This study examined whether reading fluency assessments could be automated using an indicator of Inter-Words Latency(IWL). A few issues may arise when using state-of-the-art Automated Speech Recognition(ASR). Three research questions were formulated. ...
This study examined whether reading fluency assessments could be automated using an indicator of Inter-Words Latency(IWL). A few issues may arise when using state-of-the-art Automated Speech Recognition(ASR). Three research questions were formulated. First, how accurately ASR recognizes lower-grade children's actual reading utterances? Second, what are the characteristics of children's reading utterances when ASR determines they are incorrect? Third, can IWL be effectively used to automatically evaluate reading fluency of lower-grade children? Speech data of second-grade children reading aloud were collected and analyzed. The result showed a significant difference between the human evaluator and the ASR technology. The human evaluator determined that most of the children’s utterances that the ASR assessed inaccurate were correct. IWL has been shown to have a significant correlation with Words Correct Per Minute(WCPM), the most common indicator for reading fluency. It may be reasonable to use IWL instead of WCPM when it comes to automating the assessments of reading fluency of lower grade children.