Northwestern University and Hong Kong researchers collaborated on a study to provide insight on the causes and diagnostics of this condition. This method employs machine learning to identify Cantonese and English-like speech patterns in autistic youngsters. Researchers may now be able to distinguish between genetic and environmental influences on the communication skills of autistic persons, allowing them to better understand the disorder's causes and develop novel therapies. Recently, the astounding results were also published in the journal PLOS One.
The research group successfully developed a supervised machine learning method to identify speech variations associated with autism. The training dataset consists of recordings of young children with and without autism narrating in English and Cantonese their interpretations of the events recounted in the wordless children's picture book "Frog, Where Are You?"
Observations indicate that children with autism frequently speak more slowly than typically developing children. Tone, intonation, and rhythm are other distinguishing characteristics of their speech. It is extremely difficult to consistently and objectively explain "prosodic" distinctions, and decades have passed since their origins were established.
Given the structural differences between English and Cantonese, the researchers anticipated that any commonalities identified in the speech patterns of autistic children across both languages were likely related to genetics...
In addition, the researchers saw a spectrum of variance that indicated more malleable parts of speech that could be used as intervention targets. Researchers made significant progress by employing machine learning to identify the speech patterns most symptomatic of autism. Prior to this, they were limited by the bias of the English language in autism research and the subjectivity of humans when it came to categorizing speech discrepancies between individuals with and without autism.
The research led to the identification of speech characteristics, with rhythm emerging as the most crucial in terms of its ability to properly predict an autism diagnosis. The experts have high hopes that this research will increase our understanding of autism. In addition, they believe AI has the potential to facilitate autism diagnosis by reducing the burden on medical professionals and streamlining the process. Additionally, the study seeks to identify and grasp the function of certain genes and brain processing systems associated with an inherited susceptibility to autism.
Their ultimate goal is to better comprehend the factors that contribute to the speech differences of autistic individuals. The team will study if processing differences in the brain contribute to previously identified behavioral speech patterns and their underlying neural genetics. They are quite keen to explore this new terrain further.
-----------
Author: Trevor Kingsley
Tech News CITY // New York Newsroom
