Empathetic Natural Language Processing

Professor Pascale Fung
Chaired by
Dr KAN Min Yen, Associate Professor, School of Computing

  08 Dec 2017 Friday, 10:00 AM to 11:00 AM

 MR1, COM1-03-19


Can machines understand emotion and intent of humans from natural language? Can understanding emotion and intent help machines understand natural language better? Natural language processing (NLP) deals with words and their meanings. The semantics, or meaning of words, is found to be expressed in terms of word distributions.
Whether words co-occur together points to their relatedness.
Distributional semantics has been used to find the meaning of words since the 1990s. Deep learning algorithms are an efficient approach to learn distributional semantics. However, we show that the emotion and intent of words are not represented in this type of distribution.
Instead, we need to use "emotional semantics" to better understand natural language. In this talk, I will show a novel approach to represent emotional semantics in terms of "emotional embedding" using deep learning, and how we can use it in natural language processing.


Professor Pascale Fung (http://www.ece.ust.hk/~pascale) is a leading expert in the area of speech and language processing and a joint professor of Electronic & Computer Engineering and of Computer Science & Engineering at the Hong Kong University of Science & Technology. She is a Fellow of the Institute of Electrical & Electronic Engineers
(IEEE) and of the International Speech Communication Association. She is the current president and founding board member of the Special Interest Group on Data Driven Methods of the Association for Computational Linguistics, which pioneered empirical methods in natural language processing. Her multidisciplinary work in the past few decades has accumulated in her quest for building "empathetic machines" - AI algorithms that are capable of helping humans better by learning our intent and emotions through different perception signals such as speech, audio, language, and vision. She believes that "empathy" in machines can be learned through data of human-human communications. The perception of emotion and intent, as well as the appropriate response and action can all be learned with machine learning methods. She is also very interested in the challenge of enforcing ethical AI applications. As an expert on the Global Future Council for AI and Robotics of the World Economic Forum, she frequently talks to different stakeholders on the development and applications of AI.