There is a general vision that in the future, natural language interfaces will provide access to information and services with the simplicity and flexibility required for virtually anyone to participate, eliminating both intellectual and physical barriers, and impacting the way products interface with consumers as well as creating whole new product types. In this project, the PI will test a promising new, innovative natural language processing approach called tridbit technology that may provide a practical way for humans to talk to computers and other devices. Tridbit technology uses unique knowledge structures (called tridbits) and rules for discriminating and combining different types of information to represent the meaning of natural language. The PI's goal is to test whether tridbit technology will hold up in more demanding and realistic environments than previously explored. Various aspects of an existing prototype system will be enhanced and evaluated, including performance in subject areas with varying levels of familiarity and handling of ambiguity and contradictions. Grounding of concepts and constructing natural language to express arbitrary knowledge will also be explored. The technology will be applied to a simulated household environment to control devices such as TVs, VCRs and alarm clocks, with the goal of enabling people to say in their own words what they want a device to do. The PI hopes to demonstrate that natural language interfaces are feasible within a household environment.

Broader Impacts: The technology being tested identifies key constraints within the representation of meaning, potentially simplifying many aspects of the natural language understanding problem. A departure from current paradigm, its underlying theories speak to the way information is structured. If validated, it will enhance our understanding of how natural language works, and could influence a wide range of disciplines. A positive outcome to this project could have tremendous impact for people with physical or cognitive disabilities that prevent them from using standard device controls, and would also prove to manufacturers that they need to prepare for these interfaces of the future.

Agency
National Science Foundation (NSF)
Institute
Division of Information and Intelligent Systems (IIS)
Type
Standard Grant (Standard)
Application #
0443831
Program Officer
Ephraim P. Glinert
Project Start
Project End
Budget Start
2004-11-15
Budget End
2005-10-31
Support Year
Fiscal Year
2004
Total Cost
$123,960
Indirect Cost
Name
University of Wisconsin Madison
Department
Type
DUNS #
City
Madison
State
WI
Country
United States
Zip Code
53715