1/2 Along with the Onkwawenna Kentyohkwa Kanyen'kéha immersion school, W̱SÁNEĆ School Board, University nuhelot'įne thaiyots'į nistameyimâkanak Blue Quills and @EdinburghUni, #NRCDigitalTech is developing speech synthesis systems and techniques for #IndigenousLanguages in Canada.
@istreasatuatha Thanks very much! Definitely interested to hear what you've been thinking about -- it's always nice to have a concrete application, and some in my research group have recently joined another project working on TTS in language learning settings too.
Seeking Scottish Gaelic speakers to help evaluate text-to-speech synthesis in the language!
I'm trying to make it easier to build TTS systems so that people can access speech technology in their own languages. If you speak Gaelic, you can help by taking part in a listening test.
And for a taster, you can listen to some initial samples from one of our systems here: https://t.co/dQ8pi8QjHB
The original speaker is Ruairidh MacIlleathain, who graciously gave us permission to use his vocal likeness as we get started with Gaelic TTS.
This involves listening to and rating short samples of artificial Gaelic speech. We'd like to recruit 30 participants, both native speakers and language learners, and you'll earn £5 for 20 minutes of listening.
If you're interested, please sign up here: https://t.co/dQ9dHHAaZe
#acl2022nlp Best Special Theme Paper
Requirements and Motivations of Low-Resource Speech Synthesis for Language Revitalization
(@AidanPine , Dan Wells, Nathan Brinklow, Patrick William Littell and Korin Richmond)
https://t.co/QnM2zsfbIk
It's been a really valuable experience for me to start thinking about this kind of application for TTS, and I'm very glad I got to work with Aidan on this! Hopefully more to come in the future
Very pleased with this for our paper "Requirements and Motivations of Low-Resource Speech Synthesis for Language Revitalization", based on @AidanPine's MSc project work which I had a lot of fun helping to supervise! You can read it here: https://t.co/IpYkZnTMb1
Eventually, we hope that TTS can be usefully integrated in the classroom to help language revitalisation efforts, led by Indigenous language communities in Canada as for Kanien'kéha, Gitksan and SENĆOŦEN.
Plenty still to do! We'd like to expand this to a wider set of more phonetically diverse languages, and explore methods for encouraging more information sharing in the input space. Thanks @Edin_CDT_NLP for supporting this work!
Presenting my first paper at SSW11 this week! We looked at using phonological feature vector inputs for low-resource speech synthesis: https://t.co/9xVXqdaade
See the paper for some analysis of the representations learned through each input type. We think phonological feature vectors allow us to learn more consistent representations across languages, and also share acoustic information between related sounds.