Evaluating atypical language in autism using automated language measures

Area of Science:

• Linguistics
• Developmental Psychology
• Clinical Psychology

Background:

• Measuring language atypicalities in Autism Spectrum Disorder (ASD) is challenging and expensive.
• There is a need for more efficient and valid language outcome measures for ASD.
• Existing methods for assessing language in ASD are often time-consuming and resource-intensive.

Purpose of the Study:

• To develop and validate Automated Language Measures (ALMs) using language transcripts.
• To assess the effectiveness of ALMs in differentiating between individuals with ASD and neurotypical controls.
• To determine if ALMs can serve as reliable outcome measures for language in ASD.

Main Methods:

• Collected language transcripts from 169 participants (ages 7-17) including those with ASD, typically developing (TD), and ADHD.
• Generated seven ALMs from transcripts: mean length of utterance in morphemes, number of different word roots (NDWR), um proportion, content maze proportion, unintelligible proportion, c-units per minute, and repetition proportion.
• Utilized nonparametric ANOVAs and logistic regression analyses to compare groups and predict ASD status.

Main Results:

• Significant group differences were found for most ALMs, with the ASD group generally scoring lower than TD and ADHD groups (except for repetition proportion).
• The TD and ADHD groups did not significantly differ from each other on the ALMs.
• Four ALMs accurately predicted ASD status (67.9-75.5%), with combined ALMs achieving 82.4% accuracy.

Conclusions:

• Automated Language Measures (ALMs) derived from speech transcripts are a valid and promising approach for assessing language in ASD.
• ALMs can differentiate individuals with ASD from comparison groups and predict ASD status with notable accuracy.
• These novel ALMs offer a potential solution for more efficient and cost-effective language outcome measurement in ASD research and clinical practice.