Elamur Research
☰
home / papers / 2018 / CARER: Contextualized Affect Representations for Emotion Recognition

CARER: Contextualized Affect Representations for Emotion Recognition

Written for the Research library (claude-opus-5-5) from the paper text. Original — aclanthology.org/D18-1404.pdf

Summary

Emotion in social media text is often expressed through slang, unusual spellings and coded language. Lexicon-based or hand-engineered systems struggle with these variations. Saravia et al. address this with CARER, a semi-supervised approach that treats a whole corpus as a word graph instead of analysing sentences in isolation. They collected large tweet collections from the Twitter API. Subjective tweets were labelled through hashtags (distant supervision), and objective tweets came from news accounts. The objective graph is used to down-weight arcs common in neutral text, yielding an "emotion graph". Graph measures then split tokens into two families: connector words, chosen by eigenvector centrality, and subject words, chosen by clustering coefficient. These are combined into pattern templates, expanded with word embeddings, and weighted per emotion to serve as features for conventional and deep learning classifiers. The authors report that this beats state-of-the-art baselines on emotion recognition. It matters because the patterns need little domain knowledge, handle rare and noisy words, stay useful under resource limits, and are fairly interpretable.

Key points

  • Each source corpus (subjective and objective) holds over 2 million tweets. The subjective set was gathered with 339 hashtags, which are then masked with a <hashtag> placeholder to avoid label leakage.
  • Patterns are built from templates mixing connector words (cw) and subject words (sw), such as <cw, sw> or <sw, cw, sw>, and enriched with word embeddings.
  • In a coverage test on the eight-emotion dataset, a scikit-learn random forest (depth 15, 50 estimators) reached about 65% average F1. Its results stayed fairly stable with only 30% or 10% of the patterns, whereas LIWC, word2vec and tweet2vec degraded much more when reduced.
  • In qualitative examples, the CARER EK model classified short texts, rare words and mixed-emotion posts correctly where DeepMoji and EmoNet failed.
  • Planned future work includes transfer learning, sentiment and sarcasm tasks, and Spanish and Japanese versions.

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

Fields: Natural language processing

Cited by

140 papers of our library cite it (of the 140673 whose bibliographies we read); 4 of them have a page here, 136 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/d18-1404

@inproceedings{saravia2018carer,
  title = {{CARER: Contextualized Affect Representations for Emotion Recognition}},
  author = {Saravia, Elvis and Liu, Hsien-Chi Toby and Huang, Yen-Hao and Wu, Junlin and Chen, Yi-Shin},
  booktitle = {Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing},
  year = {2018},
  doi = {10.18653/v1/d18-1404},
  url = {https://doi.org/10.18653/v1/d18-1404}
}

References

42 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Muhammad Abdul-Mageed and Lyle Ungar. 2017. Emonet: Fine-grained emotion detection with gated recurrent neural networks. In Proc. of ACL. find
  2. Karin Becker, Viviane Moreira, and Aline dos Santos. 2017. Multilingual emotion classification using supervised learning: Comparative experiments. Information Processing & Management, 53(3):684–704. find
  3. John Blitzer, Mark Dredze, and Fernando Pereira. 2007. Biographies, bollywood, boom-boxes and blenders: Domain adaptation for sentiment classification. In Proc. of ACL. find
  4. Piotr Bojanowski, Edouard Grave, Armand Joulin, and Tomas Mikolov. 2017. Enriching word vectors with subword information. Transactions of the Association for Computational Linguistics. find
  5. Y-Lan Boureau, Jean Ponce, and Yann LeCun. 2010. A theoretical analysis of feature pooling in visual recognition. In Proc. of ICML. find
  6. Chun-Hao Chang, Elvis Saravia, and Yi-Shin Chen. 2016. Subconscious crowdsourcing: A feasible data collection mechanism for mental disorder detection on social media. In Advances in Social Networks Analysis and Mining (ASONAM), 2016 IEEE/ACM International Conference on, pages 374–379. IEEE. find
  7. François Chollet et al. 2015. Keras. https:// github.com/keras-team/keras. find
  8. Cindy Chung and James Pennebaker. 2007. The psychological functions of function words. Social communication, 1:343–359. find
  9. Jan Deriu, Aurelien Lucchi, Valeria De Luca, Aliaksei Severyn, Simon Müller, Mark Cieliebak, Thomas Hofmann, and Martin Jaggi. 2017. Leveraging large amounts of weakly supervised data for multi-language sentiment classification. In Proc. of ICWSM. find
  10. Paul Ekman. 1992. An argument for basic emotions. Cognition & emotion, 6(3-4):169–200. find our card
  11. Bjarke Felbo, Alan Mislove, Anders Søgaard, Iyad Rahwan, and Sune Lehmann. 2017. Using millions of emoji occurrences to learn any-domain representations for detecting sentiment, emotion and sarcasm. In Proc of EMNLP. find our card
  12. Jure Leskovec, Anand Rajaraman, and Jeffrey David Ullman. 2014. Mining of massive datasets. Cambridge university press. find
  13. Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781. find
  14. Saif M Mohammad. 2012. Emotional tweets. In Proc. of ACL. find
  15. Saif M Mohammad and Svetlana Kiritchenko. 2015. Using hashtags to capture fine emotion categories from tweets. Computational Intelligence, 31(2):301–326. find
  16. Vinod Nair and Geoffrey Hinton. 2010. Rectified linear units improve restricted boltzmann machines. In Proc. of ICML. find
  17. Febian Pedregosa, Gal Varoquaux, Alexandre Gramfort, Vincent Michel, Bertrand Thirion, Olivier Grisel, Mathieu Blondel, et al. 2011. Scikit-learn: Machine learning in Python. Journal of Machine Learning Research, 12:2825–2830. find
  18. James Pennebaker, Roger Booth, and Martha Francis. 2007. Linguistic inquiry and word count: Liwc [computer software]. Austin, TX: liwc. net. find
  19. Robert Plutchik. 2001. The nature of emotions human emotions have deep evolutionary roots, a fact that may explain their complexity and provide tools for clinical practice. American Scientist, 89(4):344– 350. find
  20. Soujanya Poria, Iti Chaturvedi, Erik Cambria, and Amir Hussain. 2016. Convolutional mkl based multimodal emotion recognition and sentiment analysis. In Proc. of ICDM. find
  21. Ashequl Qadir and Ellen Riloff. 2013. Bootstrapped learning of emotion hashtags\# hashtags4you. In Proc. of Computational Approaches to Subjectivity, Sentiment and Social Media Analysis. find
  22. Alec Go, Richa Bhayani, and Lei Huang. 2009. Twitter sentiment classification using distant supervision. CS224N Project Report, Stanford, 1(2009):12. find
  23. Jonathon Read. 2005. Using emoticons to reduce dependency in machine learning techniques for sentiment classification. In Proc. of ACL. find
  24. Roberto González-Ibánez, Smaranda Muresan, and Nina Wacholder. 2011. Identifying sarcasm in twitter: a closer look. In Proc. of ACL. find our card
  25. Kirk Roberts, Michael Roach, Joseph Johnson, Josh Guthrie, and Sanda Harabagiu. 2012. Empatweet: Annotating and detecting emotions on twitter. In Proc. of LREC. find
  26. Geoffrey Hinton, Nitish Srivastava, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. 2012. Improving neural networks by preventing co-adaptation of feature detectors. arXiv preprint arXiv:1207.0580. find
  27. Sara Rosenthal, Noura Farra, and Preslav Nakov. 2017. SemEval-2017 task 4: Sentiment analysis in Twitter. In Proceedings of the 11th International Workshop on Semantic Evaluation, SemEval ’17, Vancouver, Canada. Association for Computational Linguistics. find our card
  28. Yoon Kim. 2014. Convolutional neural networks for sentence classification. In Proc. of EMNLP. find our card
  29. Leandro dos Santos, Edilson Corrêa Jr, Osvaldo Oliveira Jr, Diego Amancio, Letı́cia Mansur, and Sandra Aluı́sio. 2017. Enriching complex networks with word embeddings for detecting mild cognitive impairment from speech transcripts. arXiv preprint arXiv:1704.08088. find
  30. Diederik Kingma and Jimmy Ba. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980. find
  31. Elvis Saravia, Carlos Argueta, and Yi-Shin Chen. 2015. Emoviz: Mining the world’s interest through emotion analysis. In Advances in Social Networks Analysis and Mining (ASONAM), 2015 IEEE/ACM International Conference on, pages 753–756. IEEE. find
  32. Elvis Saravia, Carlos Argueta, and Yi-Shin Chen. 2016a. Unsupervised graph-based pattern extraction for multilingual emotion classification. Social Network Analysis and Mining, 6(1):92. find
  33. Elvis Saravia, Chun-Hao Chang, Renaud Jollet De Lorenzo, and Yi-Shin Chen. 2016b. Midas: Mental illness detection and analysis via social media. In Advances in Social Networks Analysis and Mining (ASONAM), 2016 IEEE/ACM International Conference on, pages 1418–1421. IEEE. find
  34. Julio Savigny and Ayu Purwarianti. 2017. Emotion classification on youtube comments using word embedding. In Proc. of ICAICTA. find
  35. Valentina Sintsova, Claudiu-Cristian Musat, and Pearl Pu. 2013. Fine-grained emotion recognition in olympic tweets based on human computation. In 4th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis. find
  36. Carlo Strapparava, Alessandro Valitutti, et al. 2004. Wordnet affect: an affective extension of wordnet. In Proc. of LREC. find
  37. Jared Suttles and Nancy Ide. 2013. Distant supervision for emotion classification with discrete binary values. In International Conference on Intelligent Text Processing and Computational Linguistics. find
  38. Svitlana Volkova and Yoram Bachrach. 2016. Inferring perceived demographics from user emotional tone and user-environment emotional contrast. In Proc. of ACL. find
  39. Svitlana Volkova, Theresa Wilson, and David Yarowsky. 2013. Exploring sentiment in social media: Bootstrapping subjectivity clues from multilingual twitter streams. In Proc. of ACL. find
  40. Wenbo Wang, Lu Chen, Krishnaprasad Thirunarayan, and Amit P Sheth. 2012. Harnessing twitter” big data” for automatic emotion identification. In Proc. of SocialCom. find
  41. Joe Ward Jr. 1963. Hierarchical grouping to optimize an objective function. Journal of the American Statistical Association. find
  42. Xiang Zhang, Junbo Zhao, and Yann LeCun. 2015. Character-level convolutional networks for text classification. In Proc. of NIPS. find our card