Lost in translation: do interpreter apps work?
Lost in translation: do interpreter appswork?By Kevin RawlinsonBBC News11 February 2015 Fromthe section TechnologyBilbaoThe BBC road-tested a host of translationapps in the Spanish Basque city BilbaoThey are going to remove language barriersforever. Or so it has been claimed.Google recently released a real-timetranslator, while Microsoft-owned Skype is beta-testing its own.Both firms make grand claims about theirservices, while other smaller rivals offer alternatives, but the questionremains: do they actually work?I took a selection of apps to Bilbao in theSpanish Basque Country to find out, testing them by completing tasks set for meby colleagues on the BBC's tech desk.The first job was to find the northerncity's Guggenheim museum and ask what was its most valuable work of art.Getting to the museum was not a challenge;another Google app saw to that. The problem was getting my question across.lineBilbao to-do listBasque flagGet to Guggenheim museum and find out whatis its most highly insured work of artGet to Moyua metro station and ask someonethere the best way to get to Plaza NuevaIn Plaza Nueva find someone to tell thestory of their first kissFind Gili-Gili and ask someone inside totake a selfie posing with you and an item sold thereCatch a taxi to Cafeteria Concha and, whenyou arrive, ask what their bestselling pintxo isGo to the city's bullfighting museum andask the staff inside how many people the bullring could hold when full and whenthe original was destroyedBuy a one euro stamp and postcard and sendit to the Tech teamlineGoogle Translate doesn't yet support spokenBasque, so I opted for its English-to-Spanish setting.Despite near-perfect conditions - indoorswith no background noise and a volunteer who was familiar with the concept ofthe real-time translators - it initially struggled to convey relatively simplephrases."OK, so I've arrived at the Guggenheimand I'm here with Begoña and I have to ask her a question, according to my listof things to do," became "ok so the rise of the guggenheim and andand i have got acid? On my list of things to do."While it got it the second time round, theanswer that came back was not exactly perfect:The piece of art mask museum's owncollection is a Rothko painting called UntitledIt was, at least, comprehensible. Aninauspicious start, nevertheless.First kissAt nearby Moyua metro station, the app hadto face an even stiffer test - one that exposed what is perhaps its biggestweaknesses: background noise and the rigours of real life.The weather was crisp but dry when Imanaged to stop a woman at the building's entrance to ask for the quickestroute to the old town's main square.By the time she had grasped the concept oflistening for a translation into Spanish and answering the question clearly andslowly - delivering the directions bit-by-bit and waiting for the app to catchup - a sudden hailstorm had struck, shaking her resolve to persevere withstruggling software and British tourist alike.Nevertheless, I eventually found PlazaNueva.There sat 22-year-old Yurena, who agreed toshare the story of her first kiss with me.Apparently, it was at a party in Bilbao.She cuddled up with Miguel on the sofa and then "muy bien". Thecouple are still together, she said.A romantic story, without doubt.Unfortunately, none of it told using Google's real-time translator, which couldnot deal with the fact that she was speaking at the speed of a normalconversation.BilbaoYurena dispensed with the realtimetranslator, which couldn't keep up with herIn the end, Yurena had to give up and typeher message manually, pressing a button to get it translated into text for meto read.Her task was made no easier when GoogleTranslate turned "can you speak slower, please" into "can youspeak Spanish Big Show".Una selfie, por favorVocreVocre supports 38 languages, but is notfree like some rival appsVocre is one of the most prominent otherapps to provide real-time translations. And, at Bilbao's bullfighting museum,it achieved a similar rate of success to Google's.Converting between English and Spanish, itmanaged to convey a question about the bullring's capacity relatively well.The answer that came back from staff memberJoaquín Vega was clearly mangled: "Around 14 thousandth people."Imperfect, but at least it wasunderstandable.Such apps performed best with measuredspeech and low background noise, as illustrated by my experience in Gili-Gili -a local clothes shop.Inside, where it was relatively quiet,Google Translate managed to convey my request for a posed selfie almostperfectly.After some explanation of how the systemworked, Pedro Aramaio - who spoke little or no English - agreed to pose for thephoto, ticking another task off my list.That was, however, a brief highlight.BilbaoPedro Aramaio agreed to pose for a selfieafter conversing using Google's appAt nearby Cafeteria Concha, which wassignificantly louder, the same app had trouble picking out any speech at all.The barman there professed to speak verylittle English.But when the software took an age to detectmy request to identify his bestselling pintxo - a word for Basque bar snacks -he opted to attempt an answer in my language, rather than to persevere with thesmartphone.BilbaoThat's definitely not what I said'Awkward' appsIf real-time translation apps can get itright, they could upend a lucrative sector.According to a recent report by theEconomist newspaper - which cited consulting firm Common Sense Advisory - thelanguage interpretation industry generates about $37bn (£24bn) worth of salesevery year.But the problems I experienced in Bilbaosuggest that processor-powered translations still have far to go.Those issues are indicative of speechrecognition tech's limitations in general, according to Joseba Abaitua, anacademic at the modern foreign languages department at Bilbao's University ofDeusto.Mr Abaitua, who specialises in onlinecommunication, suggests that interacting via a smartphone while face-to-facewith someone else is always going to be "awkward". But, he adds, appswill become more effective as people get used to them.Jump media playerMedia player helpOut ofmedia player. Press enter to return or tab to continue.Media captionTesting Skype's real-time translator"In a way, you have to… make acompromise, you have to know who are you talking to - you are talking to aspeech recogniser, a machine," he says."So, the machine can startunderstanding you quite well and, if your sentences are short and wellrecognised, the translation system… may make a good job."But, if you start talking unexpectedthings with a lot of colloquialism, then the whole system breaks down."He adds that systems need to get muchbetter at recognising people's different accents and ways of speaking.BilbaoNone of the apps had Basque, but they didhave SpanishSmartphone-based translators also facetechnical limitations, such as their reliance on internet connections andlimited battery life.And, in some cases, they simply promisemore than they can deliver.One app I tried called Interpretersteadfastly refused to live up to its name.When I attempted to use it to find apostcard shop it failed to translate either my questions or the replies I wasgiven, despite near-ideal conditions.A helpful shopkeeper ultimately had toregress to pen and paper to draw a crude map of the area.Eventually, I bought a stamp and card tosend back to London. But not before Google Translate managed to maul this lastrequest."I speak English," the bemusedcard salesman said, finally putting me out of my tech-induced misery.Share this story About
The translation industry
Technology may not replace humantranslators, but it will help them work betterTimekeeper TALK into your phone in any of the bigEuropean languages and a Google app can now turn your words into a foreignlanguage, either in text form or as an electronic voice. Skype, aninternet-telephony service, said recently that it would offer much the same (inEnglish and Spanish only). But claims that such technological marvels willspell the end of old-fashioned translation businesses are premature. Software can give the gist of a foreigntongue, but for business use (if executives are sensible), rough is not enough.And polyglot programs are a pinprick in a vast industry. The business oftranslation, interpreting and software localisation (revising websites, appsand the like for use in a foreign language) generates revenues of $37 billion ayear, reckons Common Sense Advisory (CSA), a consulting firm. ADVERTISEMENT In this sectionCrudely putPitfalls at PetrobrasLove on the rocksJust the ticketSay what?The last 90 daysReprintsThe market is growing, and widening.Translation in continental Europe was once dominated by the “FIGS” (French,Italian, German and Spanish); Japanese, Chinese and Korean were the only Asianlanguages to speak of. Roughly 90% of online spending is accounted for byspeakers of 13 languages, says Don DePalma of CSA. But others are becoming moreimportant, for reasons of both politics and commerce. The European Union’s bureaucrats now haveto communicate in 24 tongues. In Asia once-neglected languages such asVietnamese and Indonesian matter more as those countries grow. Companies activein Africa regard that continent’s languages as increasingly important. Bigsoftware firms like Microsoft find it profitable to localise their wares insmall languages like Maya or Luxembourgish. Translation is no longer usually toor from English. Technology, far from replacing humans, isinstead a tool that helps them keep up with surging demand for high-qualitytranslation. “Translation memory” (TM) was the first big useful tool. Since the1980s translators have been able to dip into vast TM databases containing wholesentences that have already been translated in a given language pair, helpingthem to speed up repetitive work, such as translating instruction manuals. “Machinetranslation” is the next step. Computers learn from huge databases ofalready-translated text to make ever-better guesses about how to render wholechunks from one language into another. Translators used to scorn this, seeing theirhuman judgment as irreplaceable. Now, says Jiri Stejskal of the AmericanTranslators’ Association, it has won respectability. Technological change has not broughtconsolidation to a fragmented industry, however. Lionbridge, which has thelargest disclosed revenues ($489m in 2013), makes much of its money fromservices other than translation. Like most of its rivals, Lionbridge talks uptechnology, but is fairly traditional. The heart of the business is managingprojects, acting as a go-between for customers and freelance translators onjobs like managing file formats and locations, client reviews and so forth. Tedious project-management tasks like thesemay offer scope for disruptive innovation—perhaps from the translation world’sequivalent of Uber, a taxi app. Software is unlikely to replace thetranslators, but it could co-ordinate their work with clients more efficiently.Smartling, an American company which seeks to cut out middlemen in this way,has clients including Tesla, an electric carmaker, and Spotify, amusic-streaming service. Jochen Hummel, a pioneer in translationmemory, says that a real breakthrough would come from combining software,memory and content management in a single database. But making money may stillbe tricky. The American tech titan has not tried to commercialise GoogleTranslate. A former executive says the firm experimented withcontent-management software but “decided to focus on easier stuff, likeself-driving cars.”axi apl�otȐ�8��nlikely to replace the translators, but it could co-ordinate their work with clients more efficiently. Smartling, an American company which seeks to cut out middlemen in this way, has clients including Tesla, an electric carmaker, and Spotify, a music-streaming service.Jochen Hummel, a pioneer in translation memory, says that a real breakthrough would come from combining software, memory and content management in a single database. But making money may still be tricky. The American tech titan has not tried to commercialise Google Translate. A former executive says the firm experimented with content-management software but “decided to focus on easier stuff, like self-driving cars.”