Generating Comparisons Using a Knowledge Graph in the ILEX System
Maria Milosavljevic
Language Technology Group
Microsoft Research Institute
Macquarie University
When: Tuesday 13th May, 1997
Time: 11:30am
Where: Room E6A102, Macquarie University
Abstract:
In this talk I will describe my current work on the implementation of comparisons in the ILEX system at Edinburgh University. ILEX (the Intelligent Labeling Explorer) is a text generation system which produces descriptions of museum artifacts as hypertext documents. The new version of this system produces descriptions of entities using a knowledge graph. This graph is constructed dynamically from a knowledge base and expresses the inter-connection between entities in the domain. The entities in the current domain include jewellery, designers, styles, materials and places. Since each of these entity types can be the focus of a description, the descriptive strategies used by the system need to be generalised. This can be achieved using a separate mechanism for expressing domain-specific information such as the importance of a particular property type to a particular user group.
A knowledge graph provides a rich representation of the links between concepts, and in particular, it allows us to generalise mechanisms for finding similarities between different types of entities. I am currently implementing mechanisms for generating comparisons between items of jewellery or between designers, styles and so on. In this seminar, I will describe this work and the benefits of using a knowledge graph.
This work is being carried out in the ILEX project at Edinburgh University. The project members include Chris Mellish and Mick O'Donnell from the Department of Artificial Intelligence and Jon Oberlander and Alistair Knott from the Human Communication Research Centre. For more information, see http://www.dai.ed.ac.uk/staff/personal_pages/micko/ILEX/hyperstart2.html
Enquiries: sals@mri.mq.edu.au
| Last modified: July 1997 |