Has your experience as an adult learner guided your work on second language acquisition methods?

Learning any language as an adult can be an artificial process. To learn a language as a child is a gold-standard not easily reproduced because of the linguistic and cultural interference we have from our first language(s) when studying a new one. However, while children learn a language because they are biologically and developmentally pushed to communicate, adults have the ability to choose and make a commitment to learning a language. This can be an advantage.

While the process is more complicated or more artificial, the motivation factor can be a major advantage for adults who can also access different resources to support their learning. While some methods seem more effective than others for building proficiency, no method can replace a consistent and intentional approach to learning.

Could you elaborate on how community-based activities and your collaborations with Mohawk-first language speakers have guided your research on reclaiming and revitalizing language?

My first two language teachers at Tyendinaga have both gone on to the spirit world. Tekahonwensere’kénha (the late Mel Diabo) spent many years of his ‘retirement’ travelling back and forth from his home community of Kahnawà:ke. He was an extraordinary storyteller and constantly encouraged his students through humour and his good-words to keep our language alive. Karihwawishon’kénha (the late Joe Brown) was a second-language learner who spent the short time he was given reclaiming the language for himself and ensuring it was passed on to countless students in our community. They both worked to make sure that our language would continue being spoken so that we, Kanyen’kehá:ka (Mohawk people) would continue existing. These two teachers, along with the other first-language speakers I’m privileged to count as teachers and friends, and people like my gramma who were never given the gift of Kanyen’kéha, motivate me to continue whatever small work I can do to ensure our words continue to be spoken.

In a recent paper, you described the development of the first neural speech synthesis systems in Canada for the purposes of Indigenous language revitalization. Could you tell us more about text-to-speech synthesis technology (TTS) and what it means for Indigenous languages in Canada to be considered 'low-resource'?

Text-to-speech (TTS) has the potential to be a transformative technology for our languages as they continue their transitions into the digital world. For learners, we have very limited resources for hearing the language spoken. Major world languages like English have billions and billions of hours of recorded speech available whereas Kanyen’kéha would have a few thousand. While TTS models cannot create new material, they can work as a force-multiplier with existing written materials and help make text available through audio.

There is also great potential for TTS to be used with online dictionaries. Because of the way our language works, a ‘complete’ dictionary would have an almost incalculable number of entries with every possible inflection. While various technologies can be used to start generating all these possible forms, it would take many lifetimes for a person to record the audio. TTS can be used to generate the spoken forms, which is a valuable addition for learners who are not always confident in their pronunciation.

The primary challenge with TTS models for Indigenous languages has been the limited availability of transcribed audio. Traditional TTS technologies have required thousands of hours of paired text and speech and this has been a challenge for us since many Indigenous languages in Canada (and around the world) can be considered ‘low-resource’ languages. And in the Indigenous language context, low resource exists on a completely different level, considering that speaker populations can sometimes be counted on one hand.

New approaches to language modelling have brought TTS within reach for Indigenous languages, reducing the minimum requirement to a few hours of transcribed audio in ideal circumstances. While these technologies have great potential, that potential is only realized ethically when the process and product are directed and controlled by the language communities. Meta has recently released a multi-lingual TTS model that can produce speech in thousands of languages. While this is exciting from a technological point of view, from my perspective, their work presents ethical concerns for its complete lack of involvement from the speakers and speech communities whose voices were used in their training. This work from Meta is the most recent example of large technology companies failing to understand our relationship with our languages.

What do you hope the impact of your research and teaching will be?

I hope that in whatever I do, it makes it easier for someone else in the future to do something better. With Indigenous research and Indigenous researchers, I think we’re in a phase where we’re preparing for something great – we’re opening the doors to Indigenous ways of knowing and doing research, we’re building foundations that will support research and researchers for generations to come, we’re Indigenizing the academy just by being Indigenous in the academy.

When I think specifically about language, my only hope is that my contributions will support students in their language learning journey. My work is just a small part of a puzzle that is larger than this country.