Research
What we measured
Eight questions about braille that have numerical answers, measured here. Six are measurements this site takes of itself; the other two are what those measurements turned up about the reference implementations everything here is measured against.
This is not a blog. Nothing here is published to a schedule, and there will not be another piece until there is another thing worth reporting.
What these have in common is that the number already exists. This site's Grade 2 output is compared against the reference implementation; its page estimator lays out a whole novel to calibrate itself; and where a contraction may sit inside a word was first settled by surveying 11,014 English words against the reference. Those measurements were sitting in the repository doing their job silently. The write-ups are the method and the reading.
The figures on these pages are measured again by the checks that run before every change, and the checks are published. When the translator improves and a figure moves, the page is corrected and the change is entered, with its date, in the log on the about page.
The eight
How accurate is this free braille translator?
“Grade 2” is easy to claim. This translator is measured against the reference implementation, and the one word where it and the reference disagree is named.
How many braille pages will my book be?
The question is usually put as a ratio, braille pages to print pages, and no single ratio is right. Here is the arithmetic, measured by laying out a whole novel.
Which contractions do the work?
Unified English Braille has 182 contractions. Counted over a whole novel, 16 of them do half the saving, “the” alone does 12.1%, and 24 are never written. Every saved cell is accounted for.
What Grade 2 does to each letter
A novel prints 107,660 letters and contracted braille writes 45.0% of them as letters. The letter h keeps 16.4%, and the cell for x is written 811 times for 148 x’s, because alone it is the word “it”.
Which shortforms can carry an ending?
One list of endings for all 75 shortforms would be simple to learn. Measured against the reference over 628 words, such a list is wrong in both directions.
Where a contraction may sit inside a word
The rule is that the letters must keep their meaning, which spelling cannot decide. Over 11,014 words, position predicts what the reference does — but the rulebook decides word by word, and since 18 September 2026 so does this site.
Eleven bugs we found in the reference braille tables
liblouis is the implementation this whole site measures itself against. Building Arabic, Spanish and Portuguese engines against it, and checking this site's punctuation against it, surfaced eleven rows where it departs from its own stated rule or from the published standard — each with a file, a line, and a test that fails the day it is fixed.
The Hindi table in liblouis before 3.37.0 is out of date
Bharati Braille 2.1 was released in January 2026, and liblouis shipped a matching Devanagari table in 3.37.0 that March. Every release before it, 3.29.0 included, carries the 2014 table, which writes 250 of the 5,000 commonest Hindi words differently.
Why publish the failures
The conformance piece names the one word where this site's translator and the reference disagree, in a table, with the reference's answer beside its own.
It is also the only version of the claim that means anything. “Accurate Grade 2 translation” is a claim anyone can make, and it is unfalsifiable. A number with a method and a list of failures can be checked, argued with, and — if it is wrong — corrected by someone who knows better. That is worth more than the fraction of a per cent the failures cost.
If you read braille and something here is wrong, say so.
Reuse any of it
The text on these pages is CC BY 4.0 and the figures are facts. Quote them, put them in a lesson, argue with them in print. The underlying tables are also available as JSON if you would rather have the data than the prose.