The Journal of Speech, Language, and Hearing Research just published some very interesting research into the use of speech-recognition technology to assess a child’s vocal development. In an article titled Automated Assessment of Child Vocalization Development Using LENA, the authors explain that they set out to use an automated speech recognition (ASR) system to predict a child’s level of vocal development accurately enough to agree with and support the typical standardized tests we use.
What does it all mean? Well, the authors caution that, while their results were encouraging, their sample size left much to be desired, as it only contained native English speakers from a small geographical area. This means that it is in no way ready to be used in the field as a valid assessment tool. That being said, it is quite exciting to think about the possible clinical implications of this course of study. First of all, the idea that a child’s range of vocalizations and phonemes is representative of their overall language development is a pretty interesting one (no more counting morphemes?!? 😉 ). And speech recognition in the natural environment has long been an exciting frontier that promises the possibility of assessing children’s true skills, not just those they demonstrate in the very weird and un-natural clinical environment. There will never be a replacement for the eyes and ears of a trained clinician, but I find it intriguing and stirring that the capacity of my eyes and ears could be enhanced in the interest of better helping kids.