Join almost 1 million
monthly
readers!

Contact Me

Coaching and Consultation

Language Hacking League!

Join over 50,000 people to get FREE weekly language hacking tips, cool links, site updates & two free chapters of the Language Hacking Guide!

No Spam. Not ever.

Current Mission:


Meeting up with readers face to face to encourage YOU to learn languages!

Previous post:

Next post:

Why your smartphone will NEVER be a universal translator

| 38 comments | Category: tools

Anyone who follows sci-fi will have come across the “universal translator“. To get around use of other languages on the shows that aliens would undoubtedly speak, (other than when part of a plot device, like Klingon or Na’vi), a computer does all the work of understanding other languages so the human explorers don’t have to.

Occasionally it’s something else, like the Hitch Hiker’s Guide to the Galaxy’s babelfish, which relies on telepathy.

Either that or everyone in the galaxy speaks English, which is just a higher level version of the delusion that English is all you need on earth.

Translator apps; created by lazy monolinguals for gullible monolinguals

The thing is, we are at the stage when our technology is sophisticated enough that there are actually apps you can install on your smartphones that almost seem to act like a universal translator!

Every day I get sent interesting links by email, or they are shared on the Fi3M forum resources page, and inevitably I see links to a couple of iPhones apps that make it seem like you don’t even need to learn the language any more – just hold your phone up and all problems will be solved.

If you have no idea how languages work, or if you are gullible enough to believe the ideal conditions of the advertisement, then you will be absolutely convinced when you see the two videos below that learning a language is now a waste of time; soon enough you will be able to stroll around in any country shoving your iPhone in people’s faces and at signs and communicate without any issues(!)

But the thing is, these videos are only impressive if you don’t understand what is actually required for communication and how restrictive they are and how far away we are from something useful. They are actually terribly inefficient and way off anything practical.

And those who designed the apps? They seem to have no understanding or appreciation of how languages work. They are programmers, who didn’t even consult natives to check if their examples were accurate, and who are blissfully unaware or unwilling to admit the many faults their app has in their commercials.

Let’s have a look at the two most famous ones I’ve seen:

Word Lens: Get your signs translated… into gibberish

This first one is something I’ve seen for a long time online. If you don’t read any Spanish it looks amazing! But the thing is, the whole ad is designed specifically to mislead you in how the app works.

The most impressive lie I see in the video is the fake sign they designed which says “Lo traduce el texto instantáneamente”, which isn’t Spanish at all! This is gibberish that you will never see in the real world; this non-Spanish was created specifically to misrepresent how good the app is.

The app works by translating word for word what it sees. Some people may feel that it’s “good enough” when they are monolingual and don’t get how languages work. I was a professional translator for several years and I can tell you that even for very basic sentences, translating word for word produces nothing but gibberish in far too many cases – even more so for languages further away from English than Spanish would be.

There are too many situations where a word has several translations. Did you know that “set” in English has 464 meanings in the Oxford English dictionary, depending on the context? And “run” has 396? You can’t just pick one, even if it comes up as a likely translation a third of the time.

“Lo” in Spanish, so overused in the video, technically can mean “it”, but it actually is way more likely going to mean “that which is”, “what”, “him”, “you”, and actually in the vast majority of cases it isn’t translated or requires serious rephrasing in the English equivalent sentence. No dictionary can portray this efficiently; only by real use do you finally understand what “lo” can mean.

There are some cases where you can indeed “get the gist”, but it requires one hell of a good imagination, especially when the word order is totally off and basic words with many translations are misused.

The app author admits to not consulting any Spanish speakers in designing the app or creating the video. This is nothing but sloppiness.

As well as these issues, it doesn’t work with other fonts, it wouldn’t handle damages to particular letters or a non-plane background around the letters, and it relies on an offline limited dictionary that is clearly really poor. The only thing I can praise in this app is the clever programming for replacing the letters in a moving image. That’s pretty impressive coding, but don’t let that fool you as to how useful it can be.

It could do OK with some signs, especially if the dictionary database is designed specifically for signs (which it isn’t, as it focuses on one word at a time rather than the phrase). I would not recommend you use it for anything but road signs, and have a real dictionary or phrasebook handy for the majority of cases when you don’t have a clue what it’s trying to tell you. Steer clear of menus, newspaper articles and the like.

In summary: don’t download it!

Vocre – the very expensive “future” of misleading instant voice translation

Here’s the latest one – an app that instantly translates spoken conversations! Once again a video portrays it as being as good as magic. On their site, they say:

Vocre [voh-krey] is a new translation app from myLanguage that allows anyone to communicate instantly with anybody from anywhere – without language being a barrier… Our world is full of people worth talking to. Now, with Vocre, nothing’s holding you back.

Burn your language learning books – the future is here and learning other languages is a waste of time it seems!

Perhaps not. Here are a few things the advertisement fails to tell you:

  • The app is free with some test translations, and then it costs a dollar per ten uses. So forget about that girl continuing an entire date using the app!
  • The interface is cumbersome according to many of the reviews – you waste your paid credits if you hesitate while rotating
  • The Google Translate app does the exact same thing for free if you don’t mind clicking some buttons.
  • That guy in the video ISN’T A NATIVE FRENCH SPEAKER! It’s pretty obvious he’s just an English speaker with good (not excellent) French. “Acheter un café” (or as he says it “ach…et…er”), literally “buy a coffee/café” sounds more like purchasing the establishment of a café – that’s not how you invite someone out in France (although it would work in Quebec, which I seriously doubt they were aiming for). This phrasing is backwards translation from English intentionally to make the video more impressive. Once again I ask if they tested this app on or consulted anyone but English speakers? And what does that really demonstrate, that the voice recognition works on non-natives?
  • I highly doubt the advertisement is actually even using the app. It’s pre-recorded audio, and the timing is suspect based on how the app is supposed to be used. If true, the ad is once again extremely misleading

If you speak slowly into it in standard formal dialect without much of an accent, it may indeed work well enough to recognise what you said and its translation may be good enough to get the gist (albeit with the exact same translation problems I’ve listed above for the previous app), but then advertise it as such.

[Edit: I've been sent more info that shows that in other demonstrations, Vocre have intentionally had natives speak unnaturally to misrepresent how good their app is. In one a Chinese phrase is reworded into Chinglish, and a basic number is translated wrong. If you don't speak these languages you are impressed, but I see it as deceitful the way they are demonstrated.]

Chatting up a girl is something that would usually involve informal register and tone, which is a lot harder to translate, and even more difficult for voice recognition to pick up. For example, I find it unlikely that many voice recognition apps would accurately pick up someone informally saying a simple French phrase like “I don’t know”. Rather than “Je… ne… sais… pas”, it would come out as “Ché pas” as described in this video (others here).

I’ve even praised the Google Translate app myself – and I always glance at a dictionary app to help me when I’m starting out in a language. But that’s the point; these tools can help us ease in while we are learning the language. I use Google Translate quite a lot in the early stages of all language learning. But suggesting that they can replace the learning process at this point in time is ludicrous.

We have a long way (decades or centuries) to go before you can rely on technology before needing to learn the language

I still recommend that people install the Google Translate app. As faulty as its translations are, I like the work Google does to try and make it produce the best it can based on context analysing (rather than dictionary translations), and especially that this tool is free. The two apps above and many others deserve some serious criticism, because they are both commercial products and their advertisements are terribly misleading.

Looking through the Youtube comments and reading the messages I get when people send me these links, I see that some people think that learning the language will very soon become obsolete thanks to this technology. As if!

Among many other issues, relying on such technology ignores how crucially important non-verbal communication is to understanding what you are being told. An app that relies entirely on audio, no matter how sophisticated it may be, and no matter what processing power the hardware has or online database it accesses, will never take this into account and always fall short of being able to translate fully.

Something I like to remind people of is that I’ve got a degree in electronic engineering. My speciality was in (digital) communications, and I worked in Paris for nine months as pre-sales technical support (for English clients) at a company that developed the voice recognition system used by Les Pages Jaunes (the French yellow pages). I am well aware of many aspects of how these systems technically work (as would the programmers), but I’m also aware of how languages and translation work (which apparently not a single iPhone app developer involved in these would).

Mark my words: Based on my experience in both fields I can tell you right now that no app will ever make learning a language obsolete.

At best, these devices may some day (definitely not now) allow you to have a superficial and basic touristy Q&A in non-noisy environments without needing to learn the language, and may indeed aid short-term tourism. But if you want anything deeper than the superficial life of an English speaking tourist and all you miss out on because of that, you will not get it from technology – at least not for the many decades.

If these zealous devs are to make a real difference in this time and create something that works beyond marketing and gimmicky and false videos, they need to consult experienced and successful language learners, foreign language natives and second language acquisition and translation linguists, and of course professional translators.

And when they do, they would do well to present its limitations honestly in their commercials.

————

As you can see, this stuff really annoys me – mostly because it encourages laziness among monolinguals.

Learning a language is extremely rewarding and very much accessible to everybody regardless of age and wealth. Despite the great advances in technology we’ve made, the universal translator (and even a fully functional single language translator) will remain science fiction for a very long time.

Any thoughts? Let me know in the comments!

***********************

Enter your email in the top right of the site to subscribe to the Language Hacking League e-mail list for way more tips sent directly to your inbox!

If you enjoyed this post, you will love my TEDx talk! You can get much better details of how I recommend learning a language if you watch it here.

This article was written by

Comments: If you liked this post or have anything to say, please leave a comment! I love reading them :)
Just keep in mind that I’ll delete any rude, trolling, spammy, irrelevant or way off-topic comments. Also, use your REAL name, not a brand or business one, and don’t link to your site in the comments unless it’s relevant to this post.
If you have a general language learning question, please ask it in the forums. Otherwise please use the search tool on the right for any other question not related to this post.

———————————–

  • http://niel.delarouviere.com NielDLR

    I wholeheartedly agree. When my friends look at these apps, they praise their “amazing” abilites, but often I try my best to bring back down to Earth. I remember that WordLens video. It did quite the rounds. I immediately saw the flaw in it. The technology is more impressive than the “proposed” use of it.

    Although I’m not a computational linguist (just a general/applied one), I’ve got some friends in my lab that I work that do research in computational linguistics. Numerous times we talked about machine translation. The conclusions time and time again: it’s almost impossible. It will always fail and be awkward.

    For other less researched languages, like Chinese, the translation between that and English sucks. Really is fails terribly, even Google Translate sucks at Chinese translation.

    For me smartphones serve an EXCELLENT complimentary use for language learners. My dictionary app is priceless. Talk to a native. Get stuck on word. Look up word. Continue.

    Furthermore, I understand why some people would love these language barrier breaking apps, but goddammit, there is nothing as rewarding as being able to communicate in another language. Even if these apps become available in the future, I’d never use them. It just doesn’t feel right. Will those that actually take the effort to learn languages become even more rare? I’d loathe to see such a dystopia.

    • http://www.fluentin3months.com/ Benny Lewis

      Such apps, if they were to exist, would be a lazy traveller’s dream come true. Those people can perhaps prepare their grandchildren for such a possibility under limited conditions, but in the next decades we will not come anything close.

  • http://pulse.yahoo.com/_LL7ZAGFTMXVYD6N5RYB342ZT4I Jeffrey C

    But… we already have universal translators that work with 80% accuracy, just not for consumers yet.

    • http://www.fluentin3months.com/ Benny Lewis

      1. Who the hell are “we”? I’ve never heard of this secret society…

      2. 80% accuracy on automatic translators is a ridiculous claim. Only people who don’t know how translation works would make up such figures.

      • Chris

        80% accuracy is not a ridiculous claim at all just depends how you measure it and what you are measuring.

        Of course 80% might not end up being particularly useful ;).

        Just like those idiots (sorry maths – challenged) who misrepresent pareto 80:20 rule as an excuse for being lazy and mediocre (forgetting to think that the rule if it applies where they think it applies states nothing about where the 20% comes in the process).  For example a language learner can learn 80% of the words (by frequency) of a text quickly and still have not a clue about it actually means.

        80% is not ridulous at all, especially when you consider that most of the ways I can imagine to think about you still end up with something pretty useless.

        • http://www.fluentin3months.com/ Benny Lewis

          Well then if you measure it as “individual term translation is understandable, ignoring context” then I’d agree that 80% is achievable. But I challenge you to compare an automatic translation and a professional translator translation of a complex technical text or a casual conversation with slang and you’d be lucky if 20% of it was a fit.

          It’s either a ridiculous claim or ridiculous way of measuring 80%.

  • http://www.fluentin3months.com/ Benny Lewis

    Good for superficial tourists PERHAPS in two to five years. Way off it now.

  • http://www.fluentin3months.com/ Benny Lewis

    Good for superficial tourists PERHAPS in two to five years. Way off it now.

  • http://www.fluentin3months.com/ Benny Lewis

    I agree. As I said I use plenty of apps myself – they have their place in aiding language learning, but never replacing it. It’s beyond laziness, cheating and unfair etc. – they are simply not good enough and never will be. It’s hard for monolinguals to appreciate that when they don’t understand the complexities of language learning, and feel it’s all a simple case of direct translation.

  • Eva Snyder

    My father was a Professor at one of the four college translation labs in the US .  He told me there will never be good machine translation because translation is the only time you have to know what someone means.  Meaning is always dependent on context.  You can write a correct sentence in your native language without knowing what it means.  You can read it without knowing what it means.  But you can’t translate it without knowing what it means.  He told me a story about a translator who was working on a text and couldn’t figure out a particular  sentence so he went to the author and asked him what it meant.  The author looked at the sentence and said “I have no idea what I meant by that”. 

    You are totally correct universal translators are just shortcuts in Sci-Fi so every first contact story won’t be taken up with overcoming communication barriers, which would be very boring.   The Star Trek Next Generation episode “Darmok” unintentionally illustrates the problem.  If the Star Trek universal translator can’t handle idioms it can’t work on any language.   English has no word for “toilet” that isn’t a euphemism. “Toilet” itself is a euphemism. 

  • Eva Snyder

    My father was a Professor at one of the four college translation labs in the US .  He told me there will never be good machine translation because translation is the only time you have to know what someone means.  Meaning is always dependent on context.  You can write a correct sentence in your native language without knowing what it means.  You can read it without knowing what it means.  But you can’t translate it without knowing what it means.  He told me a story about a translator who was working on a text and couldn’t figure out a particular  sentence so he went to the author and asked him what it meant.  The author looked at the sentence and said “I have no idea what I meant by that”. 

    You are totally correct universal translators are just shortcuts in Sci-Fi so every first contact story won’t be taken up with overcoming communication barriers, which would be very boring.   The Star Trek Next Generation episode “Darmok” unintentionally illustrates the problem.  If the Star Trek universal translator can’t handle idioms it can’t work on any language.   English has no word for “toilet” that isn’t a euphemism. “Toilet” itself is a euphemism. 

  • Eva Snyder

    My father was a Professor at one of the four college translation labs in the US .  He told me there will never be good machine translation because translation is the only time you have to know what someone means.  Meaning is always dependent on context.  You can write a correct sentence in your native language without knowing what it means.  You can read it without knowing what it means.  But you can’t translate it without knowing what it means.  He told me a story about a translator who was working on a text and couldn’t figure out a particular  sentence so he went to the author and asked him what it meant.  The author looked at the sentence and said “I have no idea what I meant by that”. 

    You are totally correct universal translators are just shortcuts in Sci-Fi so every first contact story won’t be taken up with overcoming communication barriers, which would be very boring.   The Star Trek Next Generation episode “Darmok” unintentionally illustrates the problem.  If the Star Trek universal translator can’t handle idioms it can’t work on any language.   English has no word for “toilet” that isn’t a euphemism. “Toilet” itself is a euphemism. 

  • http://twitter.com/Robodl95 Robodl95

    I agree… but only kind of. I’m really not fond of exaggerations like this, as my favorite song by the Fray goes “never say never”. You, and no one else for that matter, can predict the future. I’m sure that one day technology will have evolved so that these translators are very good, it may take 200 years for all we know but that’s not never. Second of all, what’s with the sudden monolingual hate? You can not speak all languages on the planet! That’s what translator apps are there for!!  A lot of people go through Europe in a couple weeks, a month, or maybe even a year and you simply can’t learn every language in that amount of time. It’s also pretty insulting to call someone lazy just because they can’t speak another language, for all you know they could be finding the cure to Cancer or some other time consuming thing.

    • http://www.fluentin3months.com/ Benny Lewis

      It would be very fair enough to say that there would be a useful automatic translator 200 years from now, but such long term theory really doesn’t concern me. My point is that nobody reading this article now will be wise to wait until an actual useful full translator is available. They’ll be dead or with one foot in the grave before it happens, if ever.

      Also please read the article carefully. There is no monolingual hate! I am criticising app developers for being lazy and not involving people experienced in languages in developing and marketing the app. A translation app should by definition NOT be coded and promoted exclusively by monolinguals. It’s sloppy. And only monolinguals will fall for the promises because monolinguals are simply not aware of how complex languages are beyond very basic word to word translations.

      I don’t expect tourists to learn every language of every country they visit for a weekend. Please save your rant for another time.

      Your last sentence is just plane ridiculous. You’ve missed the point of who I was criticising entirely.

  • Anonymous

    I guess it all boils down to what your true goal is.  For me, I want the satisfaction of being able to think in a variety of languages, and thus be able to talk to people face to face without always falling back on English.  I don’t travel much, so unfortunately, English is a crutch I can always rely on, however it is much more rewarding to be able to say what you want to say in words the other person can easily understand.  I see language learning as a way of educating myself, and a universal translator is not going to make me any smarter.  If all you care about is being able to order a beer in Spain or hail a cab in France, then by all means, a translator is the way to go.  For me, I see languages as a way to better myself.  Some people teach themselves how to play a musical instrument, others become master chef’s.  I happen to like the idea of being able to speak in four or five languages.  If your goal is to know a language, a translator will only hurt you, but if that is not your intention, perhaps these things can help you out.  I’ll stick to old fashioned trial and error.  :)

    • http://www.fluentin3months.com/ Benny Lewis

      As I said in the article, I think a superficial tourist-Q&A-only translator could be more or less reliable within the next few years. And that’s fine – but people will always overestimate what it can do and that’s where the potential for laziness in people who WOULD learn the language comes in and is something I want to prevent.

    • http://www.fluentin3months.com/ Benny Lewis

      As I said in the article, I think a superficial tourist-Q&A-only translator could be more or less reliable within the next few years. And that’s fine – but people will always overestimate what it can do and that’s where the potential for laziness in people who WOULD learn the language comes in and is something I want to prevent.

  • http://www.fluentin3months.com/ Benny Lewis

    I really hope I don’t have to repeat this so much in comments. I was pretty clear about the fact that I am an electronic engineer. I studied in depth how and why Moore’s law works about *certain* technology evolving exponentially.

    As I said in the article, this is not an issue of processing power or access to data. We are definitely decades off the kind of artificial intelligence required to fully appreciate language in all its contexts. Extremely restricted automatic translators can and will be available soon, that’s all.

    I’m not saying “never”, I’m saying not for a very long time.

    The Google Translate tool on my website is because a lot of non native English speakers come here, and may want to get the gist of an article, but they will never fully understand it unless they turn off the tool and read it as it was written by a human being. And as I said in this article I use Google Goggles to HELP in learning a language, and in that case I embrace all technological advancements. Replacing language learning is way off though.

  • http://www.fluentin3months.com/ Benny Lewis

    I really hope I don’t have to repeat this so much in comments. I was pretty clear about the fact that I am an electronic engineer. I studied in depth how and why Moore’s law works about *certain* technology evolving exponentially.

    As I said in the article, this is not an issue of processing power or access to data. We are definitely decades off the kind of artificial intelligence required to fully appreciate language in all its contexts. Extremely restricted automatic translators can and will be available soon, that’s all.

    I’m not saying “never”, I’m saying not for a very long time.

    The Google Translate tool on my website is because a lot of non native English speakers come here, and may want to get the gist of an article, but they will never fully understand it unless they turn off the tool and read it as it was written by a human being. And as I said in this article I use Google Goggles to HELP in learning a language, and in that case I embrace all technological advancements. Replacing language learning is way off though.

  • http://www.fluentin3months.com/ Benny Lewis

    This is the kind of potential laziness I was talking about. You don’t need anything to be invented, you need to put the work into improving your French so that it becomes part of your skillset.

    And no, there will be NO reliable translator within the next few decades. Language is by definition a human affair – it is far too complex to reduce down into if, else clauses. There are too many ways context needs to be presented, which computers can’t possibly perceive unless an advanced AI at least 50 years off is used.

  • http://www.fluentin3months.com/ Benny Lewis

    This is the kind of potential laziness I was talking about. You don’t need anything to be invented, you need to put the work into improving your French so that it becomes part of your skillset.

    And no, there will be NO reliable translator within the next few decades. Language is by definition a human affair – it is far too complex to reduce down into if, else clauses. There are too many ways context needs to be presented, which computers can’t possibly perceive unless an advanced AI at least 50 years off is used.

  • http://yetanotherlanguage.blogspot.com/ Crno Srce

    I would agree that the smaller developers don’t have the resources to give it much of a shot, but tools like Google Translate and Microsoft’s translator show how far we’ve come in a relatively short space of time already. Ten years ago we wouldn’t have believed how good Google Translate could be (far from perfect though it is). In ten years time, you might be rethinking your conclusions. Perhaps…

    Having said that, I don’t think any technological advances will ever stop people from learning languages, no matter how good the “Universal Translator”. It could just mean a lot less work for professional translators :-) 

    • http://www.fluentin3months.com/ Benny Lewis

      Google Translate is great, but it does not surprise me in the slightest. Ten years ago I would have believed you if you described its current features to me based on Google’s resources and current processing power.

      Apparently I’ll have to remind people in each comment that I’m an electronic engineer and work with and understand the exponential advancement of technology on a very intimate and technical level. Knowing this makes me very confident that I will NOT change my mind in ten years.

      Automatic translators will get better at helping tourists ask for directions and the like. I’ll have wrinkles before they can help people have a natural conversation in any useful way outside of fake or fabricated advertisements in idealised situations.

      If you trust my language skills and you trust my graduate degree in understanding the software and hardware behind the technology most people take for granted, then trust me in saying that useful all encompassing automatic translations are science fiction and will remain so for a very long time.

  • http://kevinpost.wordpress.com/ Kevin Post

    Nothing says, “romantic” like going out to dinner and communicating through an emotionless phone application. 

  • http://kevinpost.wordpress.com/ Kevin Post

    Nothing says, “romantic” like going out to dinner and communicating through an emotionless phone application. 

  • http://kevinpost.wordpress.com/ Kevin Post

    Nothing says, “romantic” like going out to dinner and communicating through an emotionless phone application. 

  • http://kevinpost.wordpress.com/ Kevin Post

    If anything it shows that girl you’re trying to impress that you are a lazy s.o.b. 

  • http://kevinpost.wordpress.com/ Kevin Post

    If anything it shows that girl you’re trying to impress that you are a lazy s.o.b. 

  • http://kevinpost.wordpress.com/ Kevin Post

    If anything it shows that girl you’re trying to impress that you are a lazy s.o.b. 

  • Gilberto de la Garza

    Hola Benny soy de Mexico y me encanta tu blog, me inspiro para empezar a atacar el idioma noruego. Sin embargo tengo curiosidad de saber por que te enfocas tanto en lenguas europeas y no en asiaticas (chino y japones).

    • http://www.fluentin3months.com/ Benny Lewis

      Por favor no dejes comentarios que no tiene nada que ver con el articulo…
      Escribí sobre la elección de idiomas acá: http://www.fluentin3months.com/which-language y si lees el blog verás que sí que he aprendido idiomas no-europeos, sólo que hasta ahora no los mantengo.

  • Anonymous

    Spot on about Google – still the best and most reliable translation tool out there.

    Couldn’t agree more with your overall argument. Technology is, and won’t be for a long time, a replacement for learning and speaking a foreign language. Besides you’d miss out on all the fun of making mistakes, getting into confusion situations and then finally seeing how far you’ve come!

  • http://twitter.com/TristanKing65 Tristan King

    Hey Benny,
    Very interesting post.  I really think some people (mostly people who don’t speak / haven’t considered the intricacies of another language) severely undervalue the idea of translating from one language to another.

    In a previous role in a big company, I was working in globalising websites, including their translations, and was requested multiple times to use Google Translate to translate client-facing technical documents and manuals which were to be posted on the company’s website, instead of using real human translators.  They wanted to save money, but I had to do a lot of convincing (and plain ‘no’s) before they realised that the output would be worthless and nothing but nonsense.

    I don’t think the tools themselves are useless – I use Google Translate and other online tools sometimes – for ‘getting the gist’, or translating individual words, they’re great.  But for full translations, the value of a real human can’t be underestimated.

    Lastly, it’s *shocking* that the creators of the first app didn’t consult any native speakers.  The usage of “lo” so many times and “ropas” with an extra ‘s’ made me cringe.

    On the other side of the coin:  maybe, if people use one of these apps to hear a few phrases and show that at least they’re making an effort (albeit a very small one), perhaps they’ll be inspired to learn the language for real?

    Interesting post, and thanks for analysing these.  I don’t like the ‘shortcut’ myths, either.

  • http://www.fluentin3months.com/ Benny Lewis

    Once again I’ll say that I am not surprised at all, and that includes Siri. I’ve been talking to my phone for a year and a half before Apple decided to try it themselves. Siri is more complicated in its range of questions than anything on Android, but it’s very easy to replicate if you think about the processes logically.

    It’s answering algorithms and predictable questions. Too many of them are not understood. Algorithms can’t replace intelligence. Not now by a long shot.

    Technology is indeed advancing exponentially, but language is a means of communication that only the monolinguals think can be emulated with some if else clauses.

  • http://www.fluentin3months.com/ Benny Lewis

    Once again I’ll say that I am not surprised at all, and that includes Siri. I’ve been talking to my phone for a year and a half before Apple decided to try it themselves. Siri is more complicated in its range of questions than anything on Android, but it’s very easy to replicate if you think about the processes logically.

    It’s answering algorithms and predictable questions. Too many of them are not understood. Algorithms can’t replace intelligence. Not now by a long shot.

    Technology is indeed advancing exponentially, but language is a means of communication that only the monolinguals think can be emulated with some if else clauses.

  • Anonymous

    Hey Benny,

    I really like your blog and I think you have a lot of great tips for how a person can learn a new language. I have been learning Portuguese and you have  been a great inspiration! However, I think you are a little out of your element when discussing machine translation. 

    It should go without saying that word for word translations will never even be remotely close to a good translation. However,  you seem to think that the computerized approach is this, or  a long list of rules and cases. This  has been tried and failed decades ago. These days, the approaches use a class of algorithms known as machine learning. Without going into too much detail, these are almost purely statistical approaches. They pore over millions of examples and calculate lots of probabilities of what a text probably meant. It’s interesting you say that machine translation researchers should consult linguists and language learners – they did, and it never worked. Statistics ALWAYS won. The human language learning process is too complex for a machine to capture, but statistical approaches give the best approximations. There are a few linguistic concepts like Chomsky grammars and de-stemming words that are left over, but most of the linguistic approaches and language-learner approaches have been abandoned. This isn’t to say things analyzing the contexts of sentences aren’t important – they absolutely are – but the approach tends to favor math over linguistics. 

    I went to a talk by the Google Translate team. They pointed out that although the algorithms have improved somewhat, the big wins in machine translation have been in computing power and data. By data, I mean bodies of texts written in both languages. Basically, even though computers aren’t very smart, when they can process terabytes of examples in seconds, they can make a pretty reasonable guess as to what a given text is supposed to mean. 

    I think you vastly underestimate how much machine translation is going to improve. This is because we are going to continue to get more data, faster computers, and to a lesser extent, smarter algorithms. A lot of smart people are working on this problem – don’t be misled because some clueless people are working on it too.

    However, I think we can both reach consensus that machine translation will never be as good as a human. This is simply because a machine can never have the human experience, which makes it virtually impossibly to get a joke, pick up innuendo in flirting, or understand a metaphor. Good  human translators can do these things and make appropriate translations (although this is difficult for humans too of course!). 

    We definitely both agree there is a ceiling to how good machines will get, I’m just pretty sure it’s higher than you’re imagining. It may never be as good as a human, but for straightforward conversations and texts expect things like Google Translate to get really, really good. MUCH better than “basic touristy Q&A”. Doubting the ability for humans to make computers do amazing things has been a source of many bad predictions throughout the 20th century – don’t be another victim of it :)

    Thanks again for the great blog,

    Bill

  • Ben E.

    How to you translate, for example, the Mool Mantra (a widely used Sikh religious song) into English? You have to have not only an understanding of Punjabi, but of the religious and cultural underlay of the words, the multiple meanings intended by the words, and then engage in a human response with how to relate to the meanings intended by the words. Even a perfectly grammatically, syntactically, phonetically, etc. sound translation cannot engage in that human process of meaning translation. Until machines can think as deeply as a human, there can be no universal translator.

    And that, my friend, is certainly possible, but is more Battlestar Galactica than Star Trek the Next Generation.

  • Greg

    I appreciate what you have to share. However, I hope you can offer me some helpful, practical, but affordable advice. For my graduate thesis, I plan to conduct focus group discussions with Latinos. Although I study Spanish, I am not fluent. A transcriber has committed to transcribing interviews and group discussions at reasonable rates, but she does not know Spanish. I doubt that I could afford to hire a professional translator. Based on your article, I am afraid I may be out of luck. What might you suggest to a cash-strapped graduate student?