Bookshelf update — controversial Russian reference

On Monday I have announced a «very special reference» to be posted the same day. Obviously it didn't happen, but it's better late than never. Without further ado I give you ГОСТ 2.105-95 Общие требования к текстовым документам. What's so special about it?

First of all, being an active [inter]national standard, this document is mandatory, and you are formally  obliged to follow it, at least for technical translation.

At the same time, it contains huge amount of outdated, unneeded, unrealistic, and controversial provisions, that nobody follows these days. Obviously, this opens a wide array of abuse possibilities for a willing reviewer. Here's some examples:

  • You can't use any acronyms, unless they are outlined in National Standards.
  • You can't use minus sign (-) in the continuous text, type the whole «minus» word or die.
  • Table of contents title should be centered.
  • You can't use borrowed words, if Russian analog exists (:trollface:).
  • Table column is actually called «графа», my friend.
  • A zillion of conventions for tables, UOMs, external, and internal references, you are definitely violating all the time.
  • Etc.

Have fun.

PS: <Serious mode on>Those who are making their first steps in this business, and aim for hardcore technical translation, should learn this this paper by heart, as well as all other relevant standards. We have all become too relaxed, too casual with all that MS Office apps, CATs, search engines, and online dictionaries around us. Being uncompromisingly rigorous, and precise in the world casual Margarita-drinking dabblers may become one's real competitive edge. Note however, that this path is tough is unrewarding for most.<Serious mode off>

PPS: I'd like to know more about applicability of government-imposed National Documentation Standards to real word translation work in different countries. Have you heard about those? Are they relevant? Usable? Feel free to post a comment, if you have something to say.

Why do I hate Logoport (aka Translation Workspace), part 2


Part 1 can be found here

As we have found in Part 1, Translation Workspace is a business model based on charging your suppliers for the privilege of working with you. It was never intended to act as a "real" CAT tool or tested as such. I don't think that TW developers have ever heard about such illusive ideas as productivity, ergonomics or user experience, and even if they have, they couldn't care less. The result is major usability flaws, which make Translation Workspace even better target for our devoted hatred. I wanted to make a numbered list, but I couldn't decide which flaw is the worst to put it on top, so it's unordered bullets.

  • Repetition handling. Surprisingly, Translation Workspace doesn't save a segment to TM immediately when you close it. I think it happens once in several minutes, so if your file contains considerable amount of reps, you'll have to retype them, use copy-paste or save the file before opening each repeated segment. Major waste of time.
  • More on repetitions. "Master" TM overrides your changes no matter what. It means, that if a segment is present in Master TM and occurs in a file more than once (imagine 20 instances), and you need to edit it, you'll have to go with retype, copy-paste or find and replace. In the ideal world, where nobody is hungry, everyone is using solar power, and can date Megan Fox or Johny Depp if they want to, Master TM would contain only perfect, thoroughly checked and approved translation units. In real world however (goodbye, Megan) Master TM is the legacy of the time, when nobody cared about localization for developing countries (Russia in my case), content was sourced to very cheap local sweatshops and published without any checks. Being a professional, you have no choice but to correct all that is wrong, so even more of your precious time goes to waste.
  • Stability. All facets of Translation Workspace are highly unstable on my scrupulously clean finely-tuned working system with the best broadband you can get (believe it or not, I can watch a 1080p movie from Netflix server on the other side of the world without any hiccups). Every 30 minutes or so Translation Workspace hangs. Sometimes for good, sometimes it returns to normal in 10+ minutes. And no, I am not the only one with this problem. Guess what it does to your precious time?
  • Latency. Moving between segments is always accompanied by a delay. It can be anywhere between several seconds and half a minute based on the server workload. Having delays like this for such a small packets of data is a major failure in the world of Telepresence and MOO shooters. This is bad for two reasons. First you loose ~2h per 1000 segments compared to any real CAT tool (This is big. 1000 segments may often correspond to ~2000 adjusted words which take approximately 4 hours to translate in a normal tool. 4h+2h=6h => ~50% overhead). The second reason is described in the next bullet.
  • Pop-ups. Lots of pop-ups. Obviously, the process of translation implies strong concentration. Open the segment -> Type your translation -> Place tags if needed -> Move to a next segment > Go on. Carefully set up hotkeys, no pauses, no distractions. For experienced translators it develops into a meditative state of mind, in which one almost completely ignores the external irritants, which may include TV, construction work outside or your wife saying something about supper or taking away the garbage. Loosing this concentration is painful, and that's exactly what happens when continuity the process is interrupted by a pop-up window. "Are you sure that you are freaking sure sure?" Pop-up when you move tags in a fuzzy TU, pop-up when you're closing a segment with different spacing around a tag, pop-up when "new segment is coming from TM" whatever that means, pop-up for no-specific-reason-just-for-the-fun-of-it, etc. I haven't used Logoport for a while, so this list may be incomplete. Anyway you work slower, and guess what, waste your time.
  • XLIFF editor doesn't support any export. Yes, it's true. Not only poor souls have to buy Logoport, they actually must work in it. No choice at all.
  • Various minor, but still irritating bugs. For example, hotkeys may just stop working for reason unknown. It can be solved by restarting the application, but given the exhaustive five-step login process, it converts to even more time waste.

No big conclusion here, sorry. Just one thing: time is money; everyone who wastes your time is as good as thief and should be treated as such. Feel free to post your experiences here, I am sure I have omitted something.


MT vs. Transcreation

Yes, it's another MT-related post.

No, I am not scared.

Really, I am not.

I can't help but notice two curious trends on modern L10N market. They may look unconnected, or even opposite, but I think they represent two sides of a single market polarization process at its earlier stage.

On one hand we see the marching Mighty Translator (MT), which is said to produce decent results in some pairs. Evangelists even predict that it will completely take over certain segments within few years. I am not sure about that, hyped statistical MTs are still useless between EN and slavic languages... algorythmic/rule-based systems are slightly better, but they don't get much investment. Anyway the trend exists — researchers are researching, customers are asking for case studies, GPMs are launching pilot or even production projects, CAT developers are implementing MT plugins... A lot is going on the scene and behind it, and many find it disturbing.

The 2nd trend is more subtle. The populariy of Transcreation concept has been growing for last few years. It was mentioned much more often in 2011 than in 2008, when I received my first Transcreation PO. Bloggers are writing slick posts (<envy mode on>What can I do to make my English that slick?<envy mode off>), theme discussion groups appear all over the Web, companies and freelancers quickly change their positioning or at least tune it with new idea. It happened with two good customers of mine, whose specialization has mutated from "marketing and communication" to "transcreation" within 2011. This is not a hype, but you surely can't miss it.

Now to the point. MT is a logical development of Henry Ford's workflow-oriented "more eyes see more" approach to localization. Supplier chains are long, resources are cheap, but sloppy, quality is "normalized". Customer is supposed to be protected by a long sequence of steps with multiple feedback links that looks fancy on the whiteboard. Every step is supposed to improve quality, though usually it does not. Good news is that nobody will ever read the bigger part of content produced this way. So if you replace the initial node (cheap human translator) with MT, nothing changes. Price drops a bit, and that's it.

But the whole workflow thing has never worked for marketing and advertising copy, and majority of translation buyers have learnt this lesson hard, harder or the hardest way. That's why agencies, who specialize on such content, are applying different approach: do it right the first time. There is one qualified translator, and one qualified editor, both not cheap. They usually work in close contact to produce a fluent copy that sells stuff. Imagine a CIO reading an MTed collateral for six-figure product.

The logical development of this "do it right the first time" approach is Transcreation, which is [IMHO] about giving a translator more freedom and responsibility to produce a copy suitable for local audience expectations. No more source limitations, the goal is to sell stuff, not to stick to the original. Do whatever you want with idioms, wordplay, and overall emotional temperature. Do whatever you want, as long as they buy it. Such translation is essentially a new copy, [loosely] based on the original with all that it implies.

So let's return to the beginning of this post — polarization thing. Here's my layman's prediction: High-volume technical projects with low to medium level of sophistication (high-volume software, consumer manuals, tender docs, etc., you name it) will be slowly moving to human-edited MT. It's happening already. Sad but true, sorry MS translators. The change will be driven not only by costs, but also by volume of translatable content.

At the same time Trascreation wave should rise for stuff with higher fluency/creativity requirements, particularly marketing and games (hopefully), training materials perhaps. May be they will even be selling MT under the Translation label, and rename regular professional human translation to Transcreation. And I don't know, who will make more cash per hour, MT-editors or Transcreators [hehe]... Think about it.

Traditional technical translation should remain under some name for highly specialized texts, which require hands-on experience in the narrow field. I won't tell why — tired of typing.

One more thing, I wanted to use "Transcreation equals Translation in hipster glasses" joke, but couldn't fit it in the main text, so I'll just put it here.


Why do I hate Logoport (aka Translation Workspace), part 1


I have seen plenty of bad CAT tools, and only two or three really good ones. Some were particularly ugly, like IBM Translation Manager or Multilizer 4.0. (I have translated several thousand words in fourth version of Multilizer in 2010. Since then I prefer to call it Mutilizer [HA-HA]).

However I have never had any emotions about those, working with bad proprietary tools is a part of the job. When you manage to export files to something you can open in your favorite environment, you win. When you don't, it's like whatever. Ultimately all CATs are the same.

But this tolerance failed me when I became familiar with Logoport/Translation Workspace. I hereby admit that I hate it from the bottom of my heart. Reasons are not entirely clear even for me, but I'll try to analyze them for your amusement. So...

First of all it's Logoport's shameful history, which started when SDL swallowed Trados. Lionbridge became dependent of one of its dire competitors, and felt somewhat uncomfortable. For SDL on the other hand this situation opened wide array of exciting opportunities. So Lionbridge went shopping, and bought small unknown company with unknown undeveloped downright useless copy of Trados macro, which communicated with server instead of local TM. Note that it was long before the cloud hype, so Liox can be considered kinda visionary from this perspective.

Then the fun started. By moving projects to Logoport Liox earned some hard cash and lost what was left of their goodwill among translators, because of "optimized" WC algorithm: Similair (but not equal) no match segments within the same project were considered fuzzy. The basic matching mechanism was also "improved" compared to other tools on the market, for example Introduction -> Implementation is considered a high fuzzy, because those two words are so much alike. Don't you agree?

The more, the better. Logoport remained free for a few years, and everyone got used to it. Just another ugly proprietary CAT. Then someone in Lionbridge came up with a brilliant idea: Let's charge for it! Our suppliers will have no choice but to pay, if they want to continue working for us, and our suppliers' suppliers will also have to pay for the same reason. Free cash! And free cash it was. Some good translators left, but SLV agencies who received over 50% of their work from Lionbridge had nowhere to go.

As you can see, greed is the only reason this tool exists. It was purchased to avoid paying SDL for Trados license on their conditions, and used to rip off vendors in several ways. Noone ever cared about insignificant stuff, like usability, testing, or customer care. Cash keeps flowing, and most "clients" don't have a choice, but to buy and use our "product". Why bother? The result: user experience nightmare.

This leads us to Part 2 of this story — technical/usability reasons why I hate Logoport.

New horse in the stable

All translators, who have been around for some time, have a roster of tools to make their life easier, and improve quality of their work. (This may not be obvious, but the former is a direct precursor of the latter.)

The easiest examples are Apsic Xbench, and Apsic Comparator. I won't be telling you, how great, and useful they are, this is not the subject of this post, but if you don't have them, go get them here right now.

Another thing worth mentioning is search/replace tools, which are essential when you revise a project that contains hundreds of files. My choice here is Text Workbench, which supports binary Office files unlike the competitors.

I can go on, but that's not our subject again. To the point:

Recently I have learnt about the tool, which seems to be just as useful as Xbench for certain kinds of jobs, more specifically liguistic QA of localized web sites, and applications via screenshots, or browser. The idea is very simple — the tool consists of two panes. You open a source screenshot or web page on one side, and a localized target one on the other. Then you compare them, and go to the next pair with a click of a single button.

The tool is opensource, and thus it is flighty, and unstable, but it's still much better than two browser, or image viewer windows.


You can also download the compiled file here

Can't wait to try it for "big" web application testing. Great addition for my roster.