Recent innovations in speech technology have made high quality TTS and ASR available even for extremely low-resource languages. This paper presents our updated work-in-progress report of an open-source speech technology project for two indigenous Sámi languages that are minority languages in Norway, Sweden and Finland. At this stage, we have designed and collected text and speech corpora for training the first neural text-to-speech (TTS) for Lule Sámi. We will update the previous North Sámi TTS by collecting additional materials and by training a new model using state-of-the-art technologies. We also describe our first experiments with developing ASR for North Sámi and discuss the next steps to be taken in our project.
| Text | Language | Gender | TTS |
|---|---|---|---|
| Dohko bohtet olbmot ja duddjojit buot lágan dujiid ja mii guossohit gáfe, ságastallat dujiid birra ja deaivvadit nuppiiguin. | North Sámi | Male |
|
| Raportta mielde ulbmilin lea maid geahpedit boazodoalu ja eará eanageavaheami ruossalasvuođaid sihke sihkkarastit guohtoneatnamiid ceavzilis anu. | North Sámi | Female |
|
| Divvun, sáme duollatjállemvædtsak, le dal ásaduvvam stuoves årnigin Tråmså universitiehtan. | Lule Sámi | Male |
|
@inproceedings{hiovain-asikainen-de-la-rosa-2023-tts-asr-sami,
title = "Unsupervised Word Segmentation from Discrete Speech Units in Low-Resource Settings",
author = "Hiovain-Asikainen, Katri and
{De la Rosa}, Javier",
booktitle = "Proceedings of the 2nd Annual Meeting of the Special Interest Group on Under-resourced Languages",
month = aug,
year = "2023",
address = "Dublin, Ireland",
publisher = "Interspeech 2023",
url = "...",
pages = "...",
}