Summary: Machine learning algorithms help researchers identify speech patterns in children with autism spectrum disorders.
A source: Northwestern University
A new study led by researchers at Northwestern University used machine learning to identify speech patterns that matched between English and Cantonese in children with autism – artificial intelligence.
A study conducted with partners in Hong Kong provided researchers with insights into the genetic and environmental factors that shape the communication skills of people with autism, which will help them learn more about the origin of the disease and develop new therapies.
Children with autism typically speak more slowly than developing children and show differences in tone, intonation, and rhythm. However, these differences (which researchers call “prosodic differences”) are surprisingly difficult to characterize consistently and objectively, and their origins have remained unknown for decades.
However, a team of researchers led by northwestern scientists Molly Losh and Joseph Si Lau, along with Hong Kong-based Patrick Wong and his team, successfully used controlled machine learning to identify speech-related differences in autism.
The data used to teach the algorithm were “Frog, where are you?” The children definitely had entries telling their own version of the story described in the picture book.
The results were published in the journal PLOS One June 8, 2022.
“When you have languages that are so structurally different, all the similarities in speech patterns observed in autism in both languages may be features that have a strong genetic response to autism,” said Losh, Joe. Anne G. and Peter F. Dolle, Professor of Northwestern Learning Disabilities.
“But the variability we’ve observed is even more interesting. It can show the flexible features of speech and good intentions for intervention.”
Lau added that the use of machine learning to identify key elements of speech that predict autism is a significant step forward for researchers who are limited by human subjectivity in classifying English one-sidedness and speech differences in the study of autism. among people with autism.
“Using this approach, we were able to identify speech features that could prevent the diagnosis of autism,” said Loch, a postdoctoral researcher at the Roxelin and Richard Pepper Department of Communication Sciences and Diseases in the Northwest.
“The most prominent of these is rhythm. We hope that this study will be the basis for future work on autism that supports machine learning. ”
Researchers believe that their work can contribute to a better understanding of autism. Artificial intelligence can help reduce the burden on medical professionals, make autism diagnosis more accessible, and make it easier to diagnose autism, Lau said. It can also provide a tool that will one day transcend cultures, thanks to the computer’s ability to quantify words and sounds, regardless of language.
Since machine-specific speech features include both English and Cantonese, as well as those of a monolingual language, Loch said, machine learning can be useful not only for identifying aspects of speech that are appropriate for therapeutic interventions, but also for developing measurement tools. The result of those interventions by assessing the speaker’s progress over time.
Finally, the results of the study could be an attempt to identify and understand the role of certain genes and mechanisms of brain function associated with genetic predisposition to autism, the authors say. Finally, their goal is to create a broad picture of the factors that shape people with autism.
“One area of the brain involved is the subcortical level of hearing, which is really closely related to the differences in how speech sounds are processed in the brains of people with autism,” Lau said.
“The next step is to determine whether those processing differences in the brain lead to the behavioral patterns we observe here and their underlying neuronal genetics. We are excited about what lies ahead. ”
It’s about AI and ASD research news
Author: Max Witinski
A source: Northwestern University
The connection: Max Witinski is a Northwestern University
Photo: Image in public domain
Original study: Open access.
“Cross-Linguistic Patterns of Prosodic Differences in Speech in Autism: A Study of Machine Learning” by Joseph C. Lau et al. PLOS ONE
Cross-Linguistic Models of Speech Prosodic Differences in Autism: Machine Learning
Differences in speech prosody are a common feature of autism spectrum disorder (ASD). However, it is unclear how prosody differences in ASD appear in different languages, indicating the variability of cross-languages in prosody.
Using a controlled machine-learning analytical approach, we examined the acoustic features related to the rhythmic and intonational aspects of prosody from descriptive samples taken in two typologically and prosodically distinct English and Cantonese languages.
Our models demonstrated a successful classification of ASD diagnostics within two languages and using rhythm-comparative features in languages. Classification with intonation features was important for English, but not Cantonese.
The results highlight differences in rhythm as the main prosodic feature influencing ASD, as well as significant variability in other prosodic properties modulated by language-related differences such as intonation.