|
Sense Disambiguation Using On-line Resources
Alpha Luk
Microsoft Institute
When: Friday, 13th May 1994
Time: 4:30pm
Where: Microsoft Institute, Conference Room 1 (near Bistro)
Abstract:
Sense Disambiguation has always been a central issue in NLP. Just like other aspects of AI, sense disambiguation requires a system to have access to a large amount of world knowledge. Earlier systems relied on hand-coding to provide the world knowledge required which allowed the system to perform sense disambiguation in a number of micro toy-domains. The dependence on hand-coding killed off any hopes of these systems' possibilities of being scaled up into real world domains of significant complexity. In the past decade, computer scientists started to look for alternative ways which are to provide the systems with data/knowledge which allow the systems to perform sense disambiguation. One way to obtain such data/knowledge economically is to extract it from machine readable on-line resources. These resources can be divided into two main types: 1.) corpora; 2.) hand-crafted resources which are originally intended for human use, which include all kinds of dictionaries, thesaurus and encyclopedias. In this talk, we will look at a number of different approaches working on each type of resources (including one developed here at MSI) and discuss their advantages and shortcomings.
SPEAKER:
Alpha Luk is a Microsoft Fellow in the final year of his Ph.D. program. His main research interest is on utilization of knowledge in on-line dictionaries. He has published in the Aust. Comp. Sci. Conference and his latest paper has been accepted at the 3rd Pacific Rim International Conference on AI.
Enquiries: Maria Milosavljevic 9850 6345 mariam@mpce.mq.edu.au
| Last modified: July 1997 |