Maturation Assessment System for Speech and Language Therapy based on Multilevel PAM and KNN

Abstract This paper presents an approach to automatically generate personalized plans for Speech and Language Therapy on children, based on five speech and language areas: linguistic formulation, oral structure and function, receptive language, expressive language and hearing. The approach relies on an implementation of the Partition Around Medoids (PAM) algorithm, which takes into account the profile of each child, and generates multi-level (hierarchical) clusters based on several distance measures. The new plans are created as per a K-Nearest Neighbors (KNN) criterion, combining elements from the five areas. This approach has been used in the assessment of 53 children with different disabilities, with promising results verified by expert therapists.