Child Development

Machine Learning Techniques for Analyzing Child Speech



David has spent two years as CTO of Mindsprout, a startup looking at early childhood development. David has spent a lot of time exploring this space, including prototyping and testing several concepts in homes with parents and their toddlers. Some of the app concepts he's built include (1) recording and analyzing audio interactions around reading to provide automatic guidance on reading technique, (2) highly specific activity suggestions that track previous success and failure to suggest progressively more developmentally challenging activities, and (3) an resource in which multiple experts responds to community generated questions about parenting.

Outside of Mindsprout, David worked with Bernd Huber (Harvard IIS PhD) to develop and test a machine learning system to track and provide emotional and behavioral insight into parent-child interactions by analyzing linguistic and paralinguistic features of recordings. They used raw audio and transcripts from the CHILDES database in this work.

David's Machine Learning Paper as part of Roz Picard's Affective Computing class is available here.