Typo Pad

The Perils of Machine Translation Illustrated

You'd think everyone would be aware of the fact that automatic machine translation tools -- such as Google Translate, Babelfish, or Systran -- do not create polished, publishing-quality translations.  They are useful tools, but what you get is a "gist translation" which requires human correction, sometimes a lot of it.

Still, more and more sites are offering automatically generated translations of their content.  Many of the users of such services are probably completely unaware of what gibberish they are putting on their web sites.  And it would be just their own problem if it were not for the fact that these junk pages end up in Google results.

In order to avoid improving their search engine rankings, I will not link directly to any such sites, but an example is at www dot raymond dot cc slash blog slash archives slash 2007 slash 06 slash 04 slash mac-osx-tiger-theme-on-windows slash fi slash

In the interest of illustrating the problem, here is my carefully human-generated translation back to English of the Finnish text on that page.  For orthogonality I have put in Finnish words where the translator had left in an English word.  This particular translation tool also left in some grammatical particles which I have simply written as PARTICLE.  I have not referred to the English original text.  Here's the first couple of paragraphs.

Boy OSX Jaguar Subject Malli jälkeen Window

I-LETTER stake olet' hear multiple PARTICLE occasionally that lot PARTICLE Window Outlook connection has to cope descend Son OSX. Want so that checkered broken Son OSX connection itself only special language' gives only that is special language' put Intelligence translation PARTICLE Son OSX degrees? Your endure preemption be so that install Son OSX subject malli jälkeen Window.

dobee lienee cause joltain excellent Son OSX subject guest OSX Jaguar duration Windows.

Laugh or cry?  Your call.

But seriously, the providers of these tools and services should make sure they don't end up polluting search engine results.  The Robots Exclusion Standard explains how to do this.  Basically, put in a "noindex" meta tag on the generated page.

Some of the translations are kind of peculiar.  The word "akkuna" means window but it's markedly dialectal.  The word "etuosto-oikeus" is very specific to stock trading; it does mean "preempt" but in the very narrow context of shareholders having the right to buy shares before they are sold to a third party (it quite literally means "preemptive buying rights").  In fact googling for 'translation "etuosto-oikeus"' mainly seems to bring up other samples of badly translated pages, probably produced by the same tool -- apparently this one originates from Google Translate (probably via some third-party tool) apparently not.

Update: this looks like the culprit: http://anaconda.taragana.net/angsumans-translator-plugin-pro-version-31-released/

Quotable quip: "Did you know Google loves our translated pages?"  Oh noes.  I rest my case.

"Dear" mail client provider,

Could you "pretty please" remove the gratuitous "double quotes" you put around my "name" and the "names" of all the "people" I correspond with?

From: "Humble Self" <me@my.example.org>
To: "Honored Correspondent" <person@example.net>
Subject: "Gratuitous" double quotes

Correct me if I'm wrong, but I believe this atrocity was introduced by Microsoft when they desperately tried to convert their legacy email software (from when completely company-internal systems were the norm) to cope with what was then the new and mysterious Internet.  Regardless of its origins, this convention has no technical merit.  Granted, the old RFC822 mandated the use of quotes if the real name happened to contain special characters, such as single quotes, but in this day and age, the preferred mechanism for ensuring transport-safe strings in email would be MIME encoding.  So (in transport) you would have, say,

From: =?iso-8859-15?Q?Humble_=36self?= <me@my.example.org>
To: Honored Correspondent <person@example.net>
Cc: =?utf-8?Q?=C3=84cc=C3=ABntric_=22weird=22_=E5=81=8F=E5=BF=83?= <centric@cn.example.com>
Subject: "Gratuitous" double quotes

... but any competent MIME-aware client should transparently display this as human-readable, with the MIME jumble decoded:

From: Humble $self <me@my.example.org>
To: Honored Correspondent <person@example.net>
Cc: Äccëntric "weird" 偏心 <centric@cn.example.com>
Subject: "Gratuitous" double quotes

Notice that a double quote in someone's name (as you often see with nicknames) is no longer a problem at all, and you have to have this mechanism in place anyway if you want to cope with names which are not purely 7-bit ASCII, or which happen to contain some of the the rather haphazard set of characters which are significant for the mail transport layer.

Grateful in anticipation of the day when I will no longer need to apologize for my choice of mail client,

"Just me"

Six Apart support emails

This would be sort of humorous if it wasn't so depressing.

I found a number of Typepad blogs which are simply shock full of spam.  It appears that the blog submission address has probably ended up on spammers' mailing lists, and so the spam goes directly to the blog.

Here's a Google search which demonstrates this.  It appears that these blogs are basically on autopilot, and only the spammers have been posting to them lately.

So I tried to contact Typepad support, which goes to the mothership, Six Apart.  I was too lazy to sign in and I wasn't reporting this as a paying customer of theirs anyway.  Well, that didn't turn out so good.  I have now received five substantially identical emails with basically no intelligent content at all.

Allow me to quote that in case the image isn't completely clear.

Recently you requested personal assistance from Six Apart Support.

Below is a summary of your request and our response.

Thank you for allowing us to be of service to you.

To respond to this message, please reply to this email and type your
reply between the lines in the space below. Text entered into any
other part of this message will be discarded.

[===> Please enter your reply below this line <===]

[===> Please enter your reply above this line <===]

There is no indication of what they are responding to (what if I had submitted multiple support requests?) and no indication of what sort of response they are soliciting, in spite of the phrase "below is a summary of your request and our response".  I replied with a question about what they need from me and what the status of the problem is, only to receive another one of these, etc.  Finally I just said forget it (to which I received the reply you see above).

So Melanie and Kymberlie, if you see this; hi, that's you in that picture.  (Sorry, couldn't resist.)

Seriously, I'm hoping their paid support is better.  (My experience is that it is.)

Suomen paras työ väli lyönti paikka

Muuan konsultointiväli lyöntiyritys on kovasti kampanjoinut viime aikoina olevansa Suomen paras työpaikka.

Jaa, kehittäjä sen kun päätti että selaimessa ei nyt vain voi käyttää tabeja? Hyvä homma.  Sopii ylpeillä.

(Eikä siellä yllättäen ollut mitään kiinnostavia paikkoja.)

My Bookmarks are Dead, Long Live the New CiteSeerX

Here's a copy of a message I sent via the CiteSeerX feedback form just now:

This new CiteSeerX is nice and all, but I have hundreds of bookmarks to the old CiteSeer which are now broken.  I sincerely hope this is a temporary condition, but even if it is, the deadline for my thesis is in two weeks.  Is there any way I can reconnect those old bookmarks to CiteSeerX pages?

I don't think "hundreds" is an over-estimate, but it's hard to know, because of course, these are not all traditional bookmarks in a single browser's configuration.  I have about a dozen instances of Firefox in various states of tabbed browsing hell, and many of those have some tabs open to CiteSeer pages I need to be able to find again.  I also have more than my fair share of social bookmarks, Linkwad depots and what not. (All of this is part of my struggle to organize some sort of scalable and pseudo-organized memory of sites I want to be able to find again.  ma.gnola and Linkwad are my current bets, but they're probably not the Final Ultimate Solution.) And of course, I have copy+pasted CiteSeer links to various places over the years.

Color me cynical, but wouldn't it be typical if they were to respond something along the lines of "sir," (or "madam;" happens all the time that silly Americans wage a 50/50 bet on what gender a foreign first name sounds like) "we have never supported the use of bookmarks on our site."

(And just to add insult to injury, "Besides, don't you think this new flashy <blink>Ajax</blink> thingy has a lot more, what's the word, usamability?"  But I digress.)

Having said that, the CiteSeer folks have been running a very useful service over the last 10 years, and always been very helpful and forthcoming when I had a comment or a problem. I just really really wish they have a plan for how to allow external bookmarks to continue to work even though they migrate to a new infrastructure.

(Oh, and sorry for not blogging. The reasons should be somewhat obvious. I have my commitments clear, and am communicating to stakeholders via email. Commitments do not include blogging. See you in April.)