Laptop with accessible icons on the screen

Intel and the Open Federated Learning (OpenFL) Project Pledge to Improve Accessibility for Developers

author-image

作者

For over a decade, Global Accessibility Awareness Day (GAAD) has served to raise awareness of the need for accessibility in the digital world. As a founding sponsor of the GAAD Foundation, Intel is on a journey to create more inclusive user experiences, and we believe enabling developers of all abilities is a critical part of the journey to building an inclusive and accessible world.

This is especially true in areas of transformation and innovation, where the desire to move fast often means accessibility is not actively considered until much later (and requires more work to implement after the fact). This year, Intel and the Open Federated Learning (OpenFL) project are taking the GAAD Pledge to make accessibility a core project value and enable developers of all abilities to effectively contribute to OpenFL’s deep learning framework.

Supporting Accessibility and Enabling Innovation

The GAAD pledge was launched in 2021 to encourage major open-source initiatives to adopt accessibility as a core principle of their projects. With Intel’s rich history of collaborating with open-source communities – a total of 19,000 software engineers, twenty years of code contributions, hundreds of open-source projects, and six architectures supported in OneAPI – we are in a prime position to have a positive impact on the accessibility of open-source projects and the communities that build them. As the world embraces new technologies, like artificial intelligence, machine learning, and confidential computing, it is crucial that the companies enabling these technologies have an accessibility-first mindset.

Federated learning is a method of distributed training in which large volumes of data can be processed across a set of decentralized systems (rather than combining all the data together in one shared location). It’s increasingly popular in industries like healthcare and finance, where privacy and data security are paramount. OpenFL is a framework for federated learning that is designed to be flexible, extensible and secure. It allows organizations to participate in collaborative multiparty machine learning without moving their confidential or regulated data off-premises. Instead, the model trains on the data where it resides, and the learnings from local data are consolidated centrally. No single party’s data is exposed to the other participants. 

Intel and OpenFL’s Commitment

Originally developed by Intel, this community-owned project is open to contributions to code and documentation from developers. In March of this year, OpenFL was accepted in the Linux Foundation* AI & Data Foundation as an incubation project to further drive collaboration, standardization, and interoperability. With the GAAD Pledge, OpenFL and the LF AI & Data Foundation are committing to the following changes in 2023:
 

  • Actively capture and remediate accessibility issues in our GitHub* repo
  • Publish accessibility guidelines and expectations in the OpenFL GitHub repo, including Code of Conduct, Contributors.md, Readme.md
  • Conform to WCAG 2.1 AA standards with our OpenFL.io web presence
  • Update developer documentation and tutorials to comply with accessible document best practices
  • Adopt the new Intel One Monospace font in all relevant documentation of OpenFL framework for optimized legibility

Letters, numbers, and special characters displayed using the Intel One monospaced font.

Beyond OpenFL: A Font Designed for All Developers

Part of OpenFL’s pledge includes adopting the newly launched Intel One Monospace font. Designed with input from a team of low-vision and legally blind developers for optimized legibility, it’s easier to read and can be used by anyone to make their work more accessible.

Here are the key design characteristics that increase legibility:
 

  • Character differentiation: letters and coding glyphs designed to increase the differentiation between "like" characters. For example, the lowercase ‘l,’ uppercase ‘L’, and the number ‘1’, where all three shapes are prone to be confused with others and need extra clarification. These treatments reduce the need to skip back and confirm what’s been read.
  • Differentiated upper- and lowercase height: unlike most monospace fonts that have increased x-heights, our research showed that bigger differences between capital and lowercase letters, along with longer ascenders and descenders could help create more radical "word-shapes".
  • Standard brace, parenthesis, bracket, and slash designs are exaggerated to quickly identify sections of code.
  • Curved strokes, like in the lowercase d and b, hit stems at different points for less visual monotony.


Available for free, with an open-source font license, we invite the developer community to use the typeface and provide insights on daily use, what you would like to see changed or what you would like to see added in future releases of the typeface family.