A 940 page book, turned into an app

If you came here from the dictionary

Then you have already used the thing this page is about. What follows is how it was made, what it took, and how the same can be done with material you already have.

If you would rather skip the story, tell me what you have. It takes about thirty seconds.

Good content, locked in the wrong format

A reference book is years of somebody's knowledge. Printed, it lives on a few shelves. As a PDF it is technically online and practically invisible: Google cannot tell one entry from the next, nobody can search it properly on a phone, and nobody shares page four hundred and twelve.

This one was a maritime dictionary of 940 pages, printed in a hundred copies. The brief was easy to say and hard to do: every word online, fully indexed and searchable.

What it became

10,943 entries, each on its own page, carrying its etymology, derived words, phrases, sources and photographs. 290,011 links between entries, and not one of them broken.

Browsing by letter, by the 26 subject fields the book itself uses, and by the 14 languages the words came from. 353 abbreviations explained on hover. A citation that gives the printed page, so the book can still be quoted properly.

Eight learning games built from the same words: guess the word from its definition, where a word came from, the winds on a compass rose, true or myth, a race against the clock. Four levels, naval ranks with their insignia as the score grows, and a link that challenges a friend to exactly the same questions in exactly the same order.

And the whole dictionary as one offline file. Double click it on a ship, in a classroom or on a laptop with no connection, and it is ready in about a fifth of a second.

Search that understands how people actually type

Most site search quietly fails anyone who does not type perfectly. This one does not ask them to.

It ignores accents, capitals and polytonic marks. It reads Greeklish, including the way people really write it. It understands a word typed with the keyboard still set to English. It finds words that sound the same but are spelled differently, and it survives typos.

It also works backwards. Type the Italian ancora and it finds the Greek word for anchor. There are 5,079 foreign origins you can search from.

Details like these decide whether somebody finds the entry or closes the tab.

The part nobody sees: reading a book correctly

Turning a typeset book into data is not copy and paste. The layout has to be understood before a single word can be trusted.

Here, the start of an entry was only visible as a hanging indent, measured across every page. The symbols marking etymology, derived words and compounds were font glyphs with no printed legend, so what each one meant had to be worked out from thousands of examples.

Some problems are invisible on paper and fatal online. A Latin letter that looks exactly like a Greek one makes a word impossible to find. Hyphens have to vanish inside words and survive inside web addresses. Small words inside a headword can be set in a different weight and still belong to it.

Every one of those was caught by checking the output against the book, never by assuming the parser was right.

Why it sits beside the existing site, not inside it

The dictionary lives in its own folder next to the institution's existing WordPress site. No plugin, no database, and no change to a single page that already exists.

That matters more than it sounds. A new section that puts your current rankings at risk is not an improvement. Added beside the site it can only add: more than ten thousand new pages for Google to index, each marked up so a search engine knows it is a defined term.

It is also fast and cheap to run. There is nothing to update, nothing to patch and nothing to fail at two in the morning.

And the rights come first. Nothing goes public until the rights holder agrees, and quoted work by other authors stays out until it is cleared. For a publisher, that is the difference between a partner and a liability.

The skills a project like this needs

It looks like one job. It is six, and they rarely sit in the same person.

Data extraction, to read a complex document into structured data you can trust. Search engineering for a language with accents, dialects and two alphabets in everyday use. Information design, so ten thousand pages stay easy to browse. SEO at scale, so every page can be found without the site competing with itself. Game design, so learning feels like playing. Interface and brand design, so it looks as if it always belonged to the client.

Twenty five years of industrial design and six of building software meet here, and that combination is why it was working in days rather than months.

Who else this is for

The same work fits anything valuable that is currently stuck in print or in a PDF.

A publisher with a reference book, a glossary or a title that is out of print. A museum or institute with a catalogue, an archive or a collection. An association with decades of bulletins. A manufacturer with manuals and spare part catalogues. A school or training centre that wants its material to be played, not only read.

If you have the content, the hard part is already done. Tell me what you have, send the file, and you get a written scope saying honestly what is possible, what is not, and what it would take, before anything is built.

What you get

• Your book, catalogue or archive read into clean, structured data that you own
• One page per entry, built to be found on Google, with no database to look after
• Search that forgives: no accents, Greeklish, the wrong keyboard, typos, sound alikes
• The whole thing as one offline file that opens with a double click
• Learning games made from your own content, with levels, ranks and challenges to share
• Styled as your own site and added beside it, so nothing you already rank for moves
• A written scope before anything is built, and every figure checked before handover