SASTA Self Assessment: An efficient human-in-the-loop strategy for developmental and pathological language analysis

Abstract

This paper introduces SASTA self-assessment, i.e., a self-assessment procedure for the SASTA application for semiautomatic analysis of spontaneous language transcripts for Dutch. We introduce SASTA and the methods that it supports. These methods are used to assess the language development of young children and to assess the language skills of patients with aphasia. We illustrate this with typical example utterances. The performance of SASTA is good but not good enough for fully automatic use. The self-assessment procedure attempts to automatically identify utterances that require revision by a human expert. The self-assessment procedure gives promising results for datasets for the ASTA and STAP methods. Significant improvements are still needed for the TARSP method, but there is still potential for such improvements, and we sketch some directions to achieve such improvements.

Resources

Details