Pilot Study on User Performance with Speech Recognition Systems
By Koester, Heidi Horstmann; Conaway, M. J.; Rorick, Kristin; Proceedings of the RESNA 25th International Conference, Vol. 22, No. 1, pp. 127-129Publication Date: June/July 2002
Paper presents results of a pilot study on user performance with automatic speech recognition (ASR) systems. The specific objective of the study was to evaluate a measurement method for possible use in a main study. Three areas were assessed: (1) the ease of using the method, (2) the amount of practice necessary for a subject to become proficient with ASR, and (3) how changes to the measurement task might influence user performance. The participant was an able-bodied engineering undergraduate student at the University of Michigan with no prior experience with ASR. The initial training involved the participant going through the standard enrollment process for Dragon Naturally Speaking. The participant then entered text by using ASR for 11 trials under three different text conditions: (1) Simple Transcription, (2) Complex Transition, and (3) Composition. Measures of dictation speed, net text entry rate, and recognition accuracy were calculated for each trial. Results varied, as the dictation speed data showed that the participant spoke at approximately 100 words per minute (wpm), but text entry rate was disparate through simple (60 wpm) and complex (30 wpm) text. Composition rates were the lowest, showing an average of 18 wpm. The authors found that they gained confidence in their measurement method and learned more about its limitations in the study.
Assistive Products Discussed: DRAGON NATURALLYSPEAKING
Published by: Rehabilitation Engineering & Assistive Technology Society of North America (RESNA) (Website:http://www.resna.org)
ISBN: 00-932101-44-5
This publication is included in the library of the National Rehabilitation Information Center (NARIC), accession number O14496
