Lunch at 12:30pm, talk at 1pm, in Hesburgh Library, Scholars Lounge

Title: Building Controllable and Efficient Natural Language Generation Systems

Abstract: Large pre-trained language models have enabled rapid progress in natural language generation (NLG). However, existing NLG systems still largely lack control over the content to be generated, and thus suffer from incoherence and unfaithfulness. In this talk, I will first introduce our argument generation work. Understanding, evaluating, and generating arguments are crucial elements of the decision-making and reasoning process. However, constructing persuasive arguments is a challenging task for both human and computers, as it requires credible evidence, rigorous logical reasoning, and sometimes emotional appeals. Our two-step argument generation model separately tackles the challenges of content planning and surface realization. I then discuss how to extend the model to conduct dynamic content planning with mixed language models. Our framework is also generic and has been applied to other text generation problems, such as news article writing. Finally, I will present our recent long document summarization work where efficient attentions are designed to handle more than 10k tokens while prior work can only process hundreds of words.

Bio: Lu Wang is an Assistant Professor of Computer Science and Engineering at University of Michigan. Previously until 2020, she was at Khoury College of Computer Sciences, Northeastern University. She completed her Ph.D. in the Department of Computer Science at Cornell University, under supervision of Professor Claire Cardie in 2015.

Lu’s research is focused on natural language processing, computational social science, and machine learning. More specifically, Lu works on algorithms for text summarization, language generation, argument mining, information extraction, and discourse analysis, as well as novel applications that apply such techniques to understand media bias and polarization and other interdisciplinary subjects. Her work won outstanding short paper award at ACL 2017, and best paper nomination award at SIGDIAL 2012.

Lu’s work has been mainly funded by National Science Foundation (NSF, including a CAREER award), Intelligence Advanced Research Projects Activity (IARPA), and several industry gifts (Tencent AI Lab, ByteDance AI Lab, NVIDIA GPU program, Oracle Research cloud credits, Amazon Web Service credits, Google Cloud Platform credits).