| Skip to main content | Skip to sub navigation |

This is now an inactive research group it's members have moved on. You can find them at their new research groups:

Best student paper award for ECS at ISWC 2010

Heather giving her paper at ISWC 2010
Heather giving her paper at ISWC 2010
Related Links:

Heather Packer, a research student in the ECS Intelligence, Agents, Multimedia research group, won Best Student Research Paper Award at the prestigious ISWC 2010 conference, held last month in Shanghai.

Heather, who is supervised by Professor Nick Jennings and Dr Nick Gibbins, is working on the development of evolution algorithms for ontologies so that they can provide timely responses where a quick informed decision is required. The award at ISWC (International Semantic Web Conference) 2010 is the second 'best paper' she has been awarded this year, both of which relate to import aspects of her research. The first award (Best Paper), presented at IAT (Intelligent Agent Technology) 2010, held at Toronto, in September, was for a paper on a learning algorithm, which enables agents to incorporate a select set of new concepts into its ontology. The ISWC award was for a forgetting algorithm, which enables the agent to focus the domain of its ontology by removing concepts that impact the agents least.

The problem can be seen by considering an ambulance team trying to save a patient and which needs to decide what course of action to take. Consulting a large knowledge base - an ontology, can have slow response times due to various factors and does not enable the ambulance team to save the patient. Whereas, consulting a subset of the knowledge base may not yield the right results to save the patient.

The ambulance team needs to find a 'good enough' answer quickly enough so that the patient's condition doesn't deteriorate. This work is situated within RoboCup Rescue, a standard multi-agent platform that simulates the aftermath of an earthquake. Using this evolution approach agents are able to make decisions faster, save more civilians and extinguish more fires than other state-of-the-art approaches, because their ontology evolves to provide the information they need in a given timeframe. While this approach does not enable the agents to use complete knowledge, it is not necessary to know everything about concepts in a domain to make an informed decision. For example it is not important to know that your fire vehicle is "red with yellow lettering" if all you want to do is extinguish fires. According to Heather, this is a practical approach and has real impact on the success of the agents.

The papers which won the awards are available in ECS EPrints:

Packer, H. S., Gibbins, N. and Jennings, N. (2010) Collaborative learning of ontology fragments by cooperating agents. In: IEEE/WIC/ACM International Conference on Intelligent Agent Technology, 1-3 September 2010, Toronto, Canada. Available at: http://eprints.ecs.soton.ac.uk/21241/

Packer, H. S., Gibbins, N. and Jennings, N. R. (2010) Forgetting Fragments from Evolving Ontologies. In: International Semantic Web Conference (ISWC), 7-11th November 2010, Shanghai, China. Available at: http://eprints.ecs.soton.ac.uk/21555/

ECS was well represented at ISWC 2010, with one of the conference keynotes being given by dr mc schraefel, also of the IAM group, on: "What does it look like, really? Imagining how citizens might effectively, usefully and easily find, explore, query and re-present Open/Linked data".

__________

If you are interested in undertaking research in the Intelligence, Agents, Multimedia research group of the University of Southampton School of Electronics and Computer Science, you can find out more information on the IAM group pages

For further information on this story contact Joyce Lewis; tel. 023 8059 5453.


Posted by Joyce Lewis on 03 Dec 2010.