Peer-reviewed research
Improving ASR Systems for Children with Autism and Language Impairment Using Domain-Focused DNN Transfer Techniques
Robert Gale et al., Interspeech 2019, ISCA Archive, 2019
Read the originalRead on October 16, 2026
- What it is
- Describes building a speech recognizer for children aged 6 to 9 with autism, language impairment or both. The system adapts a model trained on adult speech using 1.5 hours of recordings from the children, and reaches a 29.38 percent word error rate in its best configuration.
- How much weight it carries
- Peer-reviewed conference paper. It studies one age group and one structured speaking task (repeating sentences), and it reports word error rate on that task, not on everyday conversation. Results for autistic adults are not measured here.
Facts taken from this source
- The 2019 system was adapted from a model trained on adult speech using only 1.5 hours of recordings of the target children.
“Working with only 1.5 hours of target data in which children perform the Clinical Evaluation of Language Fundamentals Recalling Sentences task” (Abstract)
- A 2019 speech recognition system for children aged 6 to 9 with autism or language impairment reached a word error rate of 29.38 percent in its best configuration.
“Our best configuration yields a 29.38% word error rate (WER).” (Abstract)