Close mobile menu

Aijun An
Electrical Engineering and Computer Science Department

Professor, EECS Department
Member, ICAL
Member, Centre for Vision Research
Member, BRAIN Alliance

Website | Email

2021 – 2022 Research Highlights

Affect Detection from Text

Automatic detection of affect (such as sentiment, emotion, and sarcasm) from text has become increasingly important as governments and businesses reply on social media data to detect public opinions on policies and products. My team has been working on affect detection and its applications for over a decade. In the past year, three pieces of work were developed to advance the state-of-the-arts in affect detection. First, we formulated the task of sarcasm detection as a sequence classification problem and developed a model, called Emotion Transitions (EmoTrans), that leverages the natural shifts in various emotions over the course of a piece of text to detect sarcasm. Our experiments on two sarcasm datasets (debate corpus and news headlines) demonstrated the potential of using emotion transitions, with the proposed model outperforming several state-of-the-art baseline models. This work was published in ACM SIGIR’20. Second, we proposed a contextual embedding method that extends the architecture of state-of-the-art contextual embedding model, BERT, by incorporating both affective and contextual features of text, and proposed two novel deep neural network models that utilize the affect-extended contextual embedding model for sarcasm detection. Our extensive experiments on various datasets demonstrated that the proposed models outperform state-of-the-art models for sarcasm detection with significant margins. This work was published in COLING’20. Third, we conducted a comprehensive analysis of the role of text preprocessing for word-representation learning for affective tasks and observed that proper text processing methods at the embedding-training phase can significantly improve the downstream affect detection task. This work was published in ACL’20.

Research Highlights

  1. Ameeta Agrawal, Aijun An and Manos Papagelis, Leveraging Transitions of Emotions for Sarcasm Detection, Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR’20), Online, July 25-30, 2020.
  2. Nastaran Babanejad, Heidar Davoudi, Aijun An, Manos Papagelis, Affective and Contextual Embedding for Sarcasm Detection, Proceedings of the 28th International Conference on Computational Linguistics (COLING’20), Online, December 8-13, 2020
  3. Nastaran Babanejad, Ameeta Agrawal, Aijun An and Manos Papagelis, A Comprehensive Analysis of Preprocessing for Word Representation Learning in Affective Tasks, Proceedings of the 2020 Annual Conference of the Association for Computational Linguistics (ACL), Online, July 5-10, 2020.