This paper presents the journey towards accessible web product development in a large international publishing and information analytics company as seen by an accessibility specialist. ScienceDirect just turned 20 years old in 2017 and has grown from 50 online journals in 1997 to over 3,800 journals and more than 35,000 books across science and health domains. Virtually every major university subscribes to a collection of Elsevier journals or books and accesses them through ScienceDirect. Several transformations have happened with ScienceDirect (and parent company Elsevier) during the journey from a publisher to an information analytics company. ScienceDirect is not simply a place to read journals and books online; it is a research platform with full-text search, curated topic areas, alerts to stay current, and experimental data enhancements. The ScienceDirect development team considers accessibility with every new feature. We have continuously improved usability for people with disabilities through external collaborations and by leveraging new technology such as MathML for math content. Since ScienceDirect got on the accessibility track, strategic sites across Elsevier have adopted accessibility in some way. How did this transformation happen?
COMPANY ACCESSIBILITY POLICY
A successful accessibility programme requires collaboration between many staff from different divisions, including ‘the experts’, Engineering, Product Management, Operations, Editorial, and more. Elsevier's company accessibility policy (https://www.elsevier.com/about/our-business/policies/accessibility) has helped join our divisions to work together towards the common goal of universal design, outlining the principles of technology, standards, user-centred design, and collaborative partnership. This collaborative approach helps ensure that modern accessibility features are adopted for the good of all end users. Content accessibility standards are owned by Operations, which continuously improves the specifications for XML, PDFs, EPUBs, and other formats, based on internal and external feedback.
USER FEEDBACK FROM PEOPLE WITH DISABILITIES
Universal design and screen reader users
While the examples in this paper emphasize people who are blind using screen readers, our accessibility specialists have the goal of providing a user-friendly experience for all disability personas. Screen reader testing is a good way to start as it represents many aspects of the Web Content Accessibility Guidelines (WCAG):
- Perceivable – are alternatives for images and videos provided?
- Operable – can the site be used with keyboard?
- Understandable – are forms labelled and error messages usable?
- Robust – are user interface controls designed for compatibility with screen readers?
For example, form control labels are easy to test using screen readers and represent a feature that benefits other disability personas, including voice input/mobility impaired users and users with fine motor control issues. We have found that low vision screen magnification users have unique strategies for navigating web pages, which can greatly differ from blind screen reader users. All website developers are encouraged to utilize the WCAG 2 standards and to tap into one of the many inclusion labs, which represent all disability types, such as the Digital Accessibility Centre in Wales, UK (http://digitalaccessibilitycentre.org/), the Disability Inclusion Lab (Light for the World) (http://lab.light-for-the-world.org) in the Netherlands, and Paraquad (https://www.paraquad.org/) in the USA.
Early user feedback
In 2001, Elsevier ScienceDirect conducted its first user research on how people who are blind use online scientific journal databases with a visit to Wright State University (WSU), a school referred to as ‘one of the top disability-friendly universities in the US’ (Wright State University, 2016). As an undergraduate at WSU, I was familiar with the accessible tunnel system and Adaptive Lab where we ran the first session. The study introduced me to the web browsing strategies for people who are blind, which back then included turning off graphics, tabbing between links, and relying on the page title to understand where you are on a site. This first session allowed me to empathize with people who experience ScienceDirect differently by hearing content through a screen reader. Accessibility violations like unlabeled graphic links are a double whammy: the links themselves have no purpose, and to a listener, it basically sounds like you are being scolded with a rapid announcement of a meaningless URL. A few months later after a wider audit, accessibility enhancements were released into production and then tested with users at The National Federation for the Blind and The National Library Services for the Blind, Library of Congress. The key goal of the second study was to identify possible barriers and usability issues when using ScienceDirect with a screen reader. We measured task success of logging in, running a search query, and finding a specific journal title. As we heard blazingly fast text read by the Job Access With Speech (JAWS) screen reader, we realized that no one has time to listen to an entire web page top to bottom, and soon, we understood why accessibility best practices exist. Right away, users appreciated that ScienceDirect articles were available in HTML (see Fig. 1), which could be read well by JAWS. We concluded the research with a summary report on the user findings, a user interface specification of accessibility fixes, and finally a press release. This early cycle of user feedback, issue fix, and validation of accessibility improvements still holds today.
User feedback through collaboration
ScienceDirect's accessibility testing was largely owned by internal experts until 2011 when the team embarked on the ‘Article of the Future’ (AOTF), the name for a new article design to support enhancements, with a focus on a friendlier reading experience (Zudilova-Seinstra, Klompenhouwer, Heeman, & Aalsberg, 2014). Because so much new web technology was at play with AOTF, we looked externally to accessibility thought leaders from University of Illinois Urbana-Champaign, including Hadi Rangin (now at University of Washington) and Dr Jon Gunderson. Hadi had a stellar record participating in vendor collaborations from across the Web and publishers (e.g. Google, EBSCO Information Services, and Blackboard to name a few). Hadi's unique blend of a human computer interaction expert with software development skills and a person who relied on assistive technology allowed him to effectively advise us through uncharted territory. Dr Gunderson teaches web accessibility, works on web standards bodies, and develops widely used tools such as AInspector. Through the Illinois Assistive Technology in Higher Education Network (ATHEN) Collaboration, we assembled a group of experts to review an early AOTF prototype. Early testing ensured that we could eliminate inaccessible approaches, and we continually tested the article before launch. Accessibility was considered from design through implementation and QA. The collaboration group involves developers and product managers to help foster a greater understanding of accessible design.
Formalizing the collaboration
Since the launch of the AOTF collaboration, more than 25 members have participated in the Elsevier accessibility collaboration from universities and disability organizations across the globe. Participants appreciate the informal design discussions and that the group has a positive impact on a widely used journal resource. Lucy Greco, Web Accessibility Evangelist from UC Berkeley, is a key contributor to the collaboration group and has been featured in news articles and is an invited expert to our company accessibility forum. Lucy's participation in vendor accessibility collaborations is fully supported by the University of California Chief Technology Officer with the idea that university IT accessibility as well as other users will benefit by what our group produces. As Lucy stated, ‘I have made connections with other thought leaders in the field through the collaboration. Accessibility does not happen in a vacuum’. The diversity within the collaboration helps broaden the accessibility testing to include more assistive technology and systems (see Fig. 2). For example, we found that a Captcha security feature worked fine using one screen reader but not in another. We also address a wider range of usability and user experience issues with a variety of people with differing needs (computer scientists, librarians, researchers, and policy officers). Recently, the group has helped review products from other divisions, such as health sciences courses and three-dimensional (3D) animations around health and body systems.
OTHER CHANNELS OF FEEDBACK
Aside from the collaboration group, Elsevier receives user feedback from other channels such as the company accessibility inbox (email@example.com). This helps open up communication lines between development and users and has led to accessibility fixes turned around in less than 24 hr in one instance (in this case, it was simply adding button labels so that a screen reader user could progress through a massage therapy certification practice exam). The accessibility inbox is advertised on product-specific accessibility support pages, such as for ScienceDirect (https://www.elsevier.com/solutions/sciencedirect/support/web-accessibility), within our company accessibility policy and in each Voluntary Product Accessibility Template (VPAT). Feedback from users with disabilities also helps to guide good e-book features like table headers, alt text for figures, math content, and alternatives for video and audio. A recent example of user feedback affecting an EPUB3 book was that tables with empty cells should either have ‘blank cell’ (to denote presentation) or a dash (to denote blank experimental data).
Michigan State University Libraries have conducted ScienceDirect usability testing with people with disabilities as part of their campus accessibility initiatives. Michigan State University Libraries kindly shared videos of ScienceDirect user testing to help our teams understand opportunities for improvements. Recently, the Big 10 Academic Alliance Libraries conducted an evaluation of ScienceDirect (Big Ten Academic Alliance, 2017) and similar online databases as part of the E-resource accessibility testing initiative. (https://www.btaa.org/projects/library/accessibility/e-resource-testing). A new Benetech service called Global Certified Accessible (https://benetech.org/our-work/born-accessible/certification/) has yielded an actionable set of opportunities to improve our EPUB3 books. A simple example is that our publisher logos need better alternative text. A more substantial example of the audit suggests using a service called MathML Cloud to generate alt text for our already-in-place MathML equations. (However, MathML does not have support in current EPUB reading systems, so unfortunately, many publishers are commenting out MathML content.) Finally, peer reviewed journals such as Library & Information Science Research publish studies on publisher accessibility, such as comparing PDFs from ScienceDirect and its peers (Nganji, 2015). The ultimate goal of any of the learned best practices is to capture the semantics and accessible content in the XML documents and carry features through transformation processes to the final books and journals.
ACCESSIBILITY IN THE DEVELOPMENT PROCESS
Accessibility team composition
From experience, no one person can successfully own and execute an accessible website. The recipe for a successful accessibility team would include five key roles: accessibility expert, designer, developer, business champion, and assistive technology user(s). An accessibility expert should ideally know: design for usability, how people with disabilities interact with websites, and standards such as HTML, Accessibility Rich Internet Applications (ARIA), and WCAG 2.0. The accessibility expert and designer together recommend an accessible interaction model to meet a particular business need or feature enhancement. Considering accessibility during design will help inform matters such as selection of colour palettes and design of error messages. Including people with disabilities as early as possible will also open up innovation opportunities as Microsoft has recently demonstrated with their inclusive design strategy (see Kuang, 2016). The developer role collaboratively works with the accessibility expert on implementing an accessible solution. I have observed that, as time goes on, developers apply gained accessibility knowledge to new features on their own. Technical accessibility testing is completed with early developer versions in partnership with the quality assurance team. The business champion sanctions that accessibility is a priority and helps secure development resources. Finally, the assistive technology user brings in the perspective of experiencing a website from a person with a disability and greatly adds to the usability testing process, starting with early design and through the final delivery.
Integrated testing tools
New ScienceDirect features today are driven by ‘JIRA stories’, which are planned sets of requirements managed within an online system. As an example, a JIRA story may be used to define an enhancement such as how to approach a skip navigation link. The accessibility expert will provide details on how the feature should function and, in some cases, provides HTML snippets or links and screenshots to demonstrate a working example. Once development starts, the accessibility expert employs a number of free testing tools (Table 1), including keyboard compatibility and browser add-ons such as the AInspector Sidebar. Automated accessibility tools cannot catch all the problems and should be only understood to be one part of the toolbox. The expert tests pages using screen readers such as JAWS or NonVisual Desktop Access (NVDA). This is a true simulation of someone who has a disability and relies on assistive technology. After accessibility testing new features, any needed fixes are logged in JIRA, and development work is scheduled. In the case of a complete user interface, transformations such as the AOTF, or a complete redesign, we test a prototype page very early on with our collaboration group. Our team utilizes Camtasia software to create videos in order to illustrate accessibility issues to internal teams. A recent example demonstrated how a screen reader correctly pronounced German text but only when the correct language tag is in place.
|Keyboard||Input device||Testing device-independent input||Keyboard operability signals compatibility with a variety of assistive technology|
|Accessibility Bookmarklets||Browser add-on||Testing common mark-up issues, e.g. headings||Quickly outlines accessibility information on a handful of key issues|
|AInspector Sidebar||Browser add-on||Testing WCAG 2.0 compliance||Quickly displays comprehensive violations and fixes|
|JAWS, NVDA||Text-to-speech engine||Testing compatibility with screen readers||Provides a simulation of the website as experienced by a non-sighted person|
|Camtasia||Video editing software||Demonstrating screen reader issues||Creates understanding and empathy amongst developers|
|JIRA||Software development tool||Defining bugs and enhancements||Provides a way to organize, define, and track accessibility improvements|
|Accessibility checklist||Guidelines resource||Understanding guidelines||Provides a quick way to understand accessibility requirements and standards|
Design for mobile devices
ScienceDirect has been adopting mobile-friendly (responsive) web pages across the entire platform (see Fig. 3). In parallel, a global branding initiative has resulted in a consistent typeface and visual design. Mobile-friendly rebranded pages have been very positive for accessibility in many respects. For one, it has resulted in careful consideration of features that are underutilized or outside the user's main tasks. Reducing the number of web page elements results in a system that is simpler and easier to use for everyone, especially assistive technology users. Too many links can be overwhelming to screen reader users: ‘JAWS announced there are 400 links, ugh’. Mobile-friendly designs adapt well when using browser zoom, which is good feature for everyone. With responsive design, the mobile view and desktop view must be equally accessible, so our team tests each ‘breakpoint’ using checker tools and assistive technology. Mobile menus (commonly called hamburgers) must also be accessible, and content that is visually hidden behind mobile menus in smaller screens must also be hidden from the screen reader user; otherwise, the system state can get very confusing.
SUMMARY OF ACCESSIBILITY BEST PRACTICES AND LESSONS LEARNED
ScienceDirect aims for WCAG 2.0-level AA compliance. This standard is widely adopted as campus IT accessibility policy, and this year, the U.S. Access Board approved WCAG 2.0-level AA as the new digital standard for 508 compliance (see https://www.access-board.gov/guidelines-and-standards/communications-and-it/about-the-ict-refresh). The WCAG 2.0 standard contains several criteria that were not around when we started with ScienceDirect, such as colour contrast requirements for text. Although new techniques and standards have emerged, many of the original best practices are still relevant. Given below is a culmination of lessons learned, spanning the 16-year history of ScienceDirect accessibility.
- Ensure keyboard support: Make sure that all features of the web page are accessible through the keyboard alone.
- Keyboard focus styling: Make sure that keyboard focus styling is visible and provides sufficient contrast to people with low vision.
- Offer a variety of formats: At a minimum, provide HTML and PDF using the accessibility features of the format. EPUB3 are highly accessible formats too.
- Provide page titles: A good page title denotes the system name and state, such as ‘120,393 Search Results – Keywords (web accessibility) – ScienceDirect’.
- Provide ways to skip around pages: A ‘Skip to Main content’ link helps keyboard users navigate past information repeated at the top of each web page. ARIA landmarks allow screen reader users to navigate to the functional section of content on a page. Provide headings structure and a table of contents within articles.
- Provide ways to skip through content: Marking lists of journal article titles as headings allows a screen reader user to just hit the ‘h’ key, for instance, to skip from title to title. The beauty of the headings structure is that it provides an outline of the content and allows screen reader users to navigate to the desired sections.
- Scrutinize tables in the beginning of the editorial process: A poorly designed data table with accessibility features will still not be understandable.
- Provide table headers: Tables with headers allow screen reader users to understand which data cell goes with which heading(s).
- Ensure link and button text is unique and meaningful: If designs call for repeated links like ‘download PDF’, utilize ARIA labels to provide context.
- Use modern technologies that promote accessibility: Use MathML for math equations, which can be displayed in browsers and read by text-to-speech engines.
- Reducing noise helps all users: This means reducing the number of links on the page and careful information architecture to help break up information into logical chunks of information. For example, collapse multi-volume journal issues into expandable years.
- Make it easy to download and print figures: Users who are visually impaired may want to print out an image for discussion with a sighted peer. Students may want to print a tactile graphic or embossed copy. OCR software can extract text from graphics.
- Don't break tables or use empty cells or blank spaces for formatting. Instead, simplify the table design and use style sheets for formatting.
- Don't put too many links on a page: To a non-sighted user, hearing a large number of links present on a page can be daunting, ‘406 links, ugh’.
- Don't put too many features before the actual main content: For example, avoid too much content preceding the article text. ‘All of this information is nice to have, but I just want a linear article!’
- Don't auto-refresh pages with forms such as a Sort By Drop Down. This does not allow parallel selection of other form controls. Unexpected page refreshes can be disorienting to all users.
- Don't use Inaccessible Captchas (anti-spam programmes intended to distinguish humans from bots). Our collaboration group tested a simple checkbox Captcha, and it was rife with screen reader compatibility problems. Captchas that require deciphering blurry puzzles or noisy audio are bad usability for everyone.
Use with care
- Recreating a print-only convention in Web: For example, experiencing symbols, asterisks, superscripts, and subscripts is noisy when read by a screen reader. Understanding encoded symbols often requires visually referencing a different area of the content.
- Repeated text links such as: ‘PDF’, ‘more’, ‘download’. Non-sighted users will not be aware of relationships implied by visually styled or spatially positioned content.
- Progressive loading of content such as a journal article whereby users scroll and additional article content is loaded. Screen readers create virtual copies of web pages, and without all content loaded, the overall document structure will be incomplete. Progressive loading often happens through scrolling and may not be obvious to non-sighted users.
- Visually hidden content intended to help people who are blind. Some techniques hide the content for both sighted users and assistive technology, so be sure to test any hidden content with a reader.
- Custom user interface controls such as accordions, fly out panels, hierarchical checkboxes. To be accessible, custom controls will need to rely on appropriate keyboard commands and the use of careful ARIA mark up. Use standard HTML controls for better out-of-the box accessibility.
- Popups, for instance, in suggesting related content or in signalling alerts. The keys to an accessible popup is managing keyboard focus and providing non-sighted users notification of the popup content.
- Links that act like buttons: When elements are coded as links, they will be announced as ‘link’, and there is an expectation that the link will take a user to another page. To a sighted person, the effect of a new page load is immediate, but to a non-sighted person, hearing a link that behaves like a button is confusing.
DRIVING CULTURE CHANGE
Policy and top down
Collaboration group member Hadi Rangin stated, ‘We have the knowledge, we have the expertise, we just need management commitment’. When management speaks, employees listen and that has been especially true of accessibility at Elsevier. Company accessibility policies have existed both at Elsevier (see https://www.elsevier.com/about/our-business/policies/accessibility) and at RELX (see https://www.relx.com/~/media/Files/R/RELX-Group/documents/responsibility/policies/accessibility-policy-new.pdf) since 2011. The effect of the policy has been positive. A new product manager who was previously reluctant around accessibility cited the company policy, and it influenced his decision to prioritize accessibility fixes. It was a product manager who funded the first research into accessibility at the National Federation of the Blind and Library of Congress. Our head of health sciences books provided the edict needed to close-caption all health sciences videos going forward. Overall, several champions from the business have provided the top–down push needed to grease the accessibility wheel. This fits into the mantra that no one person can own accessibility to drive an effective programme. In the past 16 years, Elsevier went from considering the accessibility of one platform ScienceDirect to actively pursuing accessibility in all strategic platforms. Accessibility is now an initiative audited by our corporate audit and risk team.
Tracking customer demand
ScienceDirect has users, editors, and authors who experience disability. We know this because we openly work with people with disabilities to improve ScienceDirect. Tracking data on users with disability is not easily gained nor is it aligned with privacy considerations. Comparatively, tracking data from purchasing institutions on accessibility is clear and easy to perform. Each time our sales teams receive questions like ‘Is ScienceDirect 508 compliant?’, we log the inquiry and share it with decision makers, so the customer demand is clear. Over the years, I have had to do less evangelizing around accessibility because the customer data speak the business case in a clear language.
The role of the VPAT
When a customer asks about ScienceDirect accessibility compliance, we send a VPAT. VPATs are documents that provide compliance details on where a website does and does not comply with standards. Elsevier has been creating VPATs since 2007, when universities such as California State University adopted the VPAT as a tool to validate vendor databases against the campus IT accessibility policies. We update the ScienceDirect VPAT annually to ensure synchronization with the system's feature set. VPATs with honest and rich details also help instructors and disability services offices understand when accommodations may be needed or if there are any known workarounds. The ScienceDirect VPAT is readily available on the ScienceDirect web accessibility page (see https://www.elsevier.com/solutions/sciencedirect/support/web-accessibility).
Alt text and figures
Alternative (alt) text for images is a simple concept – provide a short text description of a graphic or button in the HTML source. Alt text becomes tricky when considering complex figures such as human anatomy diagrams or visualizations. Elsevier authors produce millions of new images a year. The easy answer to creating alt text may well be to treat alt text like any other part of a manuscript. Publishers could require authors to create alt text, which is then subjected to the editorial and peer review process. Elsevier's document type definition already supports alt text, so once in the XML, it will carry through to all downstream formats. The big challenge will be to entice authors to supply alt text as a new step when manuscript submission is already labour intensive. Will authors understand what alt text is, how long to make it, how to correctly describe figures? Surely, there are opportunities to gracefully link to best practice examples such as the DIAGRAM Center's image description guidelines (see http://diagramcenter.org/table-of-contents-2.html). Taking a cue from Microsoft's Office 360 accessibility checks, journal authoring systems can similarly promote accessibility checks while content is being created. Perhaps our authoring tools can provide a system-generated default alt text based on machine learning using surrounding caption text and the image itself. Then, the tools could allow authors to approve or edit the alt text. The main idea is to help authors along with creating appropriate alt text. For some complex figures, it is clear that alt text, no matter how descriptive, will not be a sufficient alternative. In that case, tactile graphics or 3D printing will be the formats to better convey spatial relationships within complex visualizations.
Third-party content, such as banner ads and article enhancements, are not exempt from accessibility as users expect the entire website to be accessible. There are two strategies to help with accessibility of third-party content. One is to research the accessibility support of the available vendors and select the superior option. A media player is a perfect example of where there are vast differences in accessibility support between vendors. Before adopting third-party web content, request a VPAT and accessibility statement, and you will quickly get a sense of a company's commitment to web accessibility. The second approach is collaborative. Communicate to the third party that accessibility is a priority with an offer to work together on accessibility enhancements. Elsevier has recently collaborated with Highsoft Highcharts to help improve the accessibility of the web chart library (see https://www.highcharts.com/docs/chart-concepts/accessibility). The result was a successful accessibility chart module with integrated screen reader and keyboard support. An additional challenge presents itself when utilizing third parties to deliver final formats. As an example, ScienceDirect and other publishers are still working on how to fully tag PDF content (Nganji, 2015). Unfortunately, screen readers today announce PDF content within tags and disregard everything else. Workarounds exist, such as the Acrobat auto tagging feature, but the headings and other semantics do not come out as well as a carefully tagged PDF.
ABOUT SCIENCEDIRECT ACCESSIBILITY
The journey to become an accessible research platform is ongoing. ScienceDirect will continue to evolve to meet researcher needs utilizing new technology. Accessibility is not an achievement to be won and done but a culture paradigm that integrates inclusive design into the very early phases of product development. As interactive content emerges in health science simulations (e.g. 3D animation depicting digestion), our collaboration group will be a key resource to improving usability and accessibility, all the while driving innovation. This paper presents several operational ideas for creating a culture of accessibility within a research platform for consuming content. For an overview of issues with authoring accessible content, such as with the manuscript review process, see Gies, Boucherie, Narup, Wise, & Giudice, 2016.
Users having trouble with ScienceDirect or people interested in joining the Elsevier accessibility collaboration may contact the accessibility inbox at firstname.lastname@example.org. A summary of accessibility features is available at: https://www.elsevier.com/solutions/sciencedirect/support/web-accessibility. Here's to the next 16 years of inclusive access on ScienceDirect.
Case Studies in Thermal Engineering is an open access journal. If articles are accepted for publication, authors are requested to pay an Article Processing Fee. Following payment of this fee, the article is made freely available to all on www.sciencedirect.com.
Case Studies in Thermal Engineering provides a forum for the rapid publication of short, structured Case Studies in Thermal Engineering and related Short Communications. It provides an essential compendium of case studies for researchers and practitioners in the field of thermal engineering and others who are interested in aspects of thermal engineering cases that could affect other engineering processes. The journal not only publishes new and novel case studies, but also provides a forum for the publication of high quality descriptions of classic thermal engineering problems. The scope of the journal includes case studies of thermal engineering problems in components, devices and systems using existing experimental and numerical techniques in the areas of mechanical, aerospace, chemical, medical, thermal management for electronics, heat exchangers, regeneration, solar thermal energy, thermal storage, building energy conservation, and power generation. Case studies of thermal problems in other areas will also be considered.
Fundamental thermal science and engineering studies are generally not included in the scope of the journal.
Manuscript length should not exceed five journal pages.Hide full Aims & Scope