A syllabus is a pipeline problem
Converting a prescribed textbook into a navigable audio edition is an afternoon of compute. Getting it into a classroom takes a term.
A school for the blind waits a full term for one accessible edition of a set text. This is not for want of technology. Speech synthesis is a commodity. Layout parsing is a commodity. Translation across twelve Indian languages is a commodity. Assembled properly, the pipeline produces a navigable audio edition of a two-hundred-page textbook in an afternoon.
The delay lives entirely in coordination. Who holds the authoritative file. Who approves the conversion. Which format the reader in the classroom actually supports. Whether the chapter markers match the printed edition the sighted children are using, because a child who cannot follow along on page 74 is still excluded.
Every part exists. Nobody owns the assembly.
Accessibility work therefore fails in the joins rather than the capability. Every part exists; nobody owns the assembly. That is a deeply fixable problem, and it is the kind of problem engineering teams are good at when someone asks them to look.
The version we deploy treats the syllabus as a pipeline with a maintainer. New editions flow through automatically. Chapter structure is preserved so a blind student can be on the same page as the class. Large print, audio and text come out of the same source, which means they cannot drift apart.
None of this is research. It is plumbing, and the plumbing has never been anybody’s job.
The five joins where accessibility work breaks
Source: nobody can say which file is authoritative, so conversion starts from a scan rather than the publisher text. Approval: an accessible edition needs a sign-off that no one department owns. Format: the reader in the classroom supports a subset of what the converter emits. Structure: chapter and page markers drift from the printed edition, so a blind student cannot follow the class. Maintenance: a new edition arrives and the whole cycle restarts from zero.
Key points
- Conversion is an afternoon of compute; coordination is a term.
- Five joins break the chain: source, approval, format, structure, maintenance.
- Page and chapter parity with the printed edition is non-negotiable.
- Treat the syllabus as a maintained pipeline, not a one-off project.
What a maintained pipeline changes
When the pipeline has an owner, new editions flow through automatically, audio and large print derive from a single source so they cannot diverge, and structure is preserved by construction rather than by manual checking. The cost per title falls to near zero after the first, which is what turns a pilot into coverage.
Questions this raises
Is synthetic narration acceptable for textbooks?
For most subjects, yes, and it is what makes coverage affordable. Literature and language texts often still merit human narration.
What about mathematics and diagrams?
These need structured markup and described figures. They are the genuinely hard part and should be scoped separately.
Who should own the pipeline?
Whoever owns the syllabus. Ownership by a vendor reintroduces the year-two problem.