 |
U.W. Bangor - School of Informatics - Computer Science Preprints 2005
Artificial Intelligence and Intelligent Agents
|
|
AIIA 05.1 : Al-Dmour, N.A. and Teahan, W.J..
The Blackboard Resource Discovery Mechanism for Distributed Computing over Peer-to-Peer Networks.
Abstract:
Peer-to-Peer computing has emerged in the last few years as an alternative to the traditional
client/server model. The Blackboard Resource Discovery Mechanism (BRDM) facilitates the discovery
of objects (files) in P2P network. We show in this paper that BRDM can also be used to find idle
computers in a Peer-to-Peer network and exploits their CPU cycles to work together on solving a
computational problem. We analyze the performance of BRDM in comparison with other searching
algorithms. The simulations show that BRDM outperforms other algorithms discovering more idle
peers than other approaches with fewer number of query messages.
Download:
WORD document file
Published in:
The IASTED International Conference on Parallel and Distributed Computing and Networks (PDCN)
as part of the 23rd IASTED International Multi-Conference on Applied Informatics,
February 15-17, Innsbruck, Austria, 2005.
AIIA 05.2 : Al-Dmour, N.A. and Teahan, W.J..
Peer-to-peer Protocols for Resource Discovery in the Grid.
Abstract:
One of the basic services in the Grid is resource discovery.
When a user request services, he gives a set of attributes
that should be satisfied by a candidate resource. The resource
discovery mechanism returns a set of best resources
for the given set of attributes. We present two P2P protocols
for resource discovery: the Query Resource Discovery
Mechanism (QRDM), and the Seeking Resource Discovery
Mechanism (SRDM). They are based on the main
characteristics of P2P systems: decentralization and self-organization.
Download:
PDF document file
Published in:
The IASTED International Conference on Parallel and Distributed Computing and Networks (PDCN)
as part of the 23rd IASTED International Multi-Conference on Applied Informatics,
February 15-17, Innsbruck, Austria, 2005.
AIIA 05.3 : Clifton, T. and Teahan, W.J.
Knowing-Aboutness: Question-Answering using a logic-based framework.
Abstract:
We describe the background and motivation for a logic-based
framework, based on the theory of Knowing-Aboutness, and its specicific
application to Question-Answering. We present the salient features of
our system, and outline the benefits of our framework in terms of a more
integrated architecture that is more easily evaluated. Favourable results
are presented in the TREC 2004 Question-Answering evaluation.
Download:
PDF document file
Published in:
27th European Conference on Information Retrieval (ECIR), March 21-23, Santiago de Compostela, Spain, 2005.
Agent-oriented Systems for Teaching, Learning and Research.
Abstract:
We describe the application of agent-oriented systems to teaching, learning
and research (TL&R). Our approach requires the establishment of an environment
called the "Knowledge Web" that facilitates TL&R. We see the Knowledge Web
as a mixed initiative network of collaborative agents who provide mediated
access to a federation of knowledge bases. The network consists of many tutors,
students and researchers, each actively assisting each other to enhance the
quality of the Knowledge Web. Importantly, the key agents in this web are
human-based rather than computer-based - that is, we do not seek to replace
human-based tutors with computer-based ones. Instead, we place an emphasis
on the notion of "agency" (hence why we use the term "agent") - each agent in
the Knowledge Web acts on each other's behalf to enhance the quality of the
web.
Download:
WORD document file
Published in:
AIIA 05.5 : Clifton, T. and Teahan, W.J.
Semi-Automated Evaluation for Question Answering Systems
Abstract:
We discuss the shortcomings of current evaluation techniques
for Question Answering, and present details of a semi-automated method,
based on generating a ground-truth of questions and answers from a source
document collection. We adopt a cascaded named entity tagging system to
generate attribute/value pairs which is then transduced using simple,
configurable, rules to form suitable test questions. We argue that this
approach not only alleviates much of the time burden associated with
preparing evaluation questions, but also allows for a more targeted
appraisal of a system, and is therefore more useful for advancing
development and recognizing areas which require further research.
Download:
PDF document file
Published in:
AIIA 05.6 : ap Cenydd, L. and Teahan, W.J.
Arachnid Simulation: Scaling Arbitray Surfaces
Abstract:
There has been little research done into the realistic simulation of
creatures with the ability to crawl across arbitrary surfaces, clamber
up walls and walk across ceilings. Realistic simulation of such feats
would be of benefit to fiels such as arthropod phobia therapy, the
animation of computer game characters and Artificial Life research.
We have implemented a system than can produce real-time simulation of
a spider traversing across an arbitrary surface. The simulation uses a
combination of a behavioral system, an orienttation system, a procedural
gait generator and an inverse kinematics solver to produce real-time dynamic
animation.
Download:
PDF document file
Published in:
AIIA 05.7 : Clifton, T. and Teahan, W.J.
QITEKAT - Question Inference Tools Employing Knowledgeable Agent
Technologies
Abstract:
We describe the QITEKAT Question Answering system which is a practical
implementation of our logic-based framework for implementing
knowledgeable agents that wil become the core component for our
multi-agent information retrieval systems. A brief overview of the
system architecture is included.
A series of applications are presented in the demonstration, including
both the regular expression and frames-based version of the system, and the
execution of various stages of the information extraction algorithms.
Initially, the systems will be demonstrated using the AQUAINT corpora of
English newswire (as used for TREC). A demonstration of the system's
flexibility will also be provided by coupling it with our Web query
and document extraction system to use the results of user specified
Web search to generate the document set. Finally, we explore the
possibilities of using Question Answering as the basis for document
level IR (as opposed to the other way round) through the use of a
novel framework exploiting the Knows relations generated by QITEKAT
as the index for document searching.
Download:
PS document file
MS Powerpoint document file
Published in:
AIIA 05.8 : Al-Dmour, N.A. and Teahan, W.J..
The Blackboard Resource Discovery Mechanism for Peer-to-Peer Networks.
Abstract:
Peer-to-Peer computing has emerged in the last few years as an alternative to
the traditional client/server model. This paper presents the Blackboard
Resource Discovery Mechanism (BRDM) for peer-to-peer networks, a novel search
algorithm for unstructured peer-to-peer networks. BRDM facilitates the
discovery of objects (files) in a P2P network. The simulation results show
that BRDM outperforms other existing search algorithms in the following ways:
it supports the discovery of objects distributed on a network as a whole
rather than on a given part of it; it achieves more successful matches than
other searching algorithms; it achieves reduced bandwidth consumption; and it
is more adaptable in dynamic P2P networks.
Download:
PDF document file
Published in: