Potential uses of Brazilian Portuguese data repositories in Forensic Phonetics

Abstract

This study aims to survey existing data repositories of Brazilian Portuguese language and speech and to evaluate their potential applications in forensic phonetics, with a particular focus on speaker comparison tasks. A total of 45 speech corpora from various Brazilian regions were identified through consultations with specialists in sociolinguistics and dialectology, as well as through bibliographic research. To assess the potential use of these corpora for forensic purposes, a questionnaire was developed, addressing a range of aspects including general corpus characteristics, types of materials available, participant profiles, geographic coverage, and conditions for access and use of the collected data. Although the identified repositories fulfill their primary goal of documenting linguistic variation—especially at the lexical, syntactic, and conversational levels—they are not always suitable for the specific demands of speaker comparison within forensic phonetics. This task requires repositories that include high-quality audio recordings, metadata on speakers and recording conditions, and broad regional and demographic coverage to support the estimation of relevant linguistic and phonetic feature distributions. Among the 45 corpora surveyed, only seven were found to be adequately suited for use in speaker comparison analyses, particularly when employing likelihood ratio-based approaches, which require representative reference data. The results underscore a significant gap in the availability of appropriate and accessible resources for forensic phonetic applications in Brazil. In particular, there is a notable lack of corpora providing audio materials from the Central-West and North regions, and only one corpus includes such materials for the South and Northeast regions. This regional imbalance limits the ability to conduct robust speaker comparison analyses nationwide. Given these findings, it is essential that current and future corpus development initiatives consider including high-quality audio data, detailed metadata, and broad geographic representation. The outcomes of this research are relevant not only for linguists and forensic phonetics specialists but also for researchers working in related areas of language study, who may benefit from greater access to phonetically and regionally representative data.