The Oxford English Dictionary is the reference book in miniature: what it has been through, and what it will become

When was the last time you opened a paper dictionary?

In 1857, a few English churchmen who belonged to a body called the Philological Society found that the dictionaries they could buy had pitifully few entries, and that what entries there were consisted mostly of obscure, odd words. So they decided to start compiling an English dictionary of their own.

In 1933, after 71 years of work, the first edition of the Oxford English Dictionary (OED) was published. The first 22 years (1857–1879) were preparation; over the 49 years of actual editing (1879–1928), the project went through four chief editors, and the team ran newspaper advertisements in Britain and America to recruit some 1,300 volunteers to supply quotations. James Murray, the fourth chief editor of the OED, had to build a small "scriptorium" beside Oxford University to hold the entries, which arrived like snow.

Barely a century later, in 2010, the OED's editors announced that the third edition — the authority on contemporary English — would probably never appear in print.

The news drew a few sighs, but most people thought it hardly mattered. For anyone who lives with the internet and smart mobile devices, looking up a word long ago stopped meaning turning the pages of a paper dictionary. You feed the word into a computer or a phone as an image, a voice, or text, and a few seconds later you have your answer.

The man who ran the digitization of the Oxford English Dictionary was John Simpson, its chief editor from 1993 to 2013.

More precisely: this Englishman watched the OED's corpus go electronic in the 1980s, and from the 1990s on he led the online OED project and the work of revising and expanding the second edition.

Simpson graduated from the University of Reading with a degree in medieval studies and joined the OED's editorial office in 1976, sorting index cards. If you have seen The Great Passage, the film adapted from the manga of the same name, you will have a fair picture of what those cards were. Before dictionary staff routinely keyed data into computers, index cards were the most important tool for gathering material: small slips, one word and its definitions to a slip.

At the OED, no word entered the dictionary until it had been collected and recorded, drafted, checked, and proofread again and again. The editors who wrote entries were split into four groups — general entries, scientific entries, new words, and Germanic and French words — and each had to judge whether a word deserved a place by where it came from, how often it was used, how far it had travelled between countries, and so on. The labor divided further as the work went on: some editors did nothing but pronunciation, tracking and verifying how each word was said by listening to television, film, and radio.

For more than a hundred years, the work went like this: editors and researchers climbing up and down vast libraries, hunting through newspapers and old books for the faint tracks a word had left as it changed through history, then setting them out, with great care, on thin sheets of dictionary paper.

Simpson happened to be standing where that era turned.

"Back in 1989, when the second edition of the OED was published, we began collecting data from printed materials and keying it into computers. (Oxford University Press hired 120 typists, who spent 18 months entering the whole of the second edition — 67 million characters — into computers using a basic markup language.) At the time we set our sights on building a huge electronic database. In the 1990s, once the database had taken rough shape, the world changed again. We began to see what the World Wide Web and the internet made possible — they could change the way we had always searched and published dictionaries," Simpson wrote in his memoir The Word Detective: A Life in Words: From Serendipity to Selfie.

Because of the peculiar nature of their trade, dictionary editors — a kind of "information craftsman" — were early users of electronic databases. As early as the 1980s, Simpson says, they were working with Lexis/Nexis, the American newspaper database. Bit by bit they got used to searching for material and entering data at a desktop computer.

In 1995 the OED team took up the Making of America database, built by the University of Michigan out of its holdings of early American books and magazines. It marked one more step away from databases on CD and toward databases online.

In 2000, Simpson and his team put the OED itself online. From then on, editors could add and revise entries over the network, and readers could subscribe to get the newest entries as they appeared.

"When you get to know a word, you have a rough sense of the product it will turn into in the end. Every word is a poem — some as small as a Shakespeare sonnet, some as large as Joyce's Ulysses. But working on a computer, we realized more and more that the single word is not really the thing to look at; a single word is one tile in the mosaic we call 'language.' In the web of language we can now build, you can see the ties between words far more clearly than before," Simpson told Time in 2013.

So will the paper in libraries die off as the electronic record grows? Simpson asked himself the same question.

He soon decided the worry was largely misplaced. Electronic databases turned up far more material than before, but they did not really make entries any faster to write. People cannot read that much, for one thing; and unreliable material online makes verification harder, not easier. "Sometimes," Simpson said, "you still need sheer human effort, and a resourceful researcher who finds the root of the problem deep in the stacks." The two ways of checking can live side by side.

The Oxford English-Chinese dictionaries became apps at the urging of a Chinese editor named Liu Haoxian.

Liu has been with the Hong Kong branch of Oxford University Press for 16 years. He joined as an editorial assistant on English-Chinese dictionaries. He is now chief editor of the Oxford English-Chinese dictionaries and head of dictionary publishing for Greater China.

Around the time Simpson was nearing retirement, Liu and his team turned the seventh edition of the Oxford Advanced Learner's English-Chinese Dictionary into an app.

"We picked the Oxford Advanced Learner's to test the water in digital publishing because its sales had always been fairly steady," Liu told QDaily.

Around 2011, on behalf of Oxford University Press, Liu put the idea of an app to The Commercial Press. The internet and search engines had sent paper dictionary sales into steep decline everywhere in the world; change could not wait.

Oxford University Press already had some grounding in digital products, including licensing its content to electronics makers. One good example was the electronic dictionary it had made with Japan's Casio since the late 1990s, which spared users the work of turning pages and packed more than a dozen dictionaries, in different editions and for different uses, into a small box.

Still, he faced a plain fact: nobody on the dictionary's editorial team knew the first thing about building an app.

And the proposal met resistance on every side from the start. The readership for e-books was still unclear, and no publisher had much confidence in the young digital business. Paper dictionaries, meanwhile, had spent a century building a complete sales network, a marketing playbook, and a set of reference standards. Who would buy an app? How would you advertise it? Nobody knew.

"It was exactly this unfamiliarity that made every partner wary in the early days of the proposal," Liu said. "The only benchmark we had was the software at the top of the charts, so we downloaded them one by one and studied them: what was wrong with these apps, what could we do, what couldn't we do? Some of our editors bought their first smartphone because of this project."

Another thing that weighed on Liu: even now, nobody has settled on how dictionary software makes money.

Just before the development plan was signed off, his boss asked him two questions. What would a digital edition do to the print edition? And how much money would the digital edition make?

"I didn't overthink it," Liu recalled. "I said: first, don't treat a digital edition as a blow to print, because people use a digital edition differently. It is a fresh opportunity — to find new readers, and to bring new material to old ones. As for the second question, I can't tell you how much we'll make, but I can tell you the most we can lose, which is what we spend on development."

Liu thought the answer looked less than perfect at the time, but at least his proposal was not dismissed out of hand. Better still, the Oxford Advanced Learner's app earned back its development costs within two months of launch, and during one discount promotion it reached No. 7 on the overall app sales chart of Hong Kong's Apple App Store.

Liu puts the success down to "doing what the times called for," and to the Oxford dictionaries' long-standing reputation, which won recommendations from many schools and academic publishers.

The Oxford Advanced Learner's English-Chinese Dictionary app is not the same thing as the online dictionary software everyone uses today, even though both draw on the corpora of professional dictionary publishers. Take Youdao Dictionary, owned by NetEase: its word bank includes the Collins English-Chinese Dictionary, the 21st Century English-Chinese Dictionary, even material from Wikipedia. Look up a word and you get competing explanations from all of them, plus English example sentences pulled in by search engines.

By comparison, a corpus built by a traditional dictionary publisher offers a narrower range of definitions.

For one thing, everything in the app comes from the Oxford English Dictionary's corpus. The way that corpus is tagged decides what the app can look up and how much it can say. Editors and engineers preparing the content are, to a large extent, bound by the tags on each piece of information.

Set against the database Simpson had just finished digitizing in the 1990s, the corpus Liu keeps returning to is now an enormous thing, refined by editors and engineers over decades and holding billions of words. It draws on many kinds of media: newspapers, magazines, literature, radio recordings, stage dialogue. The provenance of every item is recorded — when, where, from what. Editors read the corpus to write entries, deciding whether to admit a word by how often it is used, when it first appeared, how widely it has travelled, and then working out and writing its senses. The illustrative quotations come from the corpus too, which guarantees they are records of real use. Editors tend to prefer material that is tightly built and grammatically telling.

Dictionaries, electronic dictionaries, licensed content: all of them are the front end of this corpus. Once it was Murray's little scriptorium; now it is what lets you stand in a foreign restaurant and work out the menu in a few seconds.

The corpus also supports work beyond dictionary-making in the old sense. Oxford University Press licenses its content widely, including partnerships with technology companies such as Apple, Google, and Tencent.

"Beyond that, we're building an academic project to cover other languages," Liu said at the end. "We've been talking about Chinese and English dictionaries, but Oxford has dictionary projects in other countries and regions too. We're also planning a large multilingual database. In future it will produce not only dictionaries, electronic dictionaries, and translation software — it will also be part of a language preservation program at Oxford University Press. Our goal is to take in a hundred different languages."

Cover image: Examining the OED