Em 200 metros, vire à esquerda : a entoação dos comandos de GPS

Abstract

GPS (Global Positioning System) is a technology that uses a voice produced by speech synthesis to provide orientation commands and instruct listeners. As it is an instruction, GPS speech falls within Speech Act Theory, more specifically in directive acts, which aim to make the listener perform the action uttered by the speaker. Studies point to the important role of prosody in directive speech, since intonation has the function of differentiating directive speech acts in communication (SANTOS, 2010; GOMES DA SILVA, CARNAVAL & MORAES, 2020). Although synthesis systems do not present problems at the segmental level, studies have shown that they still fail at the prosodic level (TAYLOR, 2009). The present study proposed to analyze and describe the intonational patterns of directive speeches produced by GPS systems and compare them with the human voice. To achieve this goal, the analysis corpus used 20 phrases from Google Maps, 20 phrases from Waze and the realization of these phrases by two female participants, native speakers of Brazilian Portuguese in the Alagoan variety. Acoustic analysis and description of curves were done with AnalyseTier scripts (HIRST, 2012) and ProsodyPro (XU, 2013), in conjunction with Praat, and we applied statistical tests using R software. The results showed that there were differences in intonation between applications and humans when evaluating f0 variation, as well as significant acoustic differences in mean f0 and maximum f0 averages in all intonational units analyzed. The results also showed similarities between applications and humans in f0 configuration, although there are differences when checking the graphical representation of curves. Such results reinforce the advancement of speech synthesis given the similarities in f0 configuration; however, they also indicate that it is still necessary to improve these systems since there were significant differences between this speech and natural voice.