This week, we announced the Speech Accessibility Project, led by the University of Illinois Urbana-Champaign (UIUC), to further address the data desert in speech technologies. Working with tech leaders across the industry, UIUC will gather a set of high-quality, diverse speech samples that will help accelerate speech technologies. This one-of-a-kind cross-industry collaboration is rooted in the belief that inclusive speech recognition should be a universal experience.
This project highlights a real opportunity. As expectations evolve with how technology is used at home, in school and at work, Microsoft continues to prioritize partnerships to rapidly accelerate how people with disabilities are included in and represented by the systems, designs and features of technology.
While the interest in improving speech interactions is not new and not unique to any single organization, this is an important step that shows how we make progress is changing. We are recognizing as an industry that partnership is crucial to progress. Microsoft is working with partners on not only creating innovative accessible and disability-inclusive technologies, but to raise the bar for what is possible.
Disability representation data is crucial
One of the biggest challenges the research community and industry at large faces is a data desert for disability representative data. A data desert is the lack of data for a particular group, which leads to scientific discoveries and innovations developed without those groups being counted. One area where it is particularly relevant is tied to speech technologies.
Speech recognition is powered by machine learning (ML) models, and without diverse, representative data, ML models cannot learn diversity of speech, limiting customer scenarios where this technology can bring benefits. And although there are regional statistics currently available about people with disabilities, there is a lack of consistency in the types of questions asked, how disability is defined and the methods of reaching people with disabilities, especially in low- and middle-income communities.
To respond to this gap between societal need and tool capability, Microsoft has partnered with the World Bank, in collaboration with the Disability Data Initiative at Fordham University, to develop a public facing, online “disability data hub” to more accurately document representation of disability across populations, geographies and development indicators.
The World Bank works with local advisors and disability experts to inform the design and functionality of this hub, and Microsoft is grateful for the opportunity to contribute to this effort. This was needed even a decade ago, but given the growth in disability, the state of technology and data analysis, and interest from decision-makers, this could not be happening at a better time.
Incremental progress is useful
Disability-led AI investments in everyday Microsoft technologies, such as Seeing AI, continue to move the needle, not only by adding useful features but by driving innovation in ways that were not possible even two years ago. Recent updates to Windows 11 delivered new accessibility features, including the availability of Live Captions from the operating system – whether connected to the internet or not. Employees working on Windows started with an idea and co-designed with people who are deaf or hard-of-hearing over a few years to ultimately re-envision a system-wide live captioning experience. And, while live captions are useful in many scenarios, increasing options for communication access at the operating system level regardless of connectivity is real progress.
We know that our partnerships and collaboration with those most impacted by the technologies are fundamental to understanding, developing and validating the features we build. Through conversations with our customers and disability advisory boards, we continue to increase our understanding of the challenges and opportunities we have to improve and create more inclusive experiences.
The moment is now
Accessibility in 2022 is different. The need for accessibility has increased and capabilities of technology have matured. At Microsoft, we believe access to technology is a fundamental right globally. No single feature will get us where we intend to go but leaning into the innovative nature of accessibility and disability inclusion will aid progress in multiple areas, alongside the partners and community advocates with whom we are on this journey. There will be more moments, for sure, but what we do now will help expand the group of people that deeply understand the innovation potential of accessible technology when we find ourselves with the next opportunity to accelerate.
With the historic commitment to disability inclusion and the innovation experience of the Beckman Institute at UIUC and the Speech Accessibility Project, we are humbled by the opportunity to be involved, and are optimistic about how this will improve the future of inclusive speech experiences at scale. And we are grateful for the expertise, guidance and support from both Team Gleason and the Davis Phinney Foundation for the project’s initial focus on potential benefits to the ALS and Parkinson’s communities, with more community partners being added throughout.