Archive for the ‘Research’ Category

AI and Disability Interview

2019/06/17

AXS Chat recently posted to YouTube an interview of me about my artificial intelligence (AI) research and work for people with disabilities. I talk, in part, about:

  • the promise of a text-comprehension parallel between AI and people with intellectual disabilities;
  • how AI-driven Web text simplification will benefit other populations, such as non-native language speakers; and
  • my work to make sure people with intellectual disabilities and/or autism are not left out of online education.

I thank the AXS Chat members, Neil Milliken, Debra Ruh, and Antonio Santos, for their tireless work to inform the world about inclusion and technology.

Amazon re:MARS Accessibility

2019/06/14


Amazon Machine Learning Research Awards generously sponsored my colleagues and me to participate in last week’s Amazon re:MARS Conference. It was a global artificial intelligence (AI) event focused on Machine Learning, Automation, Robotics, and Space.

The conference was great with accessibility. I was assigned an employee who guided me everywhere and was just wonderful. The conference website was accessible and easy to navigate. When I identified accessibility problems with the mobile app and with SageMaker tools, Amazon personnel immediately assured me they would be fixed.

The sponsorship included participation in the re:MARS VIP Leadership Networking Reception. I was honored to speak with members of Amazon leadership as well as senior researchers from industry and academia.

We discussed:

  • my AI-driven, Web text simplification research;
  • AI fairness for people with disabilities; and
  • developing an Alexa skill for DisabilityInfo.org.

 

MIT Library Systems Inclusion Workshop

2019/04/02

MIT LibrariesI will participate this week in a

I plan to discuss my AI Web text simplification research and AI fairness for people with disabilities. More about AI fairness soon.

AI Web Text Simplification: CSUN 2019

2019/03/04

I will soon present part of my AI-Driven Web Text SiCSUN Center on Disabilitiesmplification research.

My talk:

We tested if people with intellectual disabilities understand Web text simplified with plain-language standards. (Spoiler: They do!)

We are operationalizing plain-language standards essentially to develop:

  • a reliable, easy-to-use method for human editors to create simple text; and
  • algorithms for AI to recognize and to create simple text.

 

 

 

 

AI Web Text Simplification: Partners

2019/01/21

For my AI-Driven Web Text Simplification research, I lead a coalition of corporate and academic partners. They include:

AI-Driven Web Text Simplification: Intro

2019/01/07

Research Goal

Make Web text so simple people understand it the first time they read it.

Background

Text comprises the vast majority of Web content. Poor reading comprehension presents significant challenges to many populations, including people with cognitive disabilities, non‐native speakers, and people with low literacy.

Text simplification aims to reduce text complexity while retaining its meaning. Manual text simplification research has been ongoing for decades. Yet no significant effort has been made to automate text simplification except as a preprocessor for natural-language processing tasks such as machine translation and summarization.

Short-Term Approach

In the short term, my partners and I are improving manual text simplification by creating effective, replicable methods for humans to produce it. We use national and international plain language standards. We conduct pilot studies to see if people comprehend our human-curated, simplified Web text better than typical Web text.

Long-Term Approach

In the long term, my partners and I are developing artificial intelligence (AI) capabilities to produce simple Web text on a mass scale. We are training AI with enormous sets of aligned sentence pairs (typical/simple). We will soon start crowd-sourcing the generation of training data.

I will provide details in future blog posts.

Improving Web Searching for People with Cognitive Disabilities

2011/04/25

Using a website search tool is difficult for people with cognitive disabilities. Finding a relevant result is often thwarted by spelling errors they make, their inability to detect them, or a lack of understanding about how to correct them. Determining which search results are best can be equally difficult.

This post is a synopsis of an approach to circumventing such problems. An example has been implemented on a web site of the German Institute for Human Rights, which is an easy-to-read version of a United Nations convention on the rights of people with disabilities. A typically-appearing site search incorporates novel spelling-correction features and a simplified presentation of search results.

Spelling Correction

The site search suggests spelling alternatives only for words that actually appear within the content of the website. Searches for correctly-spelled words that produce no search results would be very frustrating for anyone.

To enable spelling suggestions, a manually-edited index of syntactically-similar words was created. Point values were assigned for similarities in the number of the same letters and the word length. A higher value was given to alternative words with the same first letter, but that was not essential.

To enable search-word spelling correction within the fewest steps possible, the most-similar alternatives are displayed in a word cloud. Of those, typically three, the one with the highest probability of matching the intended search word is presented in a larger text size.

Example Spelling Correction

The German word for “contact” is “kontakt”. Initiating a search with the misspelled word “kontat” produces a word cloud as shown in the following image.

Of the displayed three words, Kommunikation Kontakt Kunst, the second is shown in a larger font. All are hyperlinks.

The developers believe the word cloud makes it very easy to recognize the correctly-spelled word, and to select a search word. I don’t know why the first letters are capitalized.

Simplified Search Results

Search results are presented in plain language. Each has a bulleted, succinct summary of information on the linked page; and a contextually-relevant image to aid comprehension.

Example Search Result

The following image shows a single search result translated from German to English using Google Translate.

Contact - Here you will find: The address and telephone number of the German Institute for Human Rights. And a contact form.

One aspect of the search results I do not favor is that links to the search-result pages are not underlined. It is only when the cursor is hovered over a link, such as “Contact” in the example search result, that an underline appears.

Conclusion

I am impressed with this approach. This is the first time I have seen search results presented so simply, and with accompanying relevant imagery. I think the spelling-correction features are also worthwhile. In a pilot study of them, 9 of 34 people with learning disabilities could use the search site independently. I expect the developers will continue user testing. With funding and time, I would like to develop a site search using similar techniques.

Notes

Sources of Research Articles on Cognitive Web Accessibility

2010/04/01

On The Clear Helper Web Site, I published my

Sources of Research Articles about Web Accessibility for People with Cognitive Disabilities.

Most are vertical search engines of research from related fields.

Each listing is annotated with an edited quote describing the source.  At the top of the page, I note the approximately twenty search terms I used.  The following blog posts are about the results of this effort.

Since I published the above, I have significantly increased the number of articles for each list.

Defining A Good Accessibility Statement

2010/03/09

This post lists the recommendations of eight Web articles I found that opine about a good accessibility statement. I also found three that advocate a site-help page be used instead.  All are referenced below.

Common Recommendations

The survey results are from articles published on the Web by accessibility-focused organizations or by site developers.  One, “Evaluating the Usability of Online Accessibility Information“, is based upon research. Publication years range from 2002 to 2009.

Mentioned In Most Articles:

  • Agreement
    • Do not just list accessibility features; explain how site visitors can use them.
    • Detail any site barriers to accessibility.
    • Provide contact info for people who experience accessibility problems.
    • Make the accessibility statement easy to locate on the Web site.
  • Disagreement
    • Don’t refer to how the site conforms to accessibility standards (50%).
    • Do refer to how the site conforms (50%). Place the info at the bottom of the statement.

Mentioned in Half of Articles:

  • Explain the site’s or the organization’s commitment to accessibility.
  • Do not use jargon.  Use clear, plain language targeted to the site’s audience.
    • Use an alternative to the term “accessibility” because many visitors do not know what it means.

Mentioned in at Least 2 Articles:

  • Separate accessibility-statement content into sections.
  • Reference authoritative, stable accessibility-help, i.e., The BBC’s My Web My Way.
  • Do not limit accessibility information to a specific impairment.
  • Do not assume knowledge visitors may not have, e.g., which browser they use.
  • Do not claim accessibility features if they are not present.

Relevancy To Current Assessment Plan

I decided to investigate this in preparation for my plan to assess the Web accessibility of 100 cognitive disability organizations. Specifically, I considered not just awarding a point for the presence of an accessibility statement, but for the presence of a good one.  To do that, I needed to determine agreed-upon characteristics.  Now that I have, I realize it would take too much time to assess the accessibility of the Web sites and whether or not their accessibility statements, if existent, are good.

Referenced Articles

Articles That Advocate Site-Help Pages Instead

Notes

  • I searched for recommendations by people who identified themselves as having a disability, but found none.
  • Did I miss an important resource?  Please comment or contact me.

50+ Readability Resources Related To Cognitive Web Accessibility

2010/03/04

I have created an index of readability resources related to plain language; measurement tools; guidelines, research; content; symbols; and  free- and commercial products and services. At the time of this writing, there are over fifty. I will add more as I find them.

Characteristics Of Readability Listings

  • All have links to the original sources.
  • All are annotated with related information, primarily edited quotes from source pages.
  • The majority are free- and commercial products and services.  The rest are research articles.
  • The publication dates of original studies and articles range from 2001 to 2009 / present.

Links to Readability Index & RSS Feed

Notes

Technorati Verification Code = 63S9AZXDSA9K


%d bloggers like this: