Inclusive Spoken Language Science and Technology – Breaking Down Barriers
The Interspeech Community has played a vital role over the decades in advancing the state of the art around the science and technology of spoken language processing. Our understanding of human speech and language enables the development of technology and policies with the ability to transform lives for the better. This creativity comes with a responsibility. We believe that our community has a special role to play in ensuring that emerging technology in the domain of speech and language should be for the benefit of the widest possible audience. We plan to use Interspeech 2023 to promote the theme of inclusion in the context of speech and language research, asking participants to focus on whether their work is relevant for the widest possible range of users.
As researchers, we are all too familiar with the challenge of variability in human speech and language. We know that male and female voices are quite distinct. Our voices and speaking style change dramatically as we age from childhood to our elderly years. There are over 6000 different languages in the world. New speakers of a language bring more new accents and idiosyncrasies. Speech pathologies result in barriers to speech technology. In Interspeech 2023, we want researchers to pause and think about inclusion. Is your algorithm, your system, your experiment, or discovery, useful for as wide a range of speech or speakers as possible? If your dataset has all native speakers of a language, can you extend it to non-native speakers? If you demonstrate performance of your system on an English language dataset, can you also show it works for Spanish, Chinese, or Hindi? Does your finding about adult speech include older speakers, or what are the limits and why? If you quote overall percentage performance figures, can you break that down for gender, by age, or other demographics?
Ireland is a country with two official languages – English and Irish. It is an ever-changing country, with 40% of our population under 30 years of age and 17% of population now not born in Ireland. It is a progressive society, being the first country in the world to legalise same-sex marriage by popular vote in 2015, and the first country to commit to sell off its investments in fossil fuel companies in 2019. The Irish are famous for speech and language- just think of the writer Oscar Wilde, the poet WB Yeats, former President and former UN High Commissioner for Human Rights Mary Robinson, and the outspoken musician Bono. What better setting for this discussion!
The theme of inclusion will resonate at multiple levels throughout the conference. The roots of Interspeech lie in our understanding of human speech communication. However, with the power of deep-learning and the improved performance it has delivered for speech technology, it is possible to get carried away on the wave of algorithmic advancement. We will also celebrate the human side of speech communication, giving full recognition to the importance of including core language knowledge in technology development.