Parts of speech and filled pauses in schizophrenia

Abstract

The aim of this work is to compare filled pause length and its distribution across parts of speech between two groups: patients with schizophrenia and people without this condition. The two hypotheses of this work are: I) there is a significant difference regarding filled pauses length between patients and no patients II) there is a relevant difference of length among filled pauses occurring right before a part of speech between the two groups. Methodologically, we compare 112 filled pauses taken from 13 audios of C-ORAL-ESQ (Rocha, 2019), a representative spoken corpus of patients with schizophrenia and 11 audios from C-ORAL-BRASIL I (Raso; Mello, 2012), which is a representative spoken corpus of informal Brazilian Portuguese. After data cleaning and treatment in the Python environment, we tagged utterances from both corpora with a tagger that we have built and trained in Mac-Morpho (Aluísio et al., 2003). The results show that there is a significant difference (p= 0,003) between filled pauses length between patients and no patients. Regarding parts of speech, we found that people with schizophrenia perform filled pauses larger than no patients, except for adverbs, with significant differences in filled pauses found right before personal pronouns.