Situated Natural Language Understanding
(Prof. Hozumi Tanaka)
Abstract: To
make our robot more intelligent, it has to understand a conversation
through natural language. The mile stone of such an NLU system was
SHRDLU system developed at MIT more than 30 years ago. NLU research
environment has changed drastically in the past two decade. Not
only tremendous amount of computing power but also better technologies
in speech recognition, natural language processing and computer
graphics are now available. The advancement in such technologies
enables to build a next-generation NLU system, where situated natural
language understanding will be the most important research theme.
In this talk we will explain why the situated natural language
understanding is so important and difficult. Along with showing
some examples, we will give you a few of solutions obtained thus
far through our NLU research.
Hozumi Tanaka was born in Japan on October
02, 1941. He graduated from Department of control engineering, Faculty
of Engineering, Tokyo Institute of Technology in 1964, and completed
his Master’s degree in 1966. He joined Electrotechnical Laboratory
of Ministry of Trade and Industry (MITI) in 1966, where he engaged
in developing a TSS and natural language processing systems, the
latter of which has been his current main research theme. He became
a chief of machine inference section and received his Dr. of Eng.
from Tokyo Institute of Technology in 1980. He joined the Fifth
Generation Computer Project supported by MITI from 1983 to 1992.
He moved from Electrotechnical Laboratory to Tokyo Institute
of Technology in 1983 and retired from the Institute in 2005. Since
April 2001, he has conducted a 5-year project named “Natural Language
Understanding and Action Control” supported by the Ministry
of Education, Culture, Sports, Science and Technology and Japan
Society for the Promotion of Sciences. At present, he is a professor
of Department of Information Science at Chukyo University.
|
|