Working with an Imperfect Medium: Speech Recognition Technology in Reading Practice

摘要

This study focused on the read-aloud and attendant behaviors of learners working with an experimental computer-based program, which made use of speech recognition technology to provide reading practice with immediate feedback for beginning readers. The study drew on data from 13 participants, who used the software as part of their twice-weekly schedule for six to eight weeks at an adult basic education site in New York. Each user's data were analyzed using various interrelated coding and analytical schemes to uncover the nature and value of the user-software interactions. In addition, observations and interviews were conducted. The analyses, based on over 10,000 practice instances and numerous responses to comprehension questions, were done in terms of two broad categories; patterns of interaction, and influence of reading practice afforded by the software. The findings showed that there were metacognitive as well as word level patterns of interaction; that the users were most commonly engaged in attempts at self-correction, and that reading and/or pronunciation tended toward the target as practice ensued. Guidelines were suggested for software improvement and the applicability of speech recognition technology in literacy instruction.