Elamur Research⌕
☰
home / papers / 2016 / Event-Driven Emotion Cause Extraction with Corpus Construction

Event-Driven Emotion Cause Extraction with Corpus Construction

Paper · 2016 · Lin Gui, Dongyin Wu, Ruifeng Xu et al.

Summary

Emotion cause extraction aims to find what triggered an emotion, not just which emotion is expressed. Progress had been limited by three problems: no public data, no formal definition of a cause "event", and annotated sets too small for most learning methods, so earlier work relied on hand-written rules. The authors address all three. They build a Chinese corpus from SINA city news. Emotion keywords are matched in raw articles, and each match keeps up to three clauses of context on each side. Spurious matches are removed, and two annotators label the remaining instances with emotion type and clause-level causes in W3C Emotion Markup Language (EML) format. They then define cause events as a 7-tuple. Candidate events are located through dependency parsing around action words and represented as syntactic trees. A multi-kernel SVM built on convolution tree kernels decides whether each event is a cause. One of these kernels is modified so that synonymous words at leaf nodes can match. This purely learned approach beats the reference methods by 11.6% F-measure, which shows that structured kernels can cope with small training sets.

Key points

  • Raw data: 20,000 SINA city news articles from 2013–2015. A lexicon of 10,259 Chinese emotion keywords yielded 15,687 keyword matches, and 2,105 valid instances remained after filtering.
  • The corpus has 11,799 clauses and 2,167 annotated causes. Multiple causes are rare, appearing in just 59 instances. Nearly all documents (2,046) contain a single emotion.
  • Emotions follow Ekman's six categories (the "Big6" set in W3C EML). Sadness (26.94%) and happiness (25.83%) dominate, and surprise is rare (4.18%).
  • The combined multi-kernel model improves on its component kernels by at least 0.2173. The authors argue it is more than simple voting between the kernels.
  • The error analysis identifies three error types:
    • chains of cascading events;
    • misleading sensory-verb templates;
    • candidates that are never generated, because 6.78% of clauses lack an action word.

Interesting ideas

  • I1. Harvest candidates by keyword, then filter by hand. High-recall keyword matching followed by manual removal of false hits (negation, word-sense ambiguity) is a cheap way to bootstrap an annotated corpus for a niche task. It could suit low-resource languages such as Ukrainian.
  • I2. Use the clause as the annotation unit. Labelling at clause level rather than sentence level keeps connectives and prepositions available as discourse cues. This could help other relation or cause-detection tasks.
  • I3. Inject lexical similarity into tree kernels. Standard convolution tree kernels ignore the words at the leaves. Letting synonyms count as matches adds lexical generalisation without needing more training data, which is useful whenever labelled data is scarce.
  • I4. Combine kernels instead of choosing one. A structural kernel and a lexically enriched kernel performed much better together than either alone. This is a simple design pattern worth testing with other structured representations.
  • I5. Choose a clean genre to isolate the target phenomenon. Formal news text was chosen deliberately, because in informal text emotion causes and emotion targets tend to overlap. Selecting a genre that separates confounded phenomena makes the task definition and evaluation cleaner.

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

🎓Harbin Institute of TechnologyHarbin · China76 works · 27,918 citations🎓Hong Kong Polytechnic UniversityHong Kong · China34 works · 4,476 citations

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 9,085 worksArtificial Intelligence · 8,888 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 worksPsychology · 798 worksExperimental and Cognitive Psychology · 463 works

🧭 Topics 1Research subjects and languages the work is about.Sentiment & emotion · 609 works

Cite

Cite as Lin Gui, Dongyin Wu, Ruifeng Xu, Qin Lu and Yu Zhou. 2016. Event-Driven Emotion Cause Extraction with Corpus Construction. Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/d16-1170

@inproceedings{gui2016event,
  title = {{Event-Driven Emotion Cause Extraction with Corpus Construction}},
  author = {Lin Gui, Dongyin Wu, Ruifeng Xu, Qin Lu and Yu Zhou},
  booktitle = {Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing},
  year = {2016},
  doi = {10.18653/v1/d16-1170},
  url = {https://doi.org/10.18653/v1/d16-1170}
}

References

32 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Daniel Beck, Trevor Cohn, and Lucia Specia. 2014. Joint emotion analysis via multi-task gaussian processes. In EMNLP, pages 1798–1803. find
  2. Yung-Chun Chang, Cen-Chieh Chen, Yu-Lun Hsieh, and WL Hsu. 2015. Linguistic template extraction for recognizing reader-emotion and emotional resonance writing assistance. ACL-IJCNLP, pages 775–780. find
  3. Ying Chen, Sophia Yat Mei Lee, Shoushan Li, and Chu-Ren Huang. 2010. Emotion cause detection with linguistic constructions. In Proceedings of the 23rd International Conference on Computational Linguistics, pages 179–187. Association for Computational Linguistics. find
  4. Michael Collins and Nigel Duffy. 2002. New ranking algorithms for parsing and tagging: Kernels over discrete structures, and the voted perceptron. In Proceedings of the 40th annual meeting on association for computational linguistics, pages 263–270. Association for Computational Linguistics. find
  5. Dipankar Das and Sivaji Bandyopadhyay. 2010. Finding emotion holder from bengali blog texts—an unsupervised syntactic approach. In PACLIC, pages 621–628. find
  6. Paul Ekman. 1984. Expression and the nature of emotion. Approaches to emotion, 3:19–344. find
  7. Wei Gao, Shoushan Li, Sophia Yat Mei Lee, Guodong Zhou, and Chu-Ren Huang. 2013. Joint learning on sentiment and emotion classification. In Proceedings of the 22nd ACM international conference on Conference on information & knowledge management, pages 1505–1508. find
  8. ACM. Kai Gao, Hua Xu, and Jiushuo Wang. 2015. A rule-based approach to emotion cause detection for chinese micro-blogs. Expert Systems with Applications, 42(9):4517–4528. find
  9. Diman Ghazi, Diana Inkpen, and Stan Szpakowicz. 2015. Detecting emotion stimuli in emotion-bearing sentences. In Computational Linguistics and Intelligent Text Processing, pages 152–165. find
  10. Springer. Lin Gui, Li Yuan, Ruifeng Xu, Bin Liu, Qin Lu, and Yu Zhou. 2014. Emotion cause detection with linguistic construction in chinese weibo text. In Natural Language Processing and Chinese Computing, pages 457–464. find
  11. Springer. Takayuki Hasegawa, Nobuhiro Kaji, Naoki Yoshinaga, and Masashi Toyoda. 2013. Predicting and eliciting addressee’s emotion in online dialogue. In ACL (1), pages 964–972. find
  12. Sophia Yat Mei Lee, Ying Chen, and Chu-Ren Huang. 2010. A text-driven rule-based system for emotion cause detection. In Proceedings of the NAACL HLT 2010 Workshop on Computational Approaches to Analysis and Generation of Emotion in Text, pages 45– 53. Association for Computational Linguistics. find
  13. Weiyuan Li and Hua Xu. 2014. Text-based emotion classification using emotion cause extraction. Expert Systems with Applications, 41(4):1742–1749. find
  14. Huanhuan Liu, Shoushan Li, Guodong Zhou, Chu-Ren Huang, and Peifeng Li. 2013. Joint modeling of news reader’s and comment writer’s emotions. In ACL (2), pages 511–515. find
  15. Bing Liu. 2015. Sentiment analysis: Mining opinions, sentiments, and emotions. find
  16. Cambridge University Press. Kun-Hu Luo, Zhi-Hong Deng, Liang-Chen Wei, and Hongliang Yu. 2015. Jeam: A novel model for cross-domain sentiment classification based on emotion analysis. In EMNLP, pages 2503–2508. find
  17. Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg S Corrado, and Jeff Dean. 2013. Distributed representations of words and phrases and their compositionality. In Advances in neural information processing systems, pages 3111–3119.
  18. Saif M Mohammad and Peter D Turney. 2013. Crowdsourcing a word–emotion association lexicon. Computational Intelligence, 29(3):436–465. find
  19. Mitra Mohtarami, Man Lan, and Chew Lim Tan. 2013. Probabilistic sense sentiment similarity through hidden emotions. In ACL (1), pages 983–992. find
  20. Alessandro Moschitti. 2006. Efficient convolution kernels for dependency and constituent syntactic trees. In Machine Learning: ECML 2006, pages 318–329. find
  21. Springer. Gaoyan Ou, Wei Chen, Tengjiao Wang, Zhongyu Wei, Binyang Li, Dongqing Yang, and Kam-Fai Wong. 2014. Exploiting community emotion for microblog event detection. In EMNLP, pages 1159–1168. find
  22. Robert Plutchik. 1980. Emotion: A psychoevolutionary synthesis. find
  23. Ashequl Qadir and Ellen Riloff. 2014. Learning emotion indicators from tweets: Hashtags, hashtag patterns, and phrases. In EMNLP, pages 1203–1209. find
  24. Changqin Quan and Fuji Ren. 2009. Construction of a blog emotion corpus for chinese emotional expression analysis. In Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing: Volume 3-Volume 3, pages 1446–1454. Association for Computational Linguistics. find
  25. Kira Radinsky, Sagie Davidovich, and Shaul Markovitch. 2012. Learning to predict from textual data. Journal of Artificial Intelligence Research, pages 641–684. Irene Russo, Tommaso Caselli, Francesco Rubino, Ester Boldrini, and Patricio Martı́nez-Barco. 2011. Emocause: an easy-adaptable approach to… find
  26. Shashank Srivastava, Dirk Hovy, and Eduard H Hovy. 2013. A walk-based semantically enriched tree kernel over distributed word representations. In EMNLP, pages 1411–1416. find
  27. Jacopo Staiano and Marco Guerini. 2014. Depechemood: a lexicon for emotion analysis from crowd-annotated news. arXiv preprint arXiv:1405.1605. — [arXiv:1405.1605](https://arxiv.org/abs/1405.1605) find
  28. Jonathan H Turner. 2000. On the origins of human emotions: A sociological inquiry into the evolution of human affect. find
  29. Stanford University Press Stanford, CA. Linhong Xu, Hongfei Lin, Yu Pan, Hui Ren, and Jianmei Chen. 2008. Constructing the affective lexicon ontology. Journal of the China Society for Scientific and Technical Information, 27(2):180–185. find
  30. Jun Xu, Ruifeng Xu, Qin Lu, and Xiaolong Wang. 2012. Coarse-to-fine sentence-level emotion classification based on the intra-sentence features and sentential context. In Proceedings of the 21st ACM international conference on Information and knowledge management, pages 2455–2458. find
  31. ACM. Ruifeng Xu, Chengtian Zou, Yanzhen Zheng, Xu Jun, Lin Gui, Bin Liu, and Xiaolong Wang. 2013. A new emotion dictionary based on the distinguish of emotion expression and emotion cognition. Journal of Chinese Information Processing, 27(6):82–90. find
  32. Min Yang, Dingju Zhu, and Kam-Pui Chow. 2014. A topic model for building fine-grained domain-specific emotion lexicon. In ACL (2), pages 421–426. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/gui-2016-event-driven-emotion-cause-extraction/