Spam still lost in translation
- May 20, 2009 3:30 PM |
- By Paul Jay
By Paul Jay, CBCNews.ca.
Yesterday Google announced that it was experimenting with GMail yet again, this time integrating the free email service with Google Translate, in what they say is a "step towards automatic email translation between 41 global languages."
The tool isn't perfect, but Google said they hope they can keep tweaking it to improve the quality of translation.
I ran a quick test this morning, asking for the French translation of this sentence: "My name is Paul Jay and I enjoy a piece of toast every now and then."
It spit back this: "Mon nom est Paul Jay et j'aime un morceau de pain grillé de temps en temps." Hey, that's not bad.
Let's see what happens when I translate it back: "My name is Paul and I like Jay a piece of toast from time to time." Uh, oh.
This got me thinking about a conversation I had with Ryan Naraine over at Kaspersky Labs a month or so ago about why spam is so full of bad grammar.
One theory, which his employer Eugene Kaspersky offered during an interview last fall, is that since spam is so successful, there is no need economically to invest in making spam pitches more believable.
Naraine offered another theory: sending out a spam mailing can involve criminals from disparate parts of the world. As he said, a guy in the U.S. could be selling a product, but it could be a guy in Brazil that writes the malicious code and a botnet owner in Bulgaria who actually sends the message. At any or all of these steps, he says, the message could be created or altered.
So I thought I'd see how Google Translate would handle one of these messages. Here's the original message, using already somewhat garbled English (I've put a few asterisks where the company/product name is.)
An incredible opportunity to try out the worlds best selling ***. We are sending out FreeTrials paks of 16 of either *** or 8 of each. You decide which ones you want. Only catch is you must act fast as we can't keep giving these away forever. See ya there.
After putting it through Portugese and Bulgarian and then back to English, here's what we come up with:
A great opportunity to experience the best of worlds sales ***. We FreeTrials Paks of 16 or *** 8 each. You can decide which ones you want. Only catch is to act quickly as we can not keep them away forever. See you there.
This adds weight to my belief translation technology is behind our nonsensical spam. The new message includes the awkward "best of worlds sales" and the ironic "we can not keep them away forever." But on the plus side they converted "ya" to the more grammatically correct "you."
Google or someone else may end up improving this and make translation software into a true Babblefish.
But there is a part of me that hopes they don't ever get too good at it.
All News blogs
- Universe hates Higgs boson, Chicago Cubs
- By John Bowman, CBCNews. A physicist working on the Large Hadron Collider doesn't think much of the theory that the universe is sabotaging the project to prevent the discovery of the Higgs boson. Might as well say that Nature hates... Continue reading this post
- Large Hadron Collider goes Back to the Future
- By Peter Evans, CBCNews.ca. Two respected physicists have put forward the theory that the Large Hadron Collider's stated aim of finding the Higgs boson might be so abhorrent to nature that mysterious forces are traveling back through time and sabotaging... Continue reading this post
- Multi-touch concept for desktops: 10/GUI
- By John Bowman, CBCNews.ca. I'm a fan of alternative ideas for human-computer interaction, so this video caught my attention. It shows an idea for a ten-finger touchpad interface and associated changes in the way a computer would handle multiple windows.... Continue reading this post