Keynote Speaker 2025
Keynote Speaker Ⅰ
Prof. Hwee Tou Ng
Fellow of ACL
National University of Singapore, Singapore
Brief Introduction: Professor Hwee Tou NG is Provost's Chair Professor of Computer Science at the National University of Singapore (NUS). He received a PhD in Computer Science from the University of Texas at Austin, USA. His research focuses on natural language processing. He is a Fellow of the Association for Computational Linguistics (ACL). He has published papers in premier journals and conferences, including Computational Linguistics, Journal of Artificial Intelligence Research (JAIR), ACM Transactions on Information Systems (TOIS), ACL, NAACL, EMNLP, SIGIR, AAAI, and IJCAI. His papers received the Best Paper Award at EMNLP 2011 and SIGIR 1997. He is an associate editor of Journal of Artificial Intelligence Research (JAIR) and an action editor of ACL Rolling Review. He served as the editor-in-chief of Computational Linguistics (July 2018 - January 2024) and ACM Transactions on Asian Language Information Processing (TALIP) (May 2007 - May 2013). He also served as an editorial board member of Computational Linguistics and Journal of Artificial Intelligence Research (JAIR), the book review editor of Computational Linguistics, and an action editor of the Transactions of the Association for Computational Linguistics (TACL). He was an elected member of the ACL executive committee and a former secretary of ACL SIGNLL. He was a program co-chair of EMNLP 2008, ACL 2005, and CoNLL 2004 conferences, and has served as an area chair of ACL, NAACL, EACL, EMNLP, SIGIR, AAAI, and IJCAI conferences and as a session chair and program committee member of many past conferences including ACL, EMNLP, SIGIR, AAAI, and IJCAI.
Speech Title: Just What You Desire: An Agentic Approach to Constrained Timeline Summarization with Self-Reflection
Abstract: Timeline summarization distills coherent event narratives from a large collection of texts, tracing the progression of events and entities over time. In this talk, I will present an agentic approach to timeline summarization, leveraging large language model (LLM) agents. We evaluate our approach on two timeline summarization tasks: clustering a stream of tweets in chronological order into their respective events with abstractive summaries, and extracting the timeline of an entity from a large collection of news articles. In addition, we introduce a novel task called constrained timeline summarization, where the timeline of an entity is generated in which all events in the timeline meet some user-specified constraint. We have collected a new human-verified dataset of constrained timelines, and we propose a novel self-reflection LLM agent that leads to improved performance.
Keynote Speaker Ⅱ
Prof. Yi Cai
South China University of Technology, China
Brief Introduction: Prof. Yi Cai serves as Dean of the School of Software Engineering at South China University of Technology. He is the Director of the Ministry of Education's Key Laboratory of Big Data and Intelligent Robotics. He received a PhD in Computer Science and Engineering from the Chinese University of Hong Kong. His research focuses on natural language processing, knowledge engineering and multi-modal processing. As a Distinguished Member of the China Computer Federation, Prof. Cai participates in several technical committees related to natural language processing and database systems in CCF. He has published more than 230 research papers in peer-reviewed journals and conference proceedings, including IEEE TKDE, IEEE TIP, IEEE TASLP, ACL, SIGIR, AAAI, IJCAI, EMNLP, etc. He has served as an Associate Editor of IEEE TASLP and CMC-Computers, Materials & Continua. He was a general co-chair of APWeb-WAIM 2021, program co-chair of APWEB-WAIM 2018, ICEBE 2021, IEEE DSC 2020 and CCAC 2024 conferences.
Speech Title: Logical-Structure-Aware Natural Language Generation
Abstract: Natural Language Generation (NLG), as a crucial branch of natural language processing, has achieved significant progress in both theoretical and applied aspects in recent years. Its goal is to generate grammatically correct and semantically coherent natural language text from structured or unstructured inputs, with wide-ranging applications in dialogue systems, text summarization, machine translation, and story generation, among others. Although existing generative models have made significant progress in terms of grammar and fluency, they still exhibit deficiencies in the comprehension and application of logical reasoning. The lack of logical capability not only leads to interpretive biases in models but also results in the generation of logically invalid conclusions, posing potential risks in practical applications. In formal logic, the validity of a conclusion depends on the correctness of the logical structure in the reasoning process. Enhancing a model's awareness of logical structures can not only improve its understanding of human reasoning but also further prevent the generation of invalid content. This speech focuses on four key scientific challenges in the critical technologies of natural language generation with logical structure awareness: (1) Logic Structure-Enhanced NLG: To improve understanding of structured logical expressions, we propose SLEtoNL, which transforms logical expressions into tree and graph structures. Using specialized encoders, SLEtoNL captures structural features better than sequential models, achieving more accurate semantic parsing. (2) Implicit Logic Structure Representation: Existing models struggle with implicit logical structures in text. Our Logic Control Framework (LCF) partitions representation spaces into logic-valid and invalid regions via contrastive learning. By adjusting representations, LCF enhances logical validity in generated content. (3) Logic-Structure Guided Generation: Data-driven models often overfit shallow features, ignoring deeper reasoning. The Abstract-level Deductive Reasoner (ADR) addresses this by abstracting concrete content into symbols, forcing the model to focus on logical structures. ADR improves generalization across data distributions. (4) Symbolic Tool-Augmented NLG: To mitigate computational inaccuracies, we integrate symbolic tools with data-driven models. Inputs are parsed into symbolic representations for precise computation, combining neural semantic understanding with symbolic precision.
Keynote Speaker Ⅲ
Assoc. Prof. David Chiang
University of Notre Dame, USA
Brief Introduction: David Chiang (PhD, University of Pennsylvania, 2004) is an associate professor in the Department of Computer Science and Engineering at the University of Notre Dame. His research is on computational models for learning human languages, particularly on connections between formal language theory and natural language, and on speech and language processing for low-resource, endangered, and historical languages. He is the recipient of best paper awards at ACL 2005 and NAACL HLT 2009, and a social impact award and outstanding paper award at ACL 2024. He has received research grants from DARPA, NSF, Google, and Amazon, has served on the executive board of NAACL and the editorial board of Computational Linguistics and JAIR, and is currently on the editorial board of Transactions of the ACL.
Keynote Speaker Ⅳ
Prof. Fang Kong
Soochow University, China
