Figure 8. The end-to-end operations in Logo-DM-Modern Tools in Patient-Centred Speech Therapy for Romanian Language

The useful data mining tasks for speech therapy fall into three categories: classification, clustering, and association rules. Classification places children with different speech impairments in predefined classes, and makes possible to track the characteristics of various groups. To model different classes we use many predictor variables (e.g. personal or familial anamnesis data or related to lifestyle). By clustering we group people with speech disorders on the basis of similarity of different features. This helps therapists to understand their patients. Clustering aims to find subsets of a predetermined segment, with homogeneous behavior towards various methods of therapy that can be effectively targeted by a specific therapy, but it is not based on the previous definition of groups (Danubianu, Tobolcea, & Pentiuc, 2009). Association rules aim to find out relationships between different data which seem to have no semantic dependence. The built patterns might be very useful to determine why a specific therapy program has been successful on a segment of patients with speech disorders, and on the other was ineffective. The Logo-DM system was designed to help the speech therapists to optimize the personalized therapy of dyslalia. To understand what kind of knowledge we could discover in TERAPERS’ dataset to improve speech therapy, we have to describe the collected data.