Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Automatic Screening for Children with Speech Disorder using Automatic Speech Recognition: Opportunities and Challenges

Dancheng Liu, Jason Yang, Ishan Albrecht-Buehler, Helen Qin, Sophie Li, Yuting Hu, Amir Nassereldine, Jinjun Xiong

TL;DR

A survey of existing techniques suitable for automating SLA pipelines, with an emphasis on adapting automatic speech recognition (ASR) models for children’s speech, and an overview of current SLAs and their automated counterparts to demonstrate the feasibility of AI-enhanced SLA pipelines.

Abstract

Speech is a fundamental aspect of human life, crucial not only for communication but also for cognitive, social, and academic development. Children with speech disorders (SD) face significant challenges that, if unaddressed, can result in lasting negative impacts. Traditionally, speech and language assessments (SLA) have been conducted by skilled speech-language pathologists (SLPs), but there is a growing need for efficient and scalable SLA methods powered by artificial intelligence. This position paper presents a survey of existing techniques suitable for automating SLA pipelines, with an emphasis on adapting automatic speech recognition (ASR) models for children's speech, an overview of current SLAs and their automated counterparts to demonstrate the feasibility of AI-enhanced SLA pipelines, and a discussion of practical considerations, including accessibility and privacy concerns, associated with the deployment of AI-powered SLAs.

Automatic Screening for Children with Speech Disorder using Automatic Speech Recognition: Opportunities and Challenges

TL;DR

A survey of existing techniques suitable for automating SLA pipelines, with an emphasis on adapting automatic speech recognition (ASR) models for children’s speech, and an overview of current SLAs and their automated counterparts to demonstrate the feasibility of AI-enhanced SLA pipelines.

Abstract

Speech is a fundamental aspect of human life, crucial not only for communication but also for cognitive, social, and academic development. Children with speech disorders (SD) face significant challenges that, if unaddressed, can result in lasting negative impacts. Traditionally, speech and language assessments (SLA) have been conducted by skilled speech-language pathologists (SLPs), but there is a growing need for efficient and scalable SLA methods powered by artificial intelligence. This position paper presents a survey of existing techniques suitable for automating SLA pipelines, with an emphasis on adapting automatic speech recognition (ASR) models for children's speech, an overview of current SLAs and their automated counterparts to demonstrate the feasibility of AI-enhanced SLA pipelines, and a discussion of practical considerations, including accessibility and privacy concerns, associated with the deployment of AI-powered SLAs.

Paper Structure

This paper contains 15 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Automatic Speech Recognition
  3. Adult ASR
  4. Adapting to Children Speech
  5. Training Directly from Children Speech
  6. Automatic Screening
  7. Traditional Approach of SLA
  8. End-to-End System
  9. Explanable System Derived from SLAs
  10. Report Generation
  11. Practical Concerns on Availability and Privacy of Automated SLA
  12. Availability of DL Models due to Privacy Concerns
  13. Fairness towards Marginalized Group
  14. Conclusion
  15. Acknowledgements

Figures (3)

  • Figure 1: A generic pipeline for automated speech and language assessments in the future. Children in need of SLA will use a local computer to perform all required components of the test which ensures privacy and usability, and such remote SLA will also alleviate the burden of SLPs, helping them focus on the treatment of children in need.
  • Figure 2: The confusion matrix on AutoRSR vs Human. Pass/Fail indicates whether the child passes/fails the RSR test. Cases when AutoRSR passes and Human fails the child are false negatives.
  • Figure 3: The distribution of differences between the automated RSR and human scorers. We can see that there is still a subtle difference between the scores even though the final classification is mostly correct.