Arabic Bidirectional Typesetting for Regulated US Documents
Arabic bidirectional typesetting is controlled by three layers: the Unicode Bidirectional Algorithm, explicit direction markup or isolation controls, and layout checks in the final format. A regulated document is correct when Arabic text, embedded English terms, numbers and punctuation display in the intended order, and when a named owner has signed off each layer with a recorded quality assurance (QA) check.
Table of Contents
What is bidirectional text and what does it cover in regulated documents?
How do numbers, punctuation and Latin terms behave inside Arabic text?
What are the steps, owners and controls for correct Arabic bidirectional typesetting?
What decision criteria separate acceptable from unacceptable Arabic layouts?
What is bidirectional text and what does it cover in regulated documents?
Bidirectional text runs in both directions on one line. Arabic letters flow right to left. Digits and embedded Latin text flow left to right. The Unicode FAQ on writing direction states that Arabic and Hebrew are more complex than simple right-to-left scripts for this reason: digits display the other way, and the text is often mixed with Latin script on the same page.
The Unicode Bidirectional Algorithm (UAX #9) defines how software turns stored text into displayed text. Unicode stores characters in logical order, the order in which they are read and typed. The algorithm computes the visual order at display time. The current version of the standard is Unicode 18.0.0, published 2026-09-01.
The ICU documentation calls the predominant direction of a text its global orientation. An Arabic paragraph has a right-to-left global orientation, with left-to-right segments nested inside it. An English paragraph with an Arabic product name is the reverse case.
Scope in regulated work includes:
Medical device labeling, instructions for use and patient information.
Clinical and regulatory submissions with product codes, dosages and units.
Contracts, court filings and affidavits with names, case numbers and dates.
Financial statements and disclosures with currency symbols and percentages.
Software user interfaces, help content and compliance portals.
Technical manuals and defense documentation with part numbers.
Typical failures fall into two groups. Shaping failures affect letters. Ordering failures affect numbers, punctuation and embedded Latin strings. Both change the meaning of a regulated document.
How do numbers, punctuation and Latin terms behave inside Arabic text?
Numbers have weak directionality. The W3C article on bidi basics explains that a number inside right-to-left text is treated as part of the surrounding Arabic run. The digits still read left to right, while the sentence flows right to left.
Punctuation is also weak. Its displayed position depends on the neighboring characters. A period or parenthesis placed after an English term at the end of an Arabic sentence can appear at the wrong end of the term. The algorithm makes assumptions about adjacent runs of opposite-direction text. The W3C bidi test results note that those assumptions do not always hold, and that isolation controls exist to give the algorithm explicit hints.
The isolation controls are:
RLI: right-to-left isolate.
LRI: left-to-right isolate.
FSI: first-strong isolate, which detects direction from the first strong character.
PDI: pop directional isolate, which ends an isolate.
Isolation prevents bidi effects from leaking between runs. The same W3C results page gives the example of a following number being affected by adjacent Arabic or Hebrew text when no isolation is used.
What are the steps, owners and controls for correct Arabic bidirectional typesetting?
Each step below has one owner and one control. Sections of this article stand alone, so each step is written to be read without the others.
Step 1. Define the direction model in the source files
Owner: localization manager, with the engineering or content lead.
Control: a written direction specification. It states the base direction of each document type, the treatment of embedded Latin terms, and the rules for numerals and units.
The base direction is set first. In HTML, the W3C inline bidi examples show that dir="rtl" on the html element sets the whole document to right to left, and that all elements inherit it. In print files, the base direction is a paragraph attribute set in the layout application.
Step 2. Prepare the source for isolation
Owner: content author and engineering.
Control: string externalization review. Variables, product names, ISBNs and codes must be separable from the sentence they sit in.
The W3C article on strings and bidi shows that wrapping inserted strings in bdi fixes punctuation, author order and ISBN display in an Arabic template. Concatenated strings produce these errors because the code joins fragments without direction information.
Step 3. Apply the right isolation method
Owner: front-end developer or DTP specialist.
Control: a method table in the style guide.
The W3C guidance on bidi isolation in HTML is specific. The Unicode Consortium recommends isolation as the default for future inline embeddings. The dir="auto" value should be used only for content of unknown direction. When the direction is known, dir must carry an explicit ltr or rtl value. The bdi element isolates its content and guesses the direction with the same first-strong heuristic.
Situation | Method | Control |
Whole document or section in Arabic | dir="rtl" on the container | Template check |
Known English term inside Arabic | dir="ltr" on a span | Style guide rule |
User-supplied or database string | bdi or dir="auto" | Code review |
Plain text outside markup | Unicode isolate characters | Encoding check |
Step 4. Typeset with RTL-capable tools and fonts
Owner: DTP specialist.
Control: approved tool and font list.
Arabic letters change shape by position and by neighboring letters. The W3C Arabic and Persian Layout Requirements note that Arabic fonts typically contain hundreds of glyphs and that lam-alef generally requires a mandatory ligature. The ReportLab Arabic documentation adds that each letter can take up to four forms, and that text without reshaping can appear as disconnected characters.
Layout software must support right-to-left composition natively. The QuarkXPress 2026 documentation lists native RTL Unicode font support, reordering by RTL orthographic rules, glyph positioning and RTL spell checking. The same document explains that a right-to-left facing-page layout begins its first page to the left of the spine, the reverse of a left-to-right layout. Page order, binding side and page numbering must follow that model.
Font developers should follow the Microsoft OpenType guidance for Arabic script on character sets and complex script features. A font that lacks required features fails shaping regardless of the layout tool.
Step 5. Preserve terminology and translation assets
Owner: translation manager.
Control: translation memory and term base governance.
Term bases must state whether a term stays in Latin script, is transliterated or is translated. That decision drives the direction markup in Step 3. A term that changes script between releases creates inconsistent layout and inconsistent regulatory language.
Step 6. Run linguistic and visual QA on the final format
Owner: Arabic subject-matter reviewer and DTP proofreader.
Control: QA checklist applied to the rendered output, not the source file.
The checklist covers:
Letter shaping and connected forms.
Position of every number, unit and percent sign.
Position of parentheses, quotation marks and final punctuation.
Embedded English terms and codes.
Table column order and text alignment.
Page order, headers, footers and page numbers.
Hyphenation and line breaks.
Step 7. Record sign-off and archive
Owner: project manager or compliance director.
Control: audit trail. It records who reviewed each layer, which version was approved and which tool and font versions produced the file. Regulated customers can then reproduce the approved output on request.
What decision criteria separate acceptable from unacceptable Arabic layouts?
The criteria below apply to any regulated Arabic deliverable.
Fidelity of data. Dosages, amounts, dates, part numbers and legal references display in the same order and value as the source. This criterion outranks appearance.
Direction correctness. The base direction is right to left. Embedded Latin terms and numbers read correctly in place.
Shaping correctness. No disconnected letters, missing ligatures or substituted glyphs.
Structural mirroring. Tables, lists, page order, navigation and progress indicators follow right-to-left logic where the layout requirements call for it.
Reproducibility. A second operator with the same files and fonts obtains the same output.
Data handling. Source documents containing personal, health or confidential data are processed in an environment that matches the contractual and legal requirements of the document owner.
Traceability. The audit trail shows who changed what and when.
A layout that fails criterion 1 is rejected regardless of the other scores.
What do mixed-direction examples look like in practice?
Example 1. Product code in an Arabic instruction sentence
An Arabic sentence ends with a device model code that contains letters, a hyphen and digits, followed by a period. Without isolation, the hyphen and digits can reorder, or the period can display at the start of the code. Wrapping the code in an isolated left-to-right span keeps the code intact. The period then attaches to the end of the Arabic sentence.
Example 2. Reviewer name and score in a template
The W3C strings and bidi article shows that isolating a reviewer’s name and score with bdi or dir="auto" fixes the score layout in an Arabic template. The same pattern applies to user names next to account balances in banking interfaces.
Example 3. Arabic title that starts with a Latin word
The same article explains that an isolated string takes its direction from its first strong character. An Arabic title that begins with the Latin word “CSS” is therefore treated as left to right. Its words and punctuation can then display in the wrong order. The fix is an explicit direction value when the direction is known.
Example 4. Legal document with names, case numbers and dates
A court filing in Arabic contains English party names, a docket number and a date. Each item needs the right isolation, and the translated text needs consistent handling of names. A reference on the legal document translation process describes how translation fits into legal service and compliance workflows. The typesetting controls above apply to the resulting files.
Example 5. Social and messaging platforms
The W3C page on bidi in social media reports that one platform appears to use first-strong detection and another appears to compare the count of RTL and LTR characters. It includes test cases where each succeeds or fails. Marketing teams that publish Arabic posts with English hashtags and brand names should preview them on each platform.
Example 6. Right-to-left email
A Litmus-style email question on RTL design is answered by the same principle: set the direction on the container and isolate English brand names and numbers inside it. Email clients vary, so each template is tested in the clients the audience uses.
What are the failure modes and mitigations?
Failure mode | Cause | Mitigation | Control |
Disconnected Arabic letters | Missing reshaping or incomplete font | Use shaping-capable tools and a complete OpenType font | Font list, visual QA |
Misplaced punctuation | Weak directionality of punctuation next to opposite-direction text | Isolate the embedded run | Style guide, QA checklist |
Reordered numbers or codes | Adjacent runs interact | Wrap in isolate or explicit LTR span | Code review, QA |
Wrong direction on dynamic strings | First-strong detection on text that starts with Latin | Use explicit direction when known; bdi for unknown | Template review |
Broken markup display in editors | Editors treat markup as ordinary text, a limitation the W3C note on authoring right-to-left HTML describes | Use editors with bidi-aware source views | Tool approval |
Reversed page order | LTR facing-page setting on an RTL book | Set RTL page layout at file creation | DTP checklist |
Inconsistent terms | No term base governance | Maintain term base and translation memory | Terminology review |
Browser differences | Rendering engines differ in edge cases | Test in target browsers; the W3C text direction tests report Gecko, Blink and WebKit passing the Arabic alignment test for dir=auto | Cross-browser QA |
Unreviewed final file | QA done on source only | Review the rendered PDF or build | Sign-off record |
The Unicode Technical Committee review notes state that the algorithm is used to display all Arabic and Hebrew text on the web and in applications, and that changes after release can cause interoperability problems. Teams should therefore pin tool versions and record them in the audit trail.
Perspective on workflow design
Arabic typesetting errors in regulated content are rarely language errors. They are process errors. A correct translation can be damaged by a missing isolate, a font without the required glyphs or a QA pass done on the wrong file. We treat direction handling as part of translation quality, not as a late formatting step. Terminology governance, direction markup and layout review belong in one controlled workflow with named owners. Our AI+HUMAN hybrid model supports this. Subject-matter linguists review the Arabic output, and the output is checked in its final format against the term base and the customer’s translation memory. The audit trail records each step. Regulated customers also need to know where documents are processed. We run our systems on private EU-hosted infrastructure under ISO 27001 and ISO 42001. Eric Brown
Where AD VERBUM fits
AD VERBUM provides translation, localization and multilingual documentation with AI+HUMAN hybrid workflows, including DTP. Its proprietary LangOps System runs on private EU-hosted infrastructure. The company holds ISO 17100, ISO 18587, ISO 27001, ISO 42001 and ISO 13485 certifications, audited by Bureau Veritas. It supports 150+ languages, integrates client translation memories and term bases, and draws on 3,500+ subject-matter linguists. See the AD VERBUM services.
Key points
Unicode stores text in logical order and computes display order with the bidirectional algorithm.
Numbers and punctuation have weak direction, so they need isolation next to opposite-direction text.
Use explicit ltr or rtl when direction is known, and bdi or dir="auto" for unknown strings.
Letter shaping depends on complete fonts and shaping-capable tools.
Right-to-left layouts reverse page order, so set the layout direction when the file is created.
Run QA on the rendered output, not only the source.
Assign one owner and one control to each step, and keep an audit trail.
Pin tool and font versions to reproduce approved files.
FAQ
What is the difference between dir, dir=“auto” and bdi?
The dir attribute sets the base direction of an element and isolates its content. The value dir="auto" detects direction from the first strong character. The bdi element isolates its content and uses the same detection. Use explicit dir values when the direction is known, as the W3C guidance states.
Why do numbers appear to move in Arabic text?
Digits in right-to-left scripts display left to right and have weak directionality. The algorithm treats a number as part of the surrounding Arabic run, as the W3C bidi basics explain. Adjacent Latin text or punctuation can change the visual result. Isolation fixes the order.
Does right-to-left DTP require special software?
Yes. The layout tool must support right-to-left composition, RTL Unicode fonts, glyph positioning and RTL page order. QuarkXPress 2026 documents these features. Fonts must also contain the glyphs and features Arabic requires.
What should a QA checklist for Arabic layouts include?
It should cover letter shaping, number and unit positions, punctuation, embedded Latin terms, table order, page order and alignment. It should be applied to the final rendered file. The reviewer, tool versions and approval date go into the audit trail.
Sources
