“For ten years I worked in the music business, doing concerts around the world. I was responsible for getting 40 people, band and crew, to the right specific entrance of a stadium, or maybe a villa which was halfway up a mountain in Italy.” Chris Sheldrick, cocreator and CEO of what3words is explaining the problems he faced in his former life as an events manager. This prompted him to imagine a new geolocation system that he hopes will transform how addresses are found around the world. “There’s one address for Wembley Stadium, even though there are 42 entrances, car parks, visiting points and so on. Inevitably, people got lost every day trying to find my events, which was annoying.”
Sheldrick tried to encourage people to use GPS latitude and longitude coordinates for satellite navigation, but it was apparently difficult to remember eight-figure numbers for each direction. “It’s very error-prone if you have to type in, one-by-one, 16 digits,” says Sheldrick. He came up with a solution by “exploiting some pretty convenient math.” If you divide the world into 57 trillion squares of three meters by three meters, there are enough words to be able to apply unique three-word sequences to each (which on the back end would correspond to latitude and longitude coordinates). Thus was born what3words, a global address system where any location could be pinpointed by using a three-word identifier. I met Sheldrick at ///index.home.raft — the what3words designation for his office space in Westbourne Park, London — to talk about the app’s uses and localization relevance.
How it works
Upon being downloaded, the what3words app can be used to find your address or location through self-entry or else via a “locate me” button — the latter option allowing you to pinpoint precisely where you wish to receive mail or be picked up on a satellite-view grid. The three-word address is displayed on a bar at the bottom of the screen — it can then be written down or memorized with much more ease than several-digit numeric coordinates. The Oval Office at the White House is represented by ///metals.deeper.hits, while MultiLingual’s headquarters in Sandpoint, Idaho, is ///punchy.nights.bond.
Beyond mobile apps, there are other tools through which what3words can be used, such as a map tool for easy discovery and sharing of addresses, an online API code through which addresses can be embedded to a website or other apps (accessible offline on a smartphone using Mobile SDK) and assorted plugins. The various pieces of software are aimed at everyone from householders and companies to governments and non-governmental organizations. For example, German automotive corporation Daimler AG has recently announced that what3words will be integrated into Mercedes Benz’s in-car navigation system.
The success of this system obviously relies upon its adoption by companies and organizations with whom the user can interact. Postal and delivery services are a key area of operations, for obvious reasons, and here there has been a strong uptake — particularly by postal services in less-developed countries around the world. The national postal service on the Dutch Caribbean territory of Sint Maarten, for example, has partnered with what3words to provide addresses everywhere there, including previously unaddressed homes and new developments. Tonga has done the same. The government of the Solomon Islands adopted the system in July 2017. The service has been used similarly by Mongol Post in Mongolia, by La Poste Cote d’Ivoire, the national post service in Djibouti, and by Carteiro Amigo, a local cooperative based in Rio’s largest favela Rocinha. In August, Nigeria’s NIPOST announced its adoption of the app.
Delivery companies in more advanced economies, such as the UK’s Direct Today, have also started using the app to improve services in rural areas (reducing first time failed deliveries by 83%), while companies in Ireland (where there was no postcode system until very recently) are also embracing the service. Even in major cities like London, companies such as Quiqup have reported a marked improvement in delivery time and accuracy using what3words, along with fewer failed deliveries — the last portion of the delivery route often being the difficult part (not knowing which precise door to use, or which office in a large building, for example). Onibag, which delivers emergency packages to hospitals and clinics, is now using the service to save precious time in their rapid response role. The technology is already being incorporated into automated delivery and transport systems, such as IBM Watson’s self-driving vehicle Olli, and Drone Scout.
Beyond being used by postal services and delivery companies, what3words has been integrated into the Navmii app for traffic and navigation (used by 26 million people). Festival Medical Services uses the app to be able to identify the exact location of an emergency at huge outdoor events such as the Glastonbury Festival, where hundreds of thousands of people are spread out across a massive area. “There are so many ways that people use us around the world, from describing where a photograph is taken with a three-word address, to people using us to coordinate where they go for a run in the Himalayas,” says Sheldrick.
Most impressively, perhaps, the United Nations has now adopted what3words for use in disaster recovery and relief programs. This is particularly useful when trying to identify the location of an emergency in a remote or poorly-addressed region. The UN estimates that three-quarters of countries have no systematic addressing system and four billion people are without a formal address.
The fact that the system can be accessed offline in a file of only 12 megabytes means that it can be used in rural and poorly-connected regions with little or no cell phone or internet coverage. GPS coverage is largely accessible around the world, so phone location identification is still possible in even the toughest terrain. Unlike commercial operations such as postal services, charities and emergency organizations are charged a much-reduced rate for embedding what3words code on their software. UN ASIGN, for example, is a crowdsourcing app for immediate post-disaster intelligence. “It was pretty humbling when the UN included us in their disaster relief app,” says Sheldrick “and the Red Cross in the Philippines used us in the last disaster there,” during Typhoon Haima in 2016. In non-emergency situations, the technology has proved useful in a program to lay pipes in earthquake-affected Haiti, for example. There was also a Red Cross mission to identify precise contaminated water sources of cholera outbreaks in Tanzania. 2017 saw the company win a “UPS International Relief and Resilience” award for its contribution to such programs.
Localization by design
Perhaps the most obvious way in which what3words has localization relevance is in the way it rectifies the problem of poor or nonexistent addressing. For example, it can be difficult for medical services in remote parts of Africa to reach patients for this very reason. One proposed solution was using local chiefs’ names to identify the relevant village, but what3words goes much further since every three-by-three meter square on Earth is identified with a unique three-word address. In terms of providing services to the large areas of the world hitherto unidentified officially, the service has great potential.
The service is a global one, and, as such, the question of language has been central to the app’s development. Any one of the 57 trillion unique location identifiers can be found in a variety of different languages — at present 14, with more on the way. These are not merely translations of the English designations into other languages, but are completely different combinations of three words in each different language used. Sheldrick gives the example of pound in English, which in French would translate as livre, which could then be translated back into English as book. Clearly, real-life language makes one-to-one direct translation impossible. However, the service can be used anywhere in any available language — so, for example, the French version can be used in the US to identify a location in the US or anywhere else using French words.
“In each language, we review a list of words,” Sheldrick explains. “We need 25,000 words in each of those languages. We go through word-by-word and we decide if the word is good enough to be included. We get a team of linguists to go through these words with us, attributing scores to words based on how easy the word is, how common it is, and that’s how we put together a final word list.” In English, homophones are eliminated — hear and here for example — to remove any identification mix-ups when the three-word combination is given orally, and to avoid the need to spell words out. “In French, however, it’s a whole different ball game. Pretty much all words are homophones due to how the tenses work. So, we made a rule for French that we would only use the infinitives of words and then a couple of other particular endings. But we have had to reconfigure it for languages like French which are very homophone-rich.”
Other languages have required decisions on types of words to be used because of how long the word endings are. “We take a lot of active advice from language consultants,” says Sheldrick. “Then we refine our master-word list with exact forms of the word that we are going to use, and then we set about getting the manual reviewers to go through them.”
Common sense was also used to decide which words were applied where in each language. “In the Russian version, we put the shortest, easiest Russian words in Russia, because that’s where it’s most likely to be used. In the French version, we put the most easy words in France, but very long or obscure words might be used in a forest in northern Russia because there aren’t many intrepid French travelers there. So generally, if you’re in London you might find ///table.chair.spoon, but if you’re visiting the Antarctic you might find ///ecologically.dodecahedron.subconsciously.”
Similar words are spaced quite far apart geographically. “///index.home.raft is our office here,” explains Sheldrick. However, “if you put an s on the end of raft, then ///index.home.rafts will take you to Western Australia.” The AutoSuggest feature takes this into account, so if you’re in London and type in something in Australia, AutoSuggest will correct you and say, “I think you meant this.” This allows for typographical errors or laziness without sending the user the wrong way, but it also means that people with literacy problems have less of a problem availing of the service. In places where illiteracy can be common, such as rural parts of Africa or India, for example, this could be a crucial feature.
Three linguists work permanently at the company’s head office in west London, coordinating freelance translators and linguists who work on the various repositories of words in each language. “If we are doing something in Arabic, we make sure to take people from across the Arabic-speaking region to get the nuances of whether a particular word is potentially offensive in one country but okay in another.” The service is aiming to have covered 25 languages soon, with the next batch in a series of Indian languages, Indonesian, Malaysian, Zulu, Xhosa, some East Asian languages and all of the Scandinavian languages.
Future plans for the app
Voice technology is the next step in the company’s vision for what3words — plans are in place to release it soon. This will be of interest to those who work in areas where illiteracy is common and for whom using a literacy-dependent app such as this would otherwise be problematic. In much the same way as typographical mistakes are identified and corrected, the voice version will pick up on mistakes of speech and ask if something more obvious was meant if that identifier seems more likely. As similar words are placed far apart geographically, the app will be correct in a large majority of cases.
“Voice recognition with addresses is normally pretty hard — you’ve got 14 Church Roads in London, and in Mexico City you’ve got 632 Juarez Streets. With ours you’ve got three words which are totally unique and the app will make sure you have an easy time using it by voice,” says Sheldrick.
Sheldrick sees this being of particular use while driving, especially with the advent of driverless vehicles. He talks of being able to get into a vehicle and say the three-word destination so that it takes you directly there. Nowadays, people put a postcode/zip code in and then often must drive around a little upon arriving to identify the exact place, but a driverless vehicle will not do that. “There’s a whole head-shift of ‘I’m going to have to start being very precise about location in a way that I haven’t before.’” The same principle applies to the use of drone delivery — street addresses will not necessarily be sufficient if, for example, it incorporates a front garden and a back garden, and the delivery needs to be somewhere precise (under cover from rain, for example). There is talk of an additional character, perhaps a number, indicating the added dimension of height.
However, for Sheldrick, the most important future goal is to make sure that people will be able to use the app in their own languages. “That requires a lot more work from our languages teams,” he tells me, “to take on some of the more difficult, complex languages for which even getting a starting dictionary is a bit of a mission — and then to be able to find 25,000 words that can actually be used with our app. We want to be able to be truly universal.”