|
|
|
Project Type/Research Category: Small Business Innovation Research (SBIR), Phase II.
Funding Priority: Technology for Access and Function.
For more information on NIDRR's funding priorities, read about NIDRR's Core Areas of Research at http://www.ed.gov/rschstat/research/pubs/core-area.html.
Natural Language Speech Access of Computers by the Disabled.
This project has completed its research activities and is now closed. Check REHABDATA for documents.
Amazability, Inc..
14 Lorraine Circle.
Waban, MA 02468.
E-mail: amazability@rcn.com.
URL(s): http://www.amazability.com.
Principal Investigator: Kenneth R. Ingham, PhD.
Public Contact Phone: 617/795-0134.
Project Number: H133S030005.
Start Date: October 1, 2003.
Length: 24 months.
NIDRR Officer: Carol Cohen.
NIDRR Funding: FY 03 $219,595; FY 04 $249,564; FY 05 $0.
Abstract:
This project implements a natural language processor to facilitate complex interactions and to make the overall system readily usable by the computer non-literate or non-technical. The Phase I work confirmed that a speech-based set of complex applications, using automatic speech recognition for command and control, could be effectively used by people with visual impairments. Mechanisms were implemented and tested that allowed users to avoid becoming lost owing to any form of interruption. The work showed that, while simple commands could be successfully recognized, longer complex commands were more difficult because users tended to lose their place while assembling their precise syntax. Therefore, the design and implementation of a natural language processor to support the issuing of complex instructions was indicated. In this phase, the automatic speech recognition subsystem is improved in accordance with the results of the Phase I work and its vocabularies increased to approximately 120,000 words composed of dynamically variable 50,000 word language models. The resulting large vocabulary recognition system, supporting natural language interaction is evaluated through focus group and human subject testing. The Adept 1, equipped with a natural language processor, can serve as a universal reading system for multipart documents (digital, analog, hypertext, etc.). As an integrated voice-based system, it can also support web access, email, and word processing, thereby becoming a universal access tool for reading and writing. By providing very large vocabulary automatic speech recognition and natural language processing of commands, the technical nature of the computer system is minimized.
Descriptors: Assistive technology, Computers, Speech recognition, Visual impairments.
Documents in REHABDATA:
There are no documents related to this project.
|
|
