learn task-specific features.
  • A relation propagation step passes clues between the AE, OE and SC modules.
  • Layers can be stacked so the subtasks interact at several semantic levels.

Word vectors come from either GloVe or BERT. On three real-world review datasets, RACL outperforms state-of-the-art pipeline and unified baselines, both on individual subtasks and on complete ABSA, at modest computational cost. The work shows that explicitly modeling cross-task dependencies helps fine-grained sentiment analysis.

Key points

  • Four relations: R1 links AE and OE, R2 connects SC with R1, R3 links SC and OE, and R4 links SC and AE. Previous unified methods used at most R3 and R4. R2 had never been used before.
  • Tagging schemes: AE and OE use B/I/O tags, and SC assigns pos/neu/neg to every word. Multi-task learning follows the shared-private scheme of Collobert and Weston (2008) and Liu et al. (2017).
  • Two implementations: RACL-GloVe and a BERT-based variant. Case studies with RACL-GloVe show R1–R4 correcting errors made by a CMLA+TCap pipeline and by IMN-D.
  • Hyper-parameters:
    • CNN kernel size K=1 performs very poorly, K=3 or 5 works best, and K=7–9 adds noise.
    • Stacking more layers (L) helps only up to a point.
  • Computational cost (Restaurant 2014, batch size 8, one 1080Ti GPU):
    • RACL-GloVe: 5,087,568 parameters, 5s per epoch.
    • IMN-D: 4,129,713 parameters, 5s per epoch.
    • DOER: 9,855,057 parameters, 116s per epoch.

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

🎓Wuhan UniversityWuhan · China7 works · 979 citations

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 9,086 worksArtificial Intelligence · 8,889 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 works

🧭 Topics 1Research subjects and languages the work is about.Sentiment & emotion · 608 works

Cite

Cite as Zhuang Chen and Tieyun Qian. 2020. Relation-Aware Collaborative Learning for Unified Aspect-Based Sentiment Analysis. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2020.acl-main.340

@inproceedings{chen2020relation,
  title = {{Relation-Aware Collaborative Learning for Unified Aspect-Based Sentiment Analysis}},
  author = {Zhuang Chen and Tieyun Qian},
  booktitle = {Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
  year = {2020},
  doi = {10.18653/v1/2020.acl-main.340},
  url = {https://doi.org/10.18653/v1/2020.acl-main.340}
}

References

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

  1. Fangtao Li, Chao Han, Minlie Huang, Xiaoyan Zhu, Yingju Xia, Shu Zhang, and Hao Yu. 2010. Structure-aware review mining and summarization. In COLING, pages 653–661. find
  2. Zhiyuan Chen, Arjun Mukherjee, and Bing Liu. 2014. Aspect extraction with automated prior knowledge learning. In ACL, pages 347–358. find
  3. Zhuang Chen and Tieyun Qian. 2019. Transfer capsule network for aspect level sentiment classification. In ACL, pages 547–556. find
  4. Maryna Chernyshevich. 2014. IHS r&d belarus: Cross-domain extraction of product features using CRF. In SemEval@COLING, pages 309–313. find
  5. Orphée De Clercq, Els Lefever, Gilles Jacobs, Tijl Carpels, and Véronique Hoste. 2017. Towards an integrated pipeline for aspect-based sentiment analysis in various domains. In WASSA@EMNLP, pages 136–142. find
  6. Xin Li, Lidong Bing, Wai Lam, and Bei Shi. 2018a. Transformation networks for target-oriented sentiment classification. In ACL, pages 946–956.
  7. Xin Li, Lidong Bing, Piji Li, and Wai Lam. 2019. A unified model for opinion target extraction and target sentiment prediction. In AAAI, pages 6714–6721. find
  8. Xin Li, Lidong Bing, Piji Li, Wai Lam, and Zhimou Yang. 2018b. Aspect term extraction with history attention and selective transformation. In IJCAI, pages 4194–4200. find
  9. Xin Li and Wai Lam. 2017. Deep multi-task learning for aspect term extraction with memory interaction. In EMNLP, pages 2886–2892. find
  10. Kang Liu, Liheng Xu, and Jun Zhao. 2012. Opinion target extraction using word-based translation model. In EMNLP, pages 1346–1356. find
  11. Zhiqiang Toh and Wenting Wang. 2014. DLIREC: aspect term extraction and term polarity classification system. In SemEval@COLING, pages 235–240. find
  12. Pengfei Liu, Shafiq R. Joty, and Helen M. Meng. 2015. Fine-grained opinion mining with recurrent neural networks and word embeddings. In EMNLP, pages 1433–1443.
  13. Iñaki San Vicente, Xabier Saralegi, and Rodrigo Agerri. 2015. Elixa: A modular and flexible ABSA platform. In SemEval@NAACL-HLT, pages 748–752. find
  14. Pengfei Liu, Xipeng Qiu, and Xuanjing Huang. 2017. Adversarial multi-task learning for text classification. In ACL, pages 1–10.
  15. Qian Liu, Bing Liu, Yuanlin Zhang, Doo Soon Kim, and Zhiqiang Gao. 2016. Improving opinion aspect extraction using semantic similarity and aspect associations. In AAAI, pages 2986–2992. find
  16. Huaishao Luo, Tianrui Li, Bing Liu, and Junbo Zhang. 2019. DOER: dual cross-shared RNN for aspect term-polarity co-extraction. In ACL, pages 591– 601. find
  17. Dehong Ma, Sujian Li, Xiaodong Zhang, and Houfeng Wang. 2017. Interactive attention networks for aspect-level sentiment classification. In IJCAI, pages 4068–4074. find
  18. Margaret Mitchell, Jacqui Aguilar, Theresa Wilson, and Benjamin Van Durme. 2013. Open domain targeted sentiment. In EMNLP, pages 1643–1654.
  19. Saif Mohammad, Svetlana Kiritchenko, and Xiaodan Zhu. 2013. Nrc-canada: Building the state-of-the-art in sentiment analysis of tweets. In NAACL-HLT, pages 321–327.
  20. Jeffrey Pennington, Richard Socher, and Christopher D. Manning. 2014. Glove: Global vectors for word representation. In EMNLP, pages 1532–1543.
  21. Maria Pontiki, Dimitris Galanis, Haris Papageorgiou, Suresh Manandhar, and Ion Androutsopoulos. 2015. Semeval-2015 task 12: Aspect based sentiment analysis. In SemEval, pages 486–495.
  22. Maria Pontiki, Dimitris Galanis, John Pavlopoulos, Harris Papageorgiou, Ion Androutsopoulos, and Suresh Manandhar. 2014. Semeval-2014 task 4: Aspect based sentiment analysis. In SemEval, pages 27–35.
  23. Duy-Tin Vo and Yue Zhang. 2015. Target-dependent twitter sentiment classification with rich automatic features. In IJCAI, pages 1347–1353. find
  24. Feixiang Wang, Man Lan, and Wenting Wang. 2018a. Towards a one-stop solution to both aspect extraction and sentiment analysis tasks with neural multi-task learning. In IJCNN, pages 1–8. find
  25. Shuai Wang, Sahisnu Mazumder, Bing Liu, Mianwei Zhou, and Yi Chang. 2018b. Target-sensitive memory networks for aspect sentiment classification. In ACL, pages 957–967. find
  26. Wenya Wang, Sinno Jialin Pan, Daniel Dahlmeier, and Xiaokui Xiao. 2016. Recursive neural conditional random fields for aspect-based sentiment analysis. In EMNLP, pages 616–626.
  27. Wenya Wang, Sinno Jialin Pan, Daniel Dahlmeier, and Xiaokui Xiao. 2017. Coupled multi-layer attentions for co-extraction of aspect and opinion terms. In AAAI, pages 3316–3322. find
  28. Yuanbin Wu, Qi Zhang, Xuanjing Huang, and Lide Wu. 2009. Phrase dependency parsing for opinion mining. In EMNLP, pages 1533–1541.
  29. Hu Xu, Bing Liu, Lei Shu, and Philip S. Yu. 2018. Double embeddings and cnn-based sequence labeling for aspect extraction. In ACL, pages 592–598.
  30. Yichun Yin, Furu Wei, Li Dong, Kaimeng Xu, Ming Zhang, and Ming Zhou. 2016. Unsupervised word and dependency path embeddings for aspect term extraction. In IJCAI, pages 2979–2985. find
  31. Jianfei Yu, Jing Jiang, and Rui Xia. 2019. Global inference for aspect and opinion terms co-extraction based on multi-task neural networks. TASLP, 27(1):168–177. find
  32. Lei Zhang, Shuai Wang, and Bing Liu. 2018. Deep learning for sentiment analysis: A survey. Wiley Interdiscip. Rev. Data Min. Knowl. Discov., 8(4). find
  33. Ana-Maria Popescu and Oren Etzioni. 2005. Extracting product features and opinions from reviews. In EMNLP, pages 339–346.
  34. Meishan Zhang, Yue Zhang, and Duy-Tin Vo. 2015. Neural networks for open domain targeted sentiment. In EMNLP, pages 612–621. find
  35. Guang Qiu, Bing Liu, Jiajun Bu, and Chun Chen. 2011. Opinion word expansion and target extraction through double propagation. Computational Linguistics, 37(1):9–27.
  36. Peisong Zhu, Zhuang Chen, Haojie Zheng, and Tieyun Qian. 2019. Aspect aware learning for aspect category sentiment analysis. TKDD, 13(6):55:1–55:21. find
  37. Duyu Tang, Bing Qin, and Ting Liu. 2016. Aspect level sentiment classification with deep memory network. In EMNLP, pages 214–224.
  38. Peisong Zhu and Tieyun Qian. 2018. Enhanced aspect level sentiment classification with auxiliary memory. In COLING, pages 1077–1087. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/chen-2020-relation-aware-collaborative-learning-unified/