[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Faculty candidate seminar of possible interest to AIDB folks
Folks:
Here is a seminar by prospective faculty candidate--that will
happen next Monday. Since the talk is related
to AIDB areas--I thought several of you might be interested in going to
the talk.
Rao
"Creation and Coordination of Internet
Agents"
Hasan Davulcu*
State University of New York at Stony Brook
Stony Brook, New York
April 22nd
10:30a.m. 11:30a.m.
GWC room 487
The growing importance of the Web as the
medium for electronic commerce has given rise to a proliferation of
Internet Agents. These are software robots that extract and process
information about products and services from Web sources. Such data,
which requires traversing links and filling out forms, is beyond the
reach of keyword-based search engines. Creating Internet agents and
coordinating a network of such agents to accomplish complex tasks, such
as web-enabled logistics planning and virtual enterprises, are problems
of increasing importance.
In this talk we first discuss the computational aspects underlying the
creation of Internet agents. We show that in general they are
computationally difficult. We then describe practical algorithms for
creating agents semi-automatically based on the idea of ``learning by
examples''. We will demonstrate a tool based on this idea.
Semi-automated techniques are adequate
for building up to several hundreds of agents, but not thousands. The key
observation is that knowledge embedded in the examples required for
semi-automatic methods can sometimes be succinctly captured in a
domain-specific ontology. We will discuss techniques that rely on
such
ontologies for creating agents automatically and thereby significantly
enhancing scalability. We will illustrate the applicability of
these techniques in mining service directories from web sources.
Finally we will discuss issues in modeling and coordinating agents that
are put together to accomplish a complex task. We introduce a novel
logic, based on game theoretic concepts, for modeling interaction among a
number of potentially non-cooperative agents. We will show how it
can be used to model, verify and coordinate efficiently. We will present
algorithms for reasoning using this logic and discuss their
complexity.
*Prospective Faculty
Hasan Davulcu is a Ph.D.
candidate from State University of New York at Stony Brook. He
received his MS in computer science from the State University of New York
at Stony Brook in 1995. He has also been involved in efforts to
develop ontology directed information extraction (IE) systems to
understand the semantics of unstructured data. His work in IE has been
supported by a Phase I, Small Business Innovation Research Grant from
NSF. His work so far has resulted in 7 publications in prestigious
conferences such as ACM SIGMOD, ACM PODS, SIAM Data Mining and several
commercial systems. His principal research area has been Intelligent
Agent and Workflow coordination technologies. You can view Hasan’s
homepage at:
http://www.cs.sunysb.edu/~davulcu.