Advances in the Identification and Agreement About Meta-Therapy for Voice Disorders: A Methodological Paper

Area of Science:

• Speech-language pathology
• Voice therapy research
• Computational linguistics

Background:

• Meta-therapy (MT) is a crucial dialogue-based component of voice therapy.
• Inconsistent identification of MT among clinicians hinders research and education.
• Natural language processing can help clarify MT's definition and clinical use.

Purpose of the Study:

• To develop and validate a reliable method for identifying Meta-therapy (MT) in voice therapy sessions.
• To analyze the linguistic characteristics of MT.
• To improve the theoretical and empirical understanding of MT.

Main Methods:

• Analysis of 10,443 clinician utterances from voice therapy sessions.
• Development of a refined annotation framework for MT.
• Assessment of inter-rater reliability using Gwet's AC1, Cohen's kappa, and percentage agreement.

Main Results:

• Gwet's AC1 demonstrated substantial to almost-perfect agreement for MT identification (up to .98).
• Distinct linguistic patterns (bigrams) differentiated MT from other therapy components.
• Lexical overlap caused some confusion between MT and direct/educational content.

Conclusions:

• A revised annotation model significantly improved MT identification reproducibility.
• The study identified MT's linguistic signature and its integrated role in therapy.
• Findings support the development of machine learning tools and formal MT curricula.