Evaluation of Speech Foundation Models for ASR on Child-Adult Conversations in Autism Diagnostic Sessions

TL;DR

This paper evaluates the performance of speech foundation models on child-adult conversations in autism diagnostic sessions, revealing significant challenges and improvements through fine-tuning in low-resource settings.

Contribution

It provides a comprehensive evaluation of ASR models on child-adult autism conversation data and demonstrates effective fine-tuning methods to improve performance.

Findings

Speech models show 15-20% higher WER on child speech compared to adult speech.

Fine-tuning Whisper-large with LoRA improves WER by 8% for child speech.

Fine-tuning also reduces WER by 13% for adult speech.

Abstract

Reliable transcription of child-adult conversations in clinical settings is crucial for diagnosing developmental disorders like Autism. Recent advances in deep learning and availability of large scale transcribed data has led to development of speech foundation models that have shown dramatic improvements in ASR performance. However, their performance on conversational child-adult interactions remains underexplored. In this work, we provide a comprehensive evaluation of ASR performance on a dataset containing child-adult interactions from autism diagnostic sessions, using Whisper, Wav2Vec2, HuBERT, and WavLM. We find that speech foundation models show a noticeable performance drop (15-20% absolute WER) for child speech compared to adult speech in the conversational setting. Then, we fine-tune the best-performing zero-shot model (Whisper-large) using LoRA in a low-resource setting, yielding 8% and 13% absolute WER improvements for child and adult speech, respectively.