Multidisciplinary Perspectives on Automatic Analysis of Children's Language Samples: Where Do We Go from Here?

Area of Science:

• Computational Linguistics
• Information Science
• Speech-Language Pathology

Background:

• Language sample analysis (LSA) is crucial for understanding child language development in clinical and research settings.
• While digital tools exist, many LSA steps remain manual, hindering efficiency.
• Automatic speech recognition (ASR), a key component of LSA, is nearing mainstream use.

Purpose of the Study:

• To explore the technological complexities and future requirements for fully automated LSA of child language.
• To review existing LSA software, compare automation levels, and synthesize research from relevant disciplines.

Main Methods:

• Multidisciplinary review integrating information science and computational linguistics perspectives.
• Characterization of requirements for a fully automated LSA pipeline.
• Comparative analysis of current LSA software features and capabilities.

Main Results:

• Existing LSA tools offer varying degrees of automation.
• Machine learning advances show promise for LSA, but child-specific data limitations pose design hurdles.
• A transdisciplinary approach is identified as essential for future LSA software development.

Conclusions:

• Developing fully automated LSA for child language requires addressing unique data and processing challenges.
• Collaboration across disciplines is vital for advancing LSA technology.
• Future LSA software development should leverage machine learning while considering the specificities of child speech and language.