Fulbright scholar will help computers understand us better
16 March 2012
Research into how computers can better understand human language has won University of Sydney PhD candidate Dominick Ng a Fulbright Scholarship to spend eight months studying at UC Berkeley in the US.
"The world is undergoing an information explosion, and many people struggle to find what they need among the noise," says Dominick, who is currently working on his PhD in the Faculty of Engineering and Information Technologies under the supervision of Dr James Curran.
"We need smarter tools to manage, search, and manipulate information - tools that can comprehend language as humans speak it."
Fulbright scholarships are awarded for a combination of academic merit and leadership potential. The Fulbright Program is part of one of the largest and most prestigious educational exchange programs in the world.
Dominick received his scholarship at a function last night attended by Senator Chris Evans, Federal Minister for Tertiary Education, Skills, Science and Research, and Gabrielle Upton, State Member for Vaucluse.
Dominick's research is investigating ways to improve parsers, software tools that automatically identify syntactic structure.
"Accurately identifying the syntactic structure of language is vital: syntax tells us that 'Alice loves Bob' is different to 'Bob loves Alice'. Parsers can identify structure, but make several unrealistic assumptions about language. The aim of my project is to make parsers read more like humans do, and dramatically improve their accuracy."
State-of-the-art parsers are fully correct on less than half of the sentences they process, which severely restricts their usefulness, Dominick says.
"One problem with parsers is that they consider sentences independently of one another, rather than using the context that a human would. When a new sentence is considered, the information gained from prior ones is discarded - even though this isn't how humans read text. Through my Fulbright project I will develop a new approach to parsing that retains and uses this lost context."
The algorithms that Dominick is developing in his research will have broad applications across existing parsers, and could lead to faster and more accurate information searching.
|Follow University of Sydney Media on Twitter|
Media enquiries: Katie Szittner, 02 9351 2261, 0478 316 809, email@example.com