Spoken Language Processing in A Voice-Enabled Procedure Navigator for the International Space Station

Beth Ann Hockey, UC Santa Cruz
Manny Rayner, ICSI

Abstract

Onboard the International Space Station, astronauts execute thousands of complex procedures to maintain life support systems, check out space suits, conduct science experiments and perform medical exams, among their many tasks. Today, when carrying out these procedures, an astronaut reads from paper procedures, or a PDF viewer on a laptop computer, which requires the astronaut to shift attention from the task to scroll PDF pages.

The goal of the Clarissa project has been to develop an experimental voice-operated procedure reader, enabling astronauts to be more efficient with their hands and eyes and give full attention to the task. The prototype version was delivered by a Russian Progress rocket on Christmas Day, 2004; tests are scheduled during the current Expedition. To the best of our knowledge, this will be the first spoken dialogue system ever to be used in space.

The talk will include a demo of the Clarissa prototype and give a tour of the research behind the system, focussing in particular on the following aspects:

  • Converting text procedures into voice-navigable XML documents.
  • Grammar-based speech recognition with the Open Source Regulus toolkit, and comparison with statistical methods.
  • Support Vector Machine methods for "open mic" speech recognition.
  • Side-effect free dialogue management.

(Joint work with Kim Farrell, Nikos Chatzichrisafis, and Jean-Michel Renders.)