Abstract

The study of oral narrative discourse allows for the description of several features of the speakers (e.g., belonging to social or age groups, pathologies or special educational needs, stages of acquisition or learning, among many other attributes). This type of semi-structured tasks also allows speakers to generate ecological discourses with comparable lexical and semantic structures, as well as being easily replicable independently of the examiner. In this framework, we present a retelling instrument composed of three stories made from 15 static images each based on a central event. In order to demonstrate its usefulness, we presented an automatic analysis of discourses generated by 50 university students and 13 people diagnosed with schizophrenia. The results showed homogeneity of the texts, based on the comparison of the types of words prompted. We observed that one of the stories generates more abstract content, which makes it especially useful for the study of certain populations.