Semi-Automatic Spontaneous Language Analysis (SASTA): Towards an efficient method for determining the language profile of children with DLD

Abstract

Grammatical analyses of spontaneous speech samples can play a key role in the action-oriented diagnosis of developmental language disorders. Speech therapists are generally familiar with analysis methods such as TARSP and STAP and recognize their importance. However, their use is limited because they are time-consuming and require specific knowledge and skills. In this article, we describe a computer application, SASTA, that semi-automatically performs grammatical analyses of children’s utterances. This can make language sample analysis faster and easier. We outline how SASTA works, focusing on the processing of deviant grammatical structures and incomplete or incorrect word realizations, which frequently occur in child language. Tests comparing the analyses generated by SASTA with those of trained human analysts show that SASTA has an accuracy of 80% or better. We expect that performance will continue to improve through ongoing research and development, and we conclude that SASTA has the potential to become a valuable tool in the diagnosis of language development disorders.