Pediatrik Konuşma Değerlendirmesi İçin LLM Tabanlı Sistemler Teknik Olarak Mümkün Görünüyor Ancak Gerçek Dünya Doğrulaması Eksik

Özet (abstract)

Pediatric speech and language assessment supports the early identification of communication difficulties and access to timely intervention. Large language models (LLMs) and related generative artificial intelligence technologies are increasingly being incorporated into systems that analyze children’s speech and language data. However, evidence regarding their application in pediatric assessment remains limited and fragmented. This scoping review aimed to map and critically examine research on LLM-enabled systems used to assess children’s speech, language, and communication abilities. Four databases were searched, and seven studies met the inclusion criteria. Data were charted according to study design, purpose, participant and target-population characteristics, assessment domain, type of LLM, model input, role of the LLM, additional artificial-intelligence components, evaluation methods, and reported outcomes. The findings provide preliminary evidence of technical feasibility and agreement with human or reference-based assessments. However, the evidence remains limited because of small or restricted datasets, the use of synthetic data in some cases, limited real-world validation, and concerns regarding fairness and generalizability across languages and clinical populations. Further research is necessary to establish clinical validity, reliability, and appropriate use of LLM-enabled systems before their broader implementation in pediatric speech and language assessment.

Yazarların özeti; kaynağından alınmıştır. Applied Sciences, 2026 · DOI ↗