Fixing tags and reading order in a PDF by hand

Automatic tagging gets a document most of the way. The rest is careful handwork, and that's what the Tags window, the content window and the reading order tool in PDF by Zaweba are for. Between them they let you see and change exactly how a screen reader will meet the document.

The Tags window

The Tags window shows the tag tree, the structure of the document from top to bottom. Here you can make a new tag, wrap existing ones in a new tag, delete, reorder and indent, and change a tag to an artifact.

Those few operations cover most fixes. Wrap is how you build a list: select the items, wrap them in a list tag, and each item becomes part of it. Reorder fixes things that will be read in the wrong place. Indent puts a tag inside another, which is how structure is expressed. Change to artifact removes decorative content from what's read aloud, such as a repeated logo or a dividing line.

Each tag has properties too. Alternate text is the description read in place of a figure. Table header scope tells a screen reader whether a header cell belongs to its row or its column, which is what lets a listener hear "March, Sales, 1,200" rather than a string of numbers.

Marked content per page

The content window shows the marked content on each page: the actual pieces of text and pictures and how they're connected to tags. It's the place to look when something on the page isn't being read at all, or is read twice. Content that isn't connected to any tag is effectively invisible to assistive technology.

The reading order tool

The reading order tool is the most visual of the three. It draws numbered boxes on the page in the order the content will be read. A glance tells you whether it makes sense. A two-column page read straight across both columns, line by line, stands out at once.

You can click a box to give it a type, such as heading or paragraph, and drag a rectangle to make a tag from what it covers. For many pages that's quicker than working in the tag tree, because you're pointing at the page rather than reading a list.

Role map and class map

Some documents use their own tag names, made up by the program that created them. The role map says what standard tag each custom one means, so a custom "Title" tag is understood as a heading, for instance. The class map holds shared attributes. Both are editable in PDF by Zaweba. You won't need them often, but when a document uses odd tag names, fixing the role map can correct the whole file at once instead of retagging every heading.

A practical way to work

Start with the reading order tool on each page, because it shows problems fastest. Fix the order and the obvious types there. Then go to the Tags window for structure that the page view can't show well: lists, tables and nesting. Add alternate text and set header scope as you meet figures and tables. Use the content window when something is missing. Finally, run the accessibility check again and save the report.

Work in a copy, and save as you go. Tag work is detailed, and a saved point to return to is reassuring.

All of this is done offline in PDF by Zaweba, so even confidential reports can be made accessible without leaving your PC.