Research

Our research questions

Digital media is central to how we engage with public information, news, government services, healthcare and welfare, culture, entertainment and education. As the volume of digital media content grows, access needs can no longer be met by human resources alone. This has led to increasing interest in AI-driven approaches to create accessible digital content.

Yet, engaging with digital media often involves complex language-based and/or multimodal interactions that draw on multiple senses. Different forms of translation are used to make media accessible across modalities (e.g., translating speech into subtitles, standard text into simpler versions or visual content into a verbal description) and languages (e.g., translation into minority language). However, language, sound, and vision present complex challenges for AI, requiring sustained, fundamental research.

The ADA network harnesses AI alongside human expertise to develop human-centric AI-enabled solutions for accessible communication that preserve accuracy, narrative coherence and other quality parameters in digital content. Research in the ADA network is guided by a set of overarching questions:

• What are the current affordances, limitations and risks of AI when addressing accessibility challenges related to digital media, and how can the limitations and risks be mitigated?
• How can AI be used to personalise digital media services to diverse individual needs, and what are the challenges associated with doing so while avoiding biases or discrimination?
• How can AI be used safely to create language-based and multimodal digital content that preserves accuracy and narrative coherence for people with diverse needs and abilities?
• How can the use of AI for digital accessibility become transparent, ethical and accountable to prevent biases and discrimination?

Our research themes

We focus on six key research areas:

1. AI for Audiovisual Accessibility: Exploring AI-enabled solutions for converting audiovisual content into accessible formats, such as audio description for people with sight loss and other audiences.
2. AI for Speech-to-Text Accessibility: Advancing AI-based speech technologies to improve real-time and post-production accessibility of speech content for different use cases.
3. AI for Text Simplification and Comprehension: Investigating how AI can enhance text accessibility by simplifying complex language, improving readability, and adapting content for diverse users, including neurodiverse users.
4. AI for Cross-Language Accessibility: Developing AI-powered tools for multilingual access through translation, interlingual subtitles and interpreting to support global content reach and inclusivity.
5. AI-Enhanced Assistive Communication Technologies: Creating AI-enabled assistive tools, such as voice assistants and chatbots to improve experiences for diverse users across digital platforms.
6. AI for Personalising Accessible Communication: Developing novel methods for tailoring digital accessibility solutions to individual preferences and needs while ensuring ethical AI practices.