Olivia Daub’s toddler son has a strong liking for “doodidees,” as he enthusiastically mentions them and demands them every morning at 5 a.m. Daub, recognizing his true desire for blueberries, understands him despite others’ confusion.
According to Daub, comprehending toddler language can be challenging for both adults and artificial intelligence (AI) due to the unique ways in which children communicate. As an assistant professor at Western University’s school of communication sciences and disorders in London, Ont., Daub is leading a new study to enhance AI’s understanding of toddler speech.
Although automatic speech recognition technology has made strides in recognizing adult speech, it struggles to accurately interpret young children’s speech. Daub emphasized the importance of leveraging AI and machine learning to improve speech recognition for toddlers and preschoolers.
Collaborating with Soodeh Nikan, an assistant professor in Western University’s electrical and computer engineering department, Daub aims to train an AI model on toddlers’ speech patterns and nuances. Nikan highlighted the need to train AI models with examples of toddler speech to distinguish between normal speech errors and potential speech disorders.
In the study, 30 children will engage in play activities, storytelling, and conversations with research assistants, whose interactions will be recorded and transcribed. Daub noted a common speech pattern among English-speaking toddlers where they struggle with pronouncing the “r” sound, often substituting it with a “w” sound.
Nikan will utilize the collected data to fine-tune a private AI model specifically designed for understanding toddler speech. The ultimate goal is to develop an AI model that can support speech language pathologists in clinical settings by analyzing and transcribing children’s speech accurately.
Looking ahead, Daub envisions broader applications for the AI model in improving tools like closed captioning and voice-activated accessibility software, while also fostering children’s interaction with technology. Ultimately, the research aims to enhance AI’s ability to comprehend and engage with preschoolers, opening up new possibilities for their contributions to society.