Time: 14:00 on 13/8/2018
Place: room B4-705, VIASM (Ta Quang Buu Library, HUST)
Presenter: Thien Huu Nguyen, Department of Computer and Information Science, University of Oregon
Title: Information Extraction in the Era of Deep Learning
Abstract:
Information Extraction (IE) is a branch of Natural Language Processing that aims to locate structured information from unstructured text. The current state-of-the-art approach to IE has involved Deep Learning, a field of machine learning that can automatically induce effective feature representations from data via multiple layers of abstraction.
However, despite the success of deep learning for the IE tasks, the current models have mainly focused on separate tasks in the IE pipeline only (i.e, relation extraction, event detection, entity linking etc). The major limitations of such models include: (i) the inability to capture the inter-dependences among the IE tasks, and (ii) the ignorance of the resources and knowledge from the related tasks.
In this talk, I will present our recent research on joint inference and transfer learning to address such limitations, introducing memory-augmented neural networks and matching networks for IE. These are among the directions that would be the emphasis of future research in this area. Finally, I will highlight some research challenges in natural language understanding and our vision to conquer such challenges in the NLP Lab at the University of Oregon.
Bio:
Thien Huu Nguyen is an assistant professor in the Department of Computer and Information Science at the University of Oregon. His long-term research goal is to create intelligent systems that can achieve a human level of understanding of natural languages. Thien obtained his Ph.D. from New York University under Ralph Grishman and Kyunghyun Cho. His Ph.D. research centers around the development of Deep Learning models for Information Extraction, including Relation Extraction, Event Extraction, Mention Detection, and Slot Filling. In 2018, Thien completed his PostDoc with Yoshua Bengio at the Montreal Institute for Learning Algorithms where he developed multi-modal and multi-task learning systems to understand human languages. Thien was a recipient of the IBM Ph.D. fellowship in 2016.