Toward universal semantic understanding of natural languages

Huaiyu Zhu (IBM Research - Almaden), Dulce Ponceleon (IBM Research - Almaden), Yunyao Li (IBM Research - Almaden)

11:55am–12:35pm Wednesday, September 11, 2019

Location: 230 B

Implementing AI, R&D and Innovation

Secondary topics: Health and Medicine, Machine Learning, Text, Language, and Speech

Average rating:

(2.00, 1 rating)

Who is this presentation for?

Researchers and users of natural language processing

Level

Beginner

Description

Understanding the semantics of the natural language is a fundamental task at the core of a wide range of applications and services. Thanks to the relatively high volume and quality of manually labeled English semantic data (e.g., PropBank and FrameNet) and the advanced models built with these labeled data, English semantic understanding has reached a relatively mature state. However, semantic understanding of other languages is much more limited due to the high cost of defining an annotation scheme for semantic understanding and building high-quality labeled data corresponding to the scheme. It’s of great value if the resources available for English can be transferred to other languages. In addition, semantic understanding across different languages captured by a single universal representation provide downstream applications with a single target.

Huaiyu Zhu, Dulce Ponceleon, and Yunyao Li examine how high-quality annotated data for different languages can be created with minimal manual effort. They outline some initial efforts toward universal semantic understanding of natural languages, including the publicly released universal Proposition Bank dataset. You’ll hear use cases of cross-lingual semantics understanding in concrete domains such as finance, healthcare, and compare and comply.

What you'll learn

Gain an introduction to semantic understanding of natural languages
Understand some challenges and approaches for multilingual semantic understanding
Learn the approach of universal semantic understanding and its advantages

Huaiyu Zhu

IBM Research - Almaden

Huaiyu Zhu is a senior software engineer and a member of the Infrastructure for Intelligent Information Systems Group at IBM Research – Almaden. His research interests include natural language processing, machine learning and statistics, and scalable information systems. He’s worked on neural networks, information geometry, text analytics, information extraction, enterprise search, knowledge discovery, enterprise analytics platforms, and multilingual natural language processing. He holds a PhD in computational mathematics and statistics from the University of Liverpool.

Dulce Ponceleon

IBM Research - Almaden

Dulce Ponceleón is a principal research staff member in the Infrastructure for Intelligent Information Systems Group at IBM Research – Almaden. Her broad interests across different disciplines include natural language processing, machine learning, blockchain, and security. She led IBM’s content protection team resulting in significant contributions to Blu-ray Content Protection Standard’s consortium. Previously, she has worked in information retrieval, multimedia content analysis, video summarization, speech recognition, numerical linear algebra, nonlinear programming, storage systems, and content protection, and she was a key contributor to QuickTime Conferencing’s video and audio compression at Apple. She earned her master’s and PhD in computer science from Stanford University and her BS (cum laude) in computer science from Universidad Simon Bolivar in Caracas, Venezuela.

Yunyao Li

IBM Research - Almaden

Yunyao Li is a senior research manager with IBM Research – Almaden, where she manages the Scalable Knowledge Intelligence Department. She’s a Master Inventor and a member of the IBM Academy of Technology. She’s also a member of the New Voices program of the American National Academies. Her expertise is in the interdisciplinary areas of natural language processing, databases, human-computer interaction, and information retrieval. Her contributions in these areas have resulted in 50+ research publications at top AI conferences and journals, 20+ patent filings, and recognition by multiple prestigious IBM internal awards. She received her PhD in computer science and engineering and dual master degrees in computer science and engineering and information science from the University of Michigan and undergraduate degrees from Tsinghua University in Beijing, China. Yunyao is also deeply passionate about improving the diversity for the STEM field. She’s been actively mentoring women and underrepresented minorities both in- and outside of IBM. She currently leads the Almaden Women’s Interest Network Group (AWING) at IBM.