PsychosisBank Portuguese C-ORAL-ESQ Corpus


Tommaso Raso
Linguistics
Federal University of Minas Gerais

Participants: 43
Type of Study: discourse -- conversations
Location: Brazil
Media type: audio, video
DOI: doi:10.21415/X9NM-ZD33

Browsable transcripts

Download transcripts

Media folder

Citation information

In accordance with TalkBank rules, any use of data from this corpus must be accompanied by at least one of these references:

Project Description

The project is part of a larger project to compile corpora (audio and audio and video) and study spontaneous speech in Brazilian Portuguese (C-ORAL-BRASIL project, coordinated by T. Raso and H. Mello). The objective of this project is the compilation and study of multimodal corpora of patients with schizophrenia. In particular, the project envisages the completion of the C-ORAL-ESQ corpus, consisting of audio recordings of interactions between patients and doctors at the Instituto Raul Soares and the Hospital das Clínicas, and the partial compilation of a multimodal corpus of the same type of interaction. Based on these data, the project proposes to analyze some aspects of patients’ speech in comparison with non-pathological speech. Some information to be analyzed can be extracted automatically or semi-automatically, while others require the annotation work of a trained team such as that of the Laboratory for Empirical and Experimental Studies of Language (LEEL) at the Faculty of Arts at UFMG. Measures can be automatically extracted that compare schizophrenic and non-pathological speech regarding (i) duration of turns (in time, utterances and words); (ii) complexity in the prosodic-informational structure of utterances; (iii) number of facial action units per utterance; (iv) type of action units used and their intensity. The main measurements that require human annotation to be extracted are: (i) quantification of the type of informational units present; (ii) the evaluation of functional prosodic prominences from an informational point of view; (iii) the quantity and quality of manual gestures per utterance; (iv) the relationship between action units and informational prominences.

A fuller description is available in this article.