Using phonetic transcription to get artificial intelligence avatars to pronounce speech in Quechua: the Illariy case; [Uso de transcrição fonética para fazer avatares de inteligência artificial pronunciarem fala na língua Quechua: caso Illariy]

Huamán-águila Ó.; Fernandez-Garcia C.; Gonzales-García C. · Lengua y Sociedad (2024) · DOI: 10.15381/lengsoc.v23i2.28422

Resumen

Artificial Intelligence (AI) has has been in the news this year. In its fifth season, the news program Letras TV Willakun of the Faculty of Letters and Human Sciences of the Universidad Nacional Mayor de San Marcos is hosted by Illariy (a Quechua word meaning “Dawn”), an avatar generated by Artificial Intelligence, the only one of its kind that speaks a native language. This fact leads us to pose the question: How does phonetic transcription influence the ability of Artificial Intelligence to generate coherent speech in the Quechua language? The aim of this research is to describe the workflow focusing on how phonetic transcription can enable Artificial Intelligence to articulate a speech in Quechua. The research methodology has a qualitative approach, based on the observation and analysis of the workflow to generate avatars to deliver a speech in Quechua. Among the results we can observe the existence of a differentiated pronunciation between Spanish and Quechua, limitations in the articulation of some consonants of the Quechua language that are irreproducible in techno-phonological contexts, as well as the use of pauses and commas for the production of a Quechua orality in the D-ID Software. © Los autores.

Instituciones: Universidad Nacional Mayor de San Marcos