Advances in artificial intelligence technologies have a certain “cool” factor. But their import goes far beyond automation, transforming the future of manufacturing and offering engaging new ways to present eLearning and performance support. The impact of AI on accessibility is profound, with implications for L&D and far beyond. AI can improve eLearning accessibility in ways unimaginable just a few years ago; here are a few of them.
Making language more accessible
Grammar tools can help most business communicators; who among us doesn’t misplace punctuation or abuse pronouns now and again? But Grammarly, a free tool that checks written content for grammatical, spelling, and punctuation errors, has gone a step farther, carving out a niche helping people with dyslexia communicate in a polished and professional way. Since dyslexia is one of the most common language-based disabilities, Grammarly and its ilk could help thousands of people land and keep jobs where written communication plays a key role.
The secret to Grammarly’s almost uncanny ability to intuit what word a writer is trying to spell is the broad education of its AI algorithms. According to Lisa Wood Shapiro, writing in Wired, “AI learns from studying millions of documents and other language-based data sets, along with computer generated misspelled words, and, yes, user feedback. For example, with any given spelling error, Grammarly presents one or a few possible corrections. As Grammarly studies the behavior of a subset of users, it sees which replacement spellings users accept and ignore. That information is incorporated into the options offered up to users in the future.” This is machine learning at work.
Automating access to text and audio
AI-powered tools offer additional language-based assistance. For instance, VocaSee uses natural language processing and machine learning to automate closed captioning and transcription services. Their initial release focuses on “lecture-type” videos and single-speaker audios, but the company plans to develop the technology to be able to distinguish multiple speakers. They’re also training their algorithms on a variety of regional US accents so transcripts and captions will be more accurate. While the results of the current product require careful editing, the process is very fast and easy, streamlining the process of captioning and creating transcripts for large amounts of video content.
While many other captioning services exist, VocaSee’s speedy turnaround and automated service has advantages. A user can “train” the service with a set of keywords and concepts to increase accuracy. This profile, called a “knowledge domain,” can be industry-specific or department-specific—useful for universities creating bodies of online learning content. The AI-powered software creates captions, a transcript, and a summary.
Other companies, such as Salesforce, have also created tools that summarize long texts. These algorithms either extract key phrases from the text or use a higher-level AI-powered process, called abstractive summarization, that can generate coherent paraphrases of the text, potentially using words that don’t actually appear in the original text.
AI-powered translation apps, similar to captioning apps, convert spoken words into another format, providing access to non-native speakers (or non-speakers) of the language of a presentation or conversation.
Conversion of spoken language into captions and transcripts is of obvious benefit to people who are deaf or hard of hearing; it can also improve access for language learners and anyone who struggles to process speech quickly. Microsoft Translator even has a “live” feature that allows multiple people to simultaneously receive real-time translations into different languages. And the ability to summarize long texts quickly can appeal to busy executives—as well as to individuals with difficulties reading or processing language.
A Seeing AI … app
Accessibility is about more than language, though. Microsoft’s Seeing AI app is one of many tools to combine AI technologies—natural language processing, visual processing, image recognition, and optical character recognition—to assist visually impaired users in interpreting their environments. Seeing AI is a fairly robust tool, with the ability to:
- Read short texts, such as the name on an envelope or a grocery list
- Read longer texts while also providing cues to the text hierarchy by noting headings and titles
- Recognize US currency bills, to assist a blind person in accurately handling money
- Guide a user to hold the camera over a bar code on a product so that the app can identify the product
- Assist users to take photos, then describe the photo
- Cover for people who didn’t add alt text to photos they’ve emailed or tweeted by describing photos in several apps—or in the user’s own photo gallery
A host of more specialized apps perform the same functions. KNFB is a well-regarded text reader, for example, and the LookTel money reader can help users recognize bills in 21 currencies. A related class of apps help those same users navigate unfamiliar environments. Blindsquare and the Sendero Group’s wayfinding apps provide walking instructions, including describing points of interest and informing users of street intersections as they approach a corner.
Increasing accessibility of text and providing information about the visual environment make eLearning tools accessible to learners who have limited vision. The AI powering these tools can also be incorporated into eLearning to offer descriptions of virtual environments or help learners recognize colleagues in person or on screen, improving collaboration and increasing social learning opportunities.
Voice commands overcome mobility barrier
Virtual assistants that respond to voice commands provide convenience to all users, but for people with limited mobility, they can be a game changer. Voice-activated assistants, such as smart speakers, allow people to search the internet, access information, play music or podcasts, or otherwise interact without using their hands. For individuals whose mobility impairments make typing challenging or impossible, this vastly expands their potential to engage with and produce content. As these assistants “learn” more tasks, the ways they can increase access to eLearning and performance support—as well as provide performance support—will continue to increase. Smart speakers’ growing repertoire of skills is based in part on AI technologies like natural language processing and machine learning.
Eye tracking improves VR accessibility
Studying a person’s gaze tells you what she is paying attention to—valuable information to marketers. In a virtual environment, eye tracking can increase the feeling of immersion or “presence”; used in multiplayer VR games, it can improve the social communication among the players’ avatars. But the utility of eye tracking extends beyond the potential to improve games or even gather data about learners or players. Knowing where a learner is looking in a VR-based training offers many opportunities to target content and questions, for example.
It also improves accessibility. If, for instance, some effects or text instructions are cued by a user’s gaze, eye tracking can also provide triggers to cue subtitles that explain the sounds to learners who cannot hear them. Without knowing where that learner is looking, there’s no way to know that she will see the subtitles.
Medium blogger Lucas Rizzotto says that eye-tracking in VR is a big deal because it allows users to interact with content directly, without a controller. For learners who cannot use a mouse or pointer or easily touch specific areas of the screen, eye tracking allows them to interact, period.