Dynamic Ontology Repair

Project Description

Dynamic ontology repair is a method for facilitating successful communication between agents with disparate ontologies, even where the mismatches between these ontologies may initially lead to communication failure. When an agent experiences communication failure with another agent, it can call our Ontology Repair System (ORS), which will diagnose the particular ontological mismatch between its ontology and the ontology of the other agent that caused the communication problem, and repair the ontology of that agent appropriately. The interaction can then be attempted again, and will this time either be successful or will reveal further mismatches that need to be repaired.

Such abilities are extremely important in large scale systems such as the Semantic Web or the Grid, where huge numbers of users with different aims, locations and backgrounds need to be able to interact and communicate easily and automatically in areas where their interests overlap. The success of such systems depends on them being easy to join and to use. Enforcing tight constraints on potential users is unreasonable; insisting on all agents using identical ontologies will mean that users must laboriously translate their entire ontologies into the required format, and may have to sacrifice a representation that was more suited to their aims. However, communication between agents with mismatched ontologies is extremely difficult. The ability to repair specific parts of ontologies during runtime, and only where necessary, is thus an important aspect of all such systems.

This project investigates this problem in a planning context, where agents are forming plans to achieve goals based on their ontologies, and these plans are sometime inexecutable due to ontological mismatches with other agents. Communication failure is indicated by plan steps failing to be successfully executed, and the repairs can be seen to be efficacious when these plan steps become executable.

Projects based on ORS

There have been several smaller projects looking at extending ORS.

People

The following people are currently working on ORS, or on work related to ORS:

• Fiona McNeill [f.j.mcneill _AT_ ed.ac.uk]

• Alan Bundy [a.bundy_AT_ed.ac.uk]

The following people have worked on ORS in the past:

• Chris Walton

• Marco Schorlemmer

• Morike Akinsola

• Ken Hu

• Theodosia Togia

• Agnieszka Bomersbach

• Andriana Gkaniatsou

• Punyanuch Borwarnginn

MASWS lecture

Running ORS

Selected Publications

Funding