Something’s Fishy

Google Translate is amazing. You can write fairly natural English and it'll produce a reasonably acceptable German equivalent most of the time. Results may vary if you start to use wildly creative colloquialisms or misspell words. Now, if you speak German reasonably well, then as you can see, even fairly well-written English sounds a little off in the translation once you write longer, more complicated sentences. These errors magnify when you start to use complex or metaphorical expressions, and then goggle translate goes completely off the rails if you start to dangle craziness in front of it, not to menshon misspellings. Four example if ewe translate this pair-o-graph into German with goggle translate, you get the following:

Google Translate ist unglaublich. Sie können ziemlich natürliches Englisch schreiben und es wird die meiste Zeit ein einigermaßen akzeptables deutsches Äquivalent produzieren. Die Ergebnisse können variieren, wenn Sie wild kreative Umgangssprachen oder falsch geschriebene Wörter verwenden. Wenn Sie nun einigermaßen gut Deutsch sprechen, klingt, wie Sie sehen können, selbst ziemlich gut geschriebenes Englisch in der Übersetzung ein wenig anders, wenn Sie längere, kompliziertere Sätze schreiben. Diese Fehler verstärken sich, wenn Sie anfangen, komplexe oder metaphorische Ausdrücke zu verwenden, und dann geht die Brillenübersetzung völlig aus dem Ruder, wenn Sie anfangen, Verrücktheit davor zu baumeln, und keine Rechtschreibfehler. Vier Beispiele: Wenn Mutterschaf dieses Paar-O-Diagramm mit Brillenübersetzung ins Deutsche übersetzt, erhalten Sie Folgendes:

So...  when teaching first-year college German around 2001-2004 in Bloomington, Indiana - at a time when Babelfish was the closest thing to an online "translator" that we had available to us, producing results ranging from surprisingly acceptable for incredibly short sentences to barely legible nonsense when you used any complex constructions, slang or incorrect spelling or grammar of any kind - you can imagine that grading essays ranged from disappointing to downright entertaining. For example, take this short essay about a favorite actor:

Ich wählte Arnold, weil ich liebte seine Filme für eine lange Zeit habe. Ich wurde vom Abschlußwiderstand geerdet, als ich 2 für das Sagen eines schmutzigen Wortes von einem seiner Filme war.

Auch ich habe Respekt für jemand, das am Tun etwas festlegen kann soviel wie, er, als er ein bodybuilder war. Er brach in die Gymnastik ein, nachdem sie geschlossen wurde, um anzuheben. Die ist Widmung.

Now, again, if you speak German reasonably well, you'll immediately see that this paragraph makes absolutely no sense. It's somehow mostly? grammatically correct (because Babelfish could get that right for the most part), but it's just completely incomprehensible. So, as a teacher dedicated to better understanding our students so that we can better guide them towards success - even when that means confronting them about possible cheating - Rachel and I of course tasked ourselves with trying to figure out exactly what this student must have typed into Babelfish to come up with the utter nonsense that they printed out to hand in for homework that day. After a couple of hours at the library computer lab that we met in every Sunday, we finally came up with this:

I chose Arnold because I have loved his films for a long time. I was grounded from the Terminator when I was 2 for saying a dirty word from one of his films.

Also, I have respect for someone that can commit to doing something as much as he did when he was a bodybuilder. He broke into the gym after it was closed to lift. That is dedication.

Feeling relatively confident that this was what they had typed in, I printed it out, stapled it to the front of their essay, graded it with a big, fat zero, handed it back, and let the entire class know that I'd accept rewrites for submissions that were clearly Babelfish this one time only if they would also spend some time working with Babelfish that weekend to think about how it makes the mistakes that it makes, and then write up a brief (English) description of what they think happened with their translation attempt.

Now, we've been talking about including a little bit about our Babelfish adventures for as long as we've been writing this book. But it wasn't until now, 17 years later, that a completely different conversation with a professor in the Spanish department at a nearby university started talking to me about a student who had clearly plagiarized some Spanish - using constructions that they clearly wouldn't know how to use in a paper on linguistics - that the issue reared its head again. She considered putting what the student had written into Google Translate to see if what came back was from Google Translate or if it was just straight-up copied from some Spanish source. I told her that putting it back in to find out what they might have tried to type wouldn't yield productive results. It's like making a copy of a copy. The conversion back into the original text was almost impossible by simple translating it backwards. Language is too diverse for that. She would have to try to think of what the student could have written in English, and then - somewhat a trial and error fashion - try to finesse it until it spat out the text that she was staring at. It was then that it hit me what what we had been doing with our Babelfish essays back in 2002 was incredibly similar to the process of checking hashed passwords when logging a user into a secure website. Let me explain...

When you choose a password for any responsible website that uses a halfway respectable level of security, your password is not stored in the database. Instead, it is "hashed", meaning it is encrypted using a one-way algorithm that cannot be un-encrypted. That hashed password is then stored in your data row. Whenever you log in, the password you enter is hashed using the same encryption method, and if the hash matches the one in the database, then it knows you entered the right thing. But if you were to contact someone in charge of the website, they would be unable to provide you with your password, because they don't actually know it. That's why we typically have to reset passwords instead of just retrieving them. They're not able to tell us our password, and that's a good thing.

I suppose I ended this post on this anecdotal note not only to make that loose comparison in itself, but rather to muse about how interesting it is when I encounter ideas relevant to both my old life and the one I live now. It's been a long time since I've been in language education. I taught my last German class in 2004. I'm a programmer now, and while I have taught my share of programming courses over the past few years too, it's a little harder to find the cheaters unless they drop large chunks of clearly stolen code into their work. And frankly, that's part of the job description, so I hardly mind it as long as their application works.