Automated extraction of speech and turn-taking parameters in autism allows for diagnostic classification using a multivariable prediction model

Authors:
  • JKJana C. Koehler
    Jana C. Koehler
  • ABAfton M. Bierlich
    Afton M. Bierlich
  • NKNikolaos Koutsouleris
    Nikolaos Koutsouleris
  • CFChristine Falter-Wagner
    Christine Falter-Wagner
Preprint
Publisher website
Google Scholar
Abstract
Speech, if present, is usually affected in terms of how people speak as a central symptom of autism spectrum disorder (ASD). Additionally, autistic people report having difficulties with small talk. Since small talk specifically, and conversations in general, are essential for building connections with others, it is important to understand how autistic and non-autistic verbal behaviour differs. In this study, we recorded dyadic conversations and used automated extraction of speech and interactional turn-taking features of 54 non-autistic and 26 autistic participants. Using Bayesian mixed ANOVAs, we found differences in speech, such that autistic interaction partners talked slower and in a more monotonous manner than non-autistic interaction partners. Turn-taking parameters also differed between homogeneous non-autistic dyads and heterogeneous (autistic and non-autistic) dyads. Specifically, heterogeneous dyads had increased periods of silence, and the intensity, i.e., loudness, of their speech was more synchronous. The extracted speech and turn-taking parameters showed high potential as a diagnostic marker. A linear support vector machine was able to predict the dyad type with 76.2% balanced accuracy (sensitivity: 73.8%, specificity: 78.6%), suggesting that digitally assisted diagnostics could significantly enhance the current clinical diagnostic process due to their objectivity and scalability.