Classification of Atypical Language in Autism

Atypical or idiosyncratic language is a characteristic of autism spectrum disorder (ASD). In this paper, we discuss previous work identifying language errors associated with atypical language in ASD and describe a procedure for reproducing those results. We describe our data set, which consists of transcribed data from a widely used clinical diagnostic instrument (the ADOS) for children with autism, children with developmental language disorder, and typically developing children. We then present methods for automatically extracting lexical and syntactic features from transcripts of children’s speech to 1) identify certain syntactic and semantic errors that have previously been found to distinguish ASD language from that of children with typical development; and 2) perform diagnostic classification. Our classifiers achieve results well above chance, demonstrating the potential for using NLP techniques to enhance neurodevelopmental diagnosis and atypical language analysis. We expect further improvement with additional data, features, and classification techniques.