14thAnnual Conference of the International Speech Communication Association

Lyon, France
August 25-29, 2013

Modeling Postcolonial Language Varieties: Challenges and Lessons Learned from Mozambican Portuguese

Simone Ashby (1), Sílvia Barbosa (1), Catarina Silva (1), Paulino Fumo (2), José Pedro Ferreira (1)

(1) ILTEC, Portugal
(2) Eduardo Mondlane University, Mozambique

While a growing number of studies deal with the modeling of accents by non-native speakers - with a strong bias towards English - relatively little is understood about postcolonial language varieties and the effects of substratal transfer (i.e. interference from local tongues) on the resulting sound systems. As the next information technology frontier, Africa presents unique challenges for language technologists tasked with adapting existing speech and language technologies (SLTs) to accommodate for this often systematic form of interference. As part of an effort to model and generate rule-based pronunciation lexicons for resource-scarce varieties, we present our work on the variety of Portuguese spoken in the Mozambican capital of Maputo. We highlight some of the segmental processes that result from contact with local Bantu languages, i.e. denasalization, aspiration, and rhotic variation. We then discuss these in the context of adapting Mozambican Portuguese (MP) to existing SLTs. Problems of interand intra-speaker variability and the lack of a well defined standard are also considered, along with implications for adopting a suitable model of MP for speech synthesis.

Full Paper

Bibliographic reference.  Ashby, Simone / Barbosa, Sílvia / Silva, Catarina / Fumo, Paulino / Ferreira, José Pedro (2013): "Modeling postcolonial language varieties: challenges and lessons learned from mozambican Portuguese", In INTERSPEECH-2013, 310-314.