PhD student

Room CS2.33

Department of Computer Science,

University of Warwick

Email: Hanqi.Yan@warwick.ac.uk

Google ScholarLink opens in a new window, CV (2022.09)

GitHubLink opens in a new window, TwitterLink opens in a new window

Hanqi Yan (颜寒祺) is a Ph.D. student (2020.10-) under the supervision of Prof. Yulan HeLink opens in a new window, doing Natural Language Processing, with a special focus on sentiment analysis and interpretable NLP models. Now, I am a visiting student in the Machine Learning Department at MBZUAILink opens in a new window, advised by Dr. Kun ZhangLink opens in a new window, focusing on counterfactual generation under identifiablity guarantee. Before Ph.D., I spent a great time visiting Prof. Wenjie LiLink opens in a new window at The Hong Kong Polytechnic University (summer, 2019). I finished my MSc. at Peking UniversityLink opens in a new window (2017-2020) with Prof. Chengqi ChengLink opens in a new window doing spatial data mining and my BEng. at Beihang UniversityLink opens in a new window (2013-2017).

I am focusing on interpretable NLP models and am inspired by the idea of "Align the machine knowledge and human knowledge" from the book [The alignment problemLink opens in a new window]. Generally, two ways to enlarge the overlap between machines and humans and improve mutually:

1) Transform the machine knowledge to be human-friendly. Mostly, I derive probabilistic models guided by Bayesian perspectives, equipped with modern neural architectures to understand the latent features in sentiment analysis. See HINT in document classification.

2) Inject the prior human knowledge to help build the robust and debiasing models, See LexicalAT using WordNet to adversarially train a robust sentiment classifier, KAG using knowledge to enhance emotion cause detection and alleviate the position bias in dataset, and SoftDecay to improve the anisotropic representation generated by large pretrained language model.

Selected Publications (CORE rank A/A*)

• H. Li*, H. Yan*, Y. Li, L. Qian, Y. He and L. Gui. Distinguishability Calibration to In-Context Learning, Findings of EACL, 2023
• R. Zhao, L. Gui, H. Yan and Y. He. Tracking Brand-Associated Polarity-Bearing Topics in User ReviewsLink opens in a new window. Transactions of the Association for Computational Linguistics (TACL).
• H. Yan*, L. Gui*, and Y. He. Hierarchical Interpretation of Neural Text ClassificationLink opens in a new window, Computational Linguistics (CL). Present in EMNLP22.
•  H. Yan, L. Gui, W. Li, and Y. He. Addressing Token Uniformity in Transformers via Singular Value TransformationLink opens in a new window. 38th Conference on Uncertainty in Artificial Intelligence (UAI), Eindhoven, Netherlands, Aug. 2022. [SoftDecay]
• H. Yan, L. Gui, G. Pergola and Y. He. Position Bias Mitigation: A Knowledge-Aware Graph Model for Emotion Cause ExtractionLink opens in a new window, The 59th Annual Meeting of the Association for Computational Linguistics (ACL), Aug. 2021. [KAG] (oral)
• Y. Lei, H Pei, H. Yan, W. Li. Reinforcement learning-based recommendation with graph convolutional q-networkLink opens in a new window, Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), July. 2020.

• J. Xu, L. Zhao, H. Yan, Q. Zeng, Y. Liang, X. Sun. LexicalAT: Lexical-Based Adversarial Reinforcement Training for Robust Sentiment ClassificationLink opens in a new window, Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), Nov. 2019. [LexicalAT]

Teaching