Integrating NLP with ASP for Intelligent Service Robots

--------------------------------------------------------------------
               ***Joint Seminar***
--------------------------------------------------------------------
The Hong Kong University of Science and Technology

Department of Computer Science and Engineering
Human Language Technology Center
--------------------------------------------------------------------

Speaker:	Professor Xiaoping CHEN
		University of Science and Technology of China

Title:		"Integrating NLP with ASP for Intelligent Service Robots"

Date:		Monday, 21 September 2009

Time:		4:00pm - 5:00pm

Venue:		Lecture Theatre F (via lifts 25/26), HKUST


Abstract:

A big challenge to research on service robots is the requirement of
communicating with humans in a natural way, especially for complicated
real-world applications where commands recognition is not sufficient. We
present a first effort to tackle this challenge by integrating Natural
Language Processing (NLP) as a separate module into the software
architecture of a service robot. Coupled with this module, we also
integrated a planning module realized with Answer Set Programming (ASP) in
the architecture, aiming at more scalable and flexible planning capacity.
A prototype system of a real robot has been implemented and tested in a
simple home environment. Standard NLP techniques such as syntactic
parsing, together with our own semantic analyzer, have been implemented in
the prototype system. With these NLP techniques, the user's commands and
descriptions about the environment expressed in English are transferred
into logical formulas in SDRT and finally in ASP, which are fed into the
task planning module, through a 5-steps procedure. ASP technique is
employed for task planning, making it feasible for the robot to conduct
task planning and reasoning about the actions and changes in a unified
way. It is shown in case studies with the prototype system that the robot
can correctly "understand" and make optimal plans for compound commands
under acceptable real-time requirement, while each compound command
consists of multiple atomic actions which may be interconnected with one
other. Some further challenges are also identified on the basis of this
work.


*******************
Biography:

Xiaoping Chen is a professor in School of Computer Science and Technology
at University of Science and Technology of China (USTC). He received a PhD
in CS from USTC, a Master in EE and a BA in Mathematics from Anhui
University. Prof. Chen is currently Director of the Center for Artificial
Intelligence Research, USTC. He established and has led the USTC
Multi-Agent Systems Lab and robot team, WrightEagle, the first Chinese
team that ever entered the international RoboCup competitions, the first
Chinese legged robot team, and the first Chinese standard platform
humanoid robot team, which serve as platforms for researches on
Intelligent Robots and Multi-Agent Systems, especially with DEC-POMDPs,
NLP, and ASP. The team has won 4 champions and 7 runner-ups in RoboCup
world championships. Professor Chen is on the Editorial Board of Journal
of Artificial Intelligence Research and Knowledge Engineering Review. He
has been a trustee of RoboCup Federation since 2008 and was a general
co-chair of RoboCup'08 and PCAR'06, a senior PC member of IJCAI'09, and a
PC member of IJCAI'07, AAAI'05, AAMAS'06-'08, and many other international
conferences.