Olivia Daub’s young son, aged two, has a strong fondness for “doodidees,” which he enthusiastically talks about and demands at an early 5 a.m. each morning. While his words may be unclear to most, Daub is quick to identify his actual desire for blueberries, the tiny, dark blue fruit he craves. Daub, an assistant professor at Western University’s school of communication sciences and disorders in London, Ont., highlights the challenge of comprehending toddler speech, especially for artificial intelligence (AI). She spearheads a new study focused on enhancing AI’s ability to understand toddler language patterns.
Although automatic speech recognition technology has made strides in deciphering adult speech, it still struggles to accurately interpret the speech of young children. Daub emphasizes the need to leverage AI and machine learning to refine recognition of toddlers’ speech nuances. Collaborating with Soodeh Nikan, an assistant professor in Western University’s electrical and computer engineering department, Daub aims to train an AI model to recognize common speech patterns and idiosyncrasies among toddlers.
Nikan points out that existing speech models predominantly trained on adult speech fail to effectively interpret toddler speech, including common errors. The study involves capturing speech data from 30 children through play, storytelling, and interactions with research assistants. Noteworthy patterns, like the substitution of “r” with “w” by many English-speaking toddlers, are documented for AI training.
The research team, having engaged with nine toddlers, continues to seek additional participants. Daub and Nikan envision the AI model developed through this study being utilized in clinical settings to aid speech language pathologists in analyzing and transcribing children’s speech. While perfection may be unattainable, Daub believes significant progress can be made in improving AI’s understanding of preschoolers, potentially enhancing tools like closed captioning and voice-activated accessibility software.
Looking ahead, Daub envisions broader applications of AI in facilitating children’s interaction with technology, emphasizing their potential contributions to society beyond being mere consumers.
