New site, new domain

This archive of The Proceedings of Cabinet Beezer is now frozen. For the latest updates: http://bzr.bzh

Is proprietary translation support software worth it?

I'm a happy Ubuntu user, but the reality is that there are still a lot of Word files flying around in the commercial environment. A combination of Open Office and OmegaT seems to handle them quite well, but on my latest job I experienced a number of formatting difficulties that had to be corrected manually, which was both time-consuming and introduced a new potential source for error.

Open Office crashed on me recently, just as I was tweaking a final .docx file before sending it off. Its automated recovery process then recovered the file to a previous version, causing errors to be reintroduced that had already been corrected in a proofing stage. This was vexing, and required another round of corrections.

One obvious way to avoid such problems in future would be to use the industry standard proprietary solutions. I priced this today: Microsoft Office plus SDL Trados Studio 2009 would be at least 650 euros, a considerable overhead given prevailing market rates for translation.

The decision for the time being is to upgrade to the latest and greatest LibreOffice, and persist with OmegaT, but it's under review. Version control problems caused by software crashes are most unwelcome.

Running with the ProZ

I regularly update a list of significant projects on my professional profile at Proz.com. I recently paid for full membership to the site in order to bid for a job going by within my fields of interest: and landed it! Which was pleasing. The short test translation of 50 French words translated straight-forwardly into 38 words in English. The site announced ten bids for the job when I went back to look a few hours later, but it was my bid that received a return offer towards the end of the afternoon last Tuesday.

The work required the translation from the French to English of several abstracts and full papers for a specialist healthcare conference to be held later this year. Of the full papers, the longer was fully within the scope of my professional interests in health care and information technology; the other had some straight-forward clinical material, combined with some technical detail in a subject that I have to confess I had contemplated only in the most superficial way in the past.

The search for background information online to assist in my comprehension of this arcana was necessarily brisk, as the first tranche of the delivery had a deadline less than forty eight hours away. That slight hump surmounted, the deadline was easily met.

The longer paper, due last, has had its moments, as my dear old Aunt Margaret used to say, but I now have the comfortable satisfaction of a fair copy of the final draft circulating in-house for minor corrections fully 24 hours before I had initially planned to send it off, and 36 hours inside the deadline.

Revealing my brilliance as a translator to the world

All my work so far has come from personal contacts, but it's not enough to live on. Sites such as proz.com and translatorscafe offer the prospect of getting work from around the world, but it's not so easy: there are 15,478 translators offering my language pair on proz.com for example. I am buried in a heap of translators, many of whom much more qualified than I, and with decades of experience. No fewer than 7313 claim competence in medical translation. As on all search engines, the question of ranking is key.
There are two obvious ways to increase this ranking: one is to cough up the cash (114 euros/year) to join the site; this would reduce my competitors to a mere 978 others. The other way is to answer so-called KudoZ questions: tricky questions of terminology asked online by other translators. I tune in to this every day using an RSS feed, but they are indeed tricky. When I do know the answer, I have mostly found that the question has already been answered by someone else. As this means that their answer is more likely to have been agreed by peers coming later to the question, it is they who get the points. Still, this morning I received an email informing me of my first four points. Hooray! But some of my colleagues have thousands... ulp!

Adding Scribus and OmegaT to the infrastructure

Set off on the latest job expecting to meander: and duly did, getting somewhat bogged down in software issues, rather than Just Doing The Work.

The job, through an agency, was delivered as reasonable photocopies of the paper originals as the source text for the job. And at the weekend, and by hand, no less. Searching around on the internet for guidance (for example), it seems that delivering a reasonable facsimile of the original layout to ease later official scrutiny seems to be the way to go.

Of course, if all we had were a manual typewriter we could do a reasonable job, providing the font sizes in our original were more or less congruent with what our typewriter could do, and we had a pasteup board, T-squares and set squares, metal rule, craft knife, cowgum et cetera to hand.

But please! This is the second decade of the twenty-first century: surely we can do a little better than that?

Well, of course you can. Running Ubuntu 10.10 on a 380 euro netbook (it's an ASUS EEEPC1015N which is already a nice start, with (a first ever, in my humble experience with the Ubuntu install since ~5.x) scan/print/text/openoffice/ all JustWorking(TM) straight off the default+latest_updates install.

Well, if I was using the Mac for this about 1995, I would have certainly used either PageMaker or Quark on Mac OS 7.x (with a laserwriter available), but the final artwork would have been camera-ready, possibly still with some pasted (physically, with Pritt(TM)) elements on the original. This means mounting it on board, which adds to the longterm storage and archiving problems.

Now, with Scribus and OmegaT both readily available for GNU/Linux (with a bit of googling, and the goodness of apt-get) there seemed no very good reason not to plunge in, and try to use both for the translation task in hand.

Ahem! I am afraid I am that both inexperienced and impatient user that they're looking for :-)

Still, it seemed reasonable to file a bug report when a preferences dialog ignored users of 10 inch screens. The Scribus project follows best practice in the free software community in enabling bug reporting and tracking in a completely open way.

Well, as is so often the way when we seek the goodness and wisdom of gnu, time passes pleasantly and copiously while you check versions and updates and reviews online in preparation for your big splashy launch. And after all, if you can set up something that's nice to use that is completely free, that is a considerable saving in annual overheads, in comparison with signing a license agreement for the proprietary software (Trados, Wordfast, etc etc) and paying handsomely for the privilege. Wouldn't it, I thought, be nice if I could actually (i) solve all the install problems, and (ii) get the first page (of ten) actually done. By five.

I had high hopes. The morning came and went. I cooked for my lovelies, but then got straight back to the office as they siesta'd after lunch. Fortunately Trevor was rockin' online.

By four, it was obvious we were going to crack it. Well, the Scribus part of the equation. But the technique is apparently to deliver up all the text to Scribus already formatted in OpenOffice (any 3.x should be OK, I run the Oracle branded 3.2 that was the default in Maverick Meerkat 10.10), and then import the text at the last minute. Along the way we sucked our teeth at the lack of undo in the built-in text editor, though that feature's apparently on its way.

So I still haven't translated a line, but it was Sunday after all, and you've got to sharpen the saw.

That final quarter

So far, work's been arriving through local networks rather than online. Recent efforts: 4 hours of coaching for an academic bound across the Atlantic to give a talk at a conference in Florida, editing a major paper in bio-inspired robotics for an international journal, and a fire notice. The latter was the trickiest, and payment for that will be only in kind, but it was also quite fun.

I have updated my CV and registered with proz.com and translatorscafe.com. I've sent off a few responses to job offers posted there, if they seem relevant to my interests, but so far I've had only one response, which merely indicated they were keeping my CV on file. There's not so much work out there, and where the rate negotiations are opened by the potential client, rates of about $0.04-0.05 seem prevalent, so I guess I'm always losing the auctions. That's fine though: I haven't seen any jobs going there by that are so interesting I'd think about doing them for free: most often it's dismal financial documents, or bizarrely obscure software manuals.

The next step is to translate all this material I have assembled in English into something coherent in French. Once that is done I'll contact local agencies, and some contacts I have at the medical faculty. I have an idea for a seminar on using the internet for translation which should aid the development of relevant networks locally.

How much to charge for translation?

The question of how much to charge for the translation of a substantial scientific paper from French into English arises once more. Some published rates on the internet: 0,10€/word, 0,15-0,25€/word, 0,20€/word, 0,20€/word 0,13-0,19 $ CAD/word (0,09-0,13€/word), 0,10€/word, and so on.

More technical texts command premium rates, and longer texts are cheaper: >10000 words, 5% reduction; >25000 words, 10% reduction. Some translators charge more for technical hassle, seemingly defined as a source file in anything other than Microsoft Word. Many have a minimum charge, with texts of <300 words being charged as 300 words.

All this seems reasonable. I plan to charge 0,10€/word, and seem to work at about 300 - 600 words/hour, yielding a gross hourly rate of 30-60€, which seems reasonable: there are expenses to deduct, taxes to pay, and the possibility that the time required will balloon upwards if the text is tricky.