Equidox AI is a fully automated solution for high-volume PDF accessibility. Federal mandates require organizations to address digital accessibility. PDF accessibility is an often overlooked part of those initiatives. Manual remediation for thousands of PDFs is inefficient and impractical. It is labor intensive, expensive to outsource, and requires staff with accessibility skills who can make errors, all while facing challenging deadlines to complete the work. Equidox developers, data scientists, and PDF accessibility experts collaborated to build a solution for these issues called Equidox AI. Equidox AI is a fully automated PDF remediation solution for high volumes of templated documents. This includes system-generated PDFs that have similar layouts and structures such as Physicians’ Directories, policy documents, statements, digital ID cards, explanations of benefits, and other templated PDFs that are produced en masse. Equidox AI is designed to fully automate the remediation of these types of documents, eliminating the need for human remediators. Equidox AI can be deployed on Prem or in the cloud. The technology can be integrated into an existing document creation workflow via an API, operated on demand through the user interface, or handled by Equidox staff as a managed service. To remediate a batch of PDFs using Equidox AI, simply upload your documents either individually in .zip folders or using our API, which would seamlessly integrate into a document creation workflow. In this small-scale demonstration, we’re using a .zip file with 20 sample bank statements, and running the batch through our Equidox AI interface. With the .zip file uploaded, I will select the correct ML model to apply to the PDFs and click “Run Batch.” The interface shows the number of documents uploaded, the time elapsed, the status of the remediation, and the completion of each document. As you can see the documents are already in process and some are finished in mere seconds. While Equidox AI is running, let’s take a quick look at what the tag tree looks like before remediation. As you can see, there are no tags in this document at all, and it will not provide any information to an assistive technology user. A screen reader would report “empty document.” Many people with disabilities couldn’t read this document because their assisted technology relies on digital tags to communicate with their screen reader or Braille display. Returning to Equidox AI, this sample batch of 20 documents was remediated in 84 seconds–less than a minute and a half. With increased processing power, Equidox AI can be scaled and sped up to remediate hundreds, thousands, or millions of pages. The remediated documents can be downloaded by the user or delivered to its destination directly through our API. Now, if we look at the tag tree on the remediated document, it clearly has tags. These accurately reflect the text, headings, and images with alt text. Even the tables have been tagged. And the reading order is correct as well. Not only will this remediated document pass an automated checker but it will pass a manual review using a screen reader as well. Now that you’ve seen Equidox AI in action, imagine how this cutting-edge technology will streamline the remediation process for your organization. Equidox is the leading provider of PDF accessibility solutions. To learn more about Equidox AI and how it can streamline the PDF remediation process for your high-volume PDF documents email us at EquidoxSales@Equidox.co or call 216-529-3030, or visit our website www.Equidox.co.

My name is Dan Tuleta, Senior Sales Engineer  with the Equidox accessibility team. Today   I’m going to be giving a quick demonstration  of our Equidox PDF accessibility software.   Equidox is a web-based application that can  be operated directly through your browser. It   also uses a concurrent user licensing model,  which allows for a larger number of users to   share access. It is only limited to the number  of people who are logged in simultaneously.   Another nice aspect of being web-based  is that Equidox allows for collaboration.   You can share documents with other  users in your account, and you can   even have multiple users remediating  the same document at the same time. The first thing I will do is import  a document. I can open my folders and   select a document or multiple documents to  drag and drop into the import screen. Once   the document finishes uploading to the  cloud, a blue import button will appear.   Pressing this button will trigger an automated  process where Equidox will begin to analyze the   document for either existing tag structure or, if  the document has not yet been previously tagged,   Equidox will automatically identify things like  text, images, hyperlinks, and even fillable form   fields. Returning to the document list, I can  now see the document that I just imported. When   I click on the document, I will arrive at the  document detail page. From here, I can see a   thumbnail of all pages in the document and adjust  some of the basic properties of the document,   such as the title, the author, or the language  attribute. There is also an Images tab that   allows me to add alt text or artifact any  images in the document from one location. If I click on the thumbnail for the  one and only page in this document,   I will enter the remediation page, where I  can make changes to the reading zones on the   document. The reading zones are represented by  these yellow boxes that cover up the content,   and they will ultimately create the PDF tag  structure when we export the document. Since   this document was untagged to begin with, most  of the zones on this page say “TX” in the upper   left-hand corner to indicate that they are text  zones. I can also see zones with the IMG label   to indicate where the images are located. In the  upper right-hand corner of the zone, I can see   a number that reflects the reading order of the  page. If I am unsatisfied with the size, location,   and number of reading zones on the page based on  where Equidox has auto-detected the content, I can   use the zone detection slider to adjust the way  that Equidox analyzes the page. If I slide further   to the left, Equidox will be much more focused  and create very granular reading zones, whereas   if I slide further to the right, Equidox zooms  its focus out and gives me much larger zones. The purpose of the slider is to give you different  options as starting points, so there’s less manual   work required to establish zones and manipulate  their size and location. Another very important   feature of Equidox is the button that looks like  a computer monitor. When I press this button,   a separate tab will open in my browser and show  me an HTML preview of the page I’m currently   working on. The reason this HTML preview is  important is because this is a simple and   linear representation of how a screen reader would  read the content on this page. If I were to stop   working and export the document, I can use this  HTML preview to validate the accuracy of my work,   because this HTML represents what the PDF  tags will become when I export the document. Returning to the PDF page, I’m going to start  adjusting my zones to create the proper tags   on this document. The first zones that I  like to set are headings. The main title   on the top of this page, in the large bold  font, is going to be my heading level one.   Under the zone properties, I can use the  dropdown menu and change it to a heading,   or I can use keyboard shortcuts. In  order to set a heading level one,   I can simply press one on my keyboard, and I  will repeat that step with my heading level   twos and my heading level threes by hitting  two and three, respectively, on my keyboard. Returning to the HTML preview, I can now see  the headings identified by this large bold font.   The next element I will work on is this nested  list, which you can probably tell from the HTML   preview is not tagged as a list at all. We use  computer vision and machine learning techniques,   and we have trained Equidox to identify list  items and nested list items automatically.   All I have to do is hit L on my keyboard to  set the element as a list and then use my   list detection slider. In just a second  or two, Equidox has located a list item   and multiple layers of nesting inside of them.  Returning to the HTML preview, instead of that   list being read as a strange run-on sentence  without any punctuation, it will now have a   proper list tag structure and be fully usable and  understandable for an assistive technology user. The next element to focus on is this table, which  currently looks nothing like a table at all in   the HTML preview. If this table was ignored, an  assistive technology user would hear all that data   read in a random order without any ability to  navigate or understand what they were listening   to. Tables are notoriously difficult and  time-consuming to tag properly. However,   with Equidox, all I need to do is click and drag  to create a single zone on top of the table. Next,   I will hit “T” on my keyboard to change it  to a table and double-click on the zone to   open the table editor. Inside the table editor,  there are now some green grid lines that Equidox   has drawn for me. I can click to manipulate the  location of these grid lines, but I prefer to use   the table detector. Similar to the list detector,  the table detector sliders are also using computer   vision and machine learning to identify  the location of all the cells in the table. If I take a look at the HTML preview of this  table, I will see that instead of having a   random string of information, I have a real table  structure with rows, columns, and table data.   However, there are a couple of other things  I need to do to make this table perfect. The   years 2023 and 2024 need to be spanned or  merged across multiple columns. To span,   I simply select the cell where the  span begins, hold shift on my keyboard,   and select the cell where the span ends. Next,  I just hit “S” on my keyboard. I will repeat   this process for the other part of the table.  Returning to the preview, I can see that all of   those extra cells have now been consolidated. The  other thing I need to correct on this table is to   identify that there are two levels of column  headers. Not only is the year a column header,   but the individual quarters within each year  need to be tagged as headers as well. By default,   Equidox assumes that each table has a single  column header and a single row header. In this   example, we have two column headers, so I just  have to hit the up arrow to change the column   header from one to two. Lastly, Equidox is able  to provide an accurate table summary by simply   hitting this check box. Although not necessarily  a requirement, certain accessibility checkers will   flag tables for not having table summaries.  We have automated this task for our users. One final look at the HTML preview will  show that both of the top two rows and the   far left column are in bold font, confirming  that those cells will be tagged as headers,   while the standard font will be correctly  tagged as table data. Returning to the full PDF,   I cannot forget about the two images on this  page. In the upper left-hand corner of the page,   we have our Equidox logo. I can add alt text to  this image by typing into the description field   that appears when I select the image zone. Halfway  down the page, I have another image of our logo,   but this time there is a dog in the picture.  This is not a document about dogs, and there’s   nothing critical being displayed in this image  that will allow an assistive technology user to   better understand the information on this page.  So I might elect to artifact this image to avoid   filling up the page with redundant alt text. I  can artifact this image or any zone by hitting   backspace on my keyboard. The zone will disappear,  removing my responsibility to add alt text to it.   The visual representation of the image will still  be there in the PDF, but an assistive technology   user will skip past it because it will not  be included in the tagged reading order. If I now take a final look at my HTML preview,  I have just one final issue to fix: my reading   order is not exactly what I want it to be. We can  see that the table has found its way to the very   bottom of the reading order because I manually  drew that zone. If I return to the page editor,   I can press the reorder button, and Equidox will  reorder the page with the default top-to-bottom   reading order. There are also multi-column layout  options depending on the structure of the page.   You can select individual zones to reorder  them, and there’s even a method for selecting   multiple zones to reorder zones in clusters for  pages requiring very complex reading orders. One final look at the HTML preview will show that  the reading order is correct, and I can save my   work and close out of the document. The final  step is to export the PDF. Keep in mind that with   Equidox, you can also export the PDF as an HTML or  an ePub file; however, the majority of our users   prefer to maintain the PDF structure. By pressing  “Generate PDF,” Equidox will produce a brand new   document for me. This document will be exactly  the same as the original except it will be tagged,   fully accessible, and most importantly, fully  usable for assistive technology users. Keep   in mind the original version of this document  was completely untagged and therefore useless   to a screen reader user, and of course, fully  non-compliant with all Web Content Accessibility   Guidelines. The version we just exported from  Equidox, however, is fully and accurately tagged,   100% compliant, and completely usable for  people who rely on assistive technology. Equidox also offers an add-on feature that I would  like to briefly mention. This feature is called   “Zone transfer,” which allows users to instantly  transfer the reading zones from one version   of a document to another version of the same  or similar document. There are a couple of   different methods of doing this, but the simplest  way is to just copy and paste the document ID out   of the URL and paste it into this text field  on the user interface. This functionality is   extremely valuable for our customers who generate  documents from consistently formatted templates,   make last-minute revisions or typo  corrections to documents, or if they   just find themselves working through a backlog  of documents that have a lot of similarities. For more information about how Equidox Software  Company can help you with PDF accessibility,   email us at EquidoxSales@Equidox.co  or give us a call at 216-529-3030,   or visit our website at www.Equidox.co.