Carnegie Mellon Robotics Institute
doctoral dissertation, 1999
|The goal of this study was to determine if real-time captions (RTC) benefit both hearing and deaf students and if the format of the display has an effect on verbal comprehension. Comprehension was measured with an established working memory test that has been shown to be sensitive to perception difficulty.
For the first experiment, deaf and hearing subjects were found to have similar abilities to recall written verbal material. RTC produced improved performance for both groups. Analysis of the RTC conditions identified significant effects for hearing type (hearing students performed 9.8% better than deaf students) and the number of captioned lines (four lines were 4.3% better than two lines) but not display location. The conditions with four lines resulted in better performance than those with two lines, especially for the trials with larger memory demands.
In the second experiment, presentation rate (160, 200 wpm) was found to have an impact only during 3 and 4-way interactions including hearing type (deaf, hearing) and the number of lines the captions lagged behind the speaker's voice (0, 1, 2). Performance was seen to improve slightly as sentence lag increased except for deaf subjects at 200 wpm. Under the 160 wpm conditions, it was possible to predict performance of the deaf subjects from the hearing subjects by applying an adjustment factor derived from the first experiment.
Both experiments suggested that, under most scenarios, subject preference was correlated with performance. The exception was that deaf subjects tended to dislike conditions with sentence lag despite higher performance.
Two aggregate factors, buffer and speed, were shown to be effective descriptors of performance. Buffer describes the size of a visual buffer external to the user (e.g., captions) and speed describes how long information takes to be presented, processed, and recalled.
In summary, the manner in which RTC are provided to the user can impact the degree of verbal comprehension. The number of lines, the rate of presentation, and the sentence lag between the captions and the speaker were all found to have an effect. However, location did not show an effect.
|Aaron Steinfeld, "The Benefit to the Deaf of Real-Time Captions in a Mainstream Classroom Environment," doctoral dissertation, 1999|
author = "Aaron Steinfeld",
title = "The Benefit to the Deaf of Real-Time Captions in a Mainstream Classroom Environment",
booktitle = "doctoral dissertation",
publisher = "University of Michigan",
address = "Ann Arbor, MI",
year = "1999",
|The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.|
Contact Us | Update Instructions