25-07-2012, 12:10 PM
Survey of Continuous Speech Recognition Software Usability
Survey of Continuous Speech Recognition.pdf (Size: 125.39 KB / Downloads: 61)
Abstract
This research examines the usability of English language continuous speech recognition software programs. Continuous speech recognition programs (CSRP) were brought to market at the end of 1997, with claims that they are capable of recognizing users' continuously spoken speech and translating this directly into text processing software. A comprehensive usability survey questionnaire was administered via the web to collect information from both CSRP users and non-users. Survey respondents were recruited via newsgroup announcements, email referrals, and email-lists. Over a two month period, 162 respondents answered the survey, and 80 were CSRP users. Results from the survey profile patterns of general computer usage and opinions on CSRP. Results from the CSRP users indicates user behavior pattern, the pros and cons of current CSRP, new design opportunities, and levels of user satisfaction. Results from the survey are relevant to ergonomic guidelines for improving interface design for future CSRP.
Introduction
The pace of development in microprocessor development continues to follow Moore's law, and recent generations of computer chips have the power to meet the processing demands of speech recognition. Improvement in computer technology coupled with new approaches to speech recognition algorithms has recently produced several commercially available software programs that claim to support continuous speech recognition. The use of continuous speech may provide for a more natural and user-friendly interface. Continuous speech recognition programs (CSRP) also may obviate the need for extensive use of the computer keyboard and mouse, thereby reducing risks of musculoskeletal discomfort and repetitive strain injuries.
Procedure
A sample of users and non-users of CSRP was obtained by targeting potential respondents with an electronic research request announcement, which described the project and provided a hyperlink to the Web survey. This request was posted to selected newsgroups and email-lists (Table 1). Respondents were also encouraged to forward the request to colleagues.
The first section of the Web survey for all respondents took 3-5 minutes to complete. The second section for CSRP users took around 20 minutes to complete. Respondents received confirmation of successful submissions. Survey data were collected automatically for subsequent statistical analysis.
RESULTS
General Computer Usage
The demographic profile of respondents shows that 67.3% were male; 68.6% were aged 26-50 years; 93% used a computer >3 years; 79% used a computer 6 to 7 days a week; 51.8% used computers 7-9 hours a day; 25.9% use 4-6 hours a day; and 66.7% use a windows (Win95/98/NT) platform.
Eighty four percent indicated that they knew about computer continuous speech recognition programs, such as Dragon NatuallySpeaking, IBM Via Voice, and L & H Voice Xpress (Kurzweil) and 11.7% had not heard of these. 51.2% of respondents think CSRP will replace keyboards for 'composing documents' and 24.0% think CSRP will replace keyboards for 'computer image manipulation'.