1. The End of the Five-Year-Rule
(Premium Content)
|
|
At the above-mentioned TCD conference there was a (semi-) official
declaration that machine translation's five-year-cycle has ended. The five-year-cycle
started in 1954 after the IBM 701 "dashed off its English translations [of
a handful of Russian sentences] on an automatic printer at the breakneck speed
of two and a half lines per second." The project
leader Leon Dostert declared:
"Those in
charge of this experiment now consider it to be definitely established that
meaning conversion through electronic language translation is feasible."
And
although he
emphasized that it is not yet possible "to insert a Russian book at one
end and come out with an English book at the other," Doctor Dostert
predicted that "five, perhaps three years hence, interlingual meaning
conversion by electronic process in important functional areas of several
languages may well be an accomplished fact."
We all know what happened afterward: This imagined five-year-goal was
pushed out every five years for another five years, until we suddenly find
ourselves in 2010. And the output of machine translation is mostly still not
particularly "good," but it certainly is used. Since in many cases
the concept of "good" has been replaced with "usable," it
has become, well, usable (sort of). And therefore the five-year goal is a thing
of the past. We now know that MT's limitations will stay (and gradually become
"better"), and we have all accepted that there is a certain level of
usability within those limitations.
And what we certainly do see is a lot of creative juices flowing to
quickly collect more humanly translated data to train the machine translation
engines to make them even more usable. Some of these efforts tend to go against
our grain (for instance, see this blog post by someone with strong feelings on Google's
efforts through the Google Translation Toolkit), while others may be viewed
as more generous because there is more given in return.
A case in point is Microsoft's new Website Translation Widget (I was alerted to this by a blog posting by Dave Grunwald).
Like many before, this little MT application sits on a webpage and
allows you to translate the content of that webpage into any of the two dozen
offered languages. And the result is often, well, not very good. See for yourself right here. What's different about this is that it
combines crowdsourcing (sorry, I know that many of you don't like this term!)
and machine translation. Again, the result of the machine translation is less
than stellar, but you as a visitor to the website (or as a collaborator or
owner, for that matter) can click on any of the sentences and edit any of the
segments. If in our example you choose German as the target language and
click on the segments, you will see that the first three segments have been
edited and approved (by me, as the website owner).
The difference between this and Google's machine translation
correction efforts is that the human efforts do not (only) disappear into a
black hole for the betterment of Google; instead, these corrected and
approved translations will from now on always be used for your specific webpage.
(The program is in beta right now, so sometimes the corrected translations will
be listed as alternates rather than the actual translation. This will be
changed and may even already have been changed when you read this.)
These translations will be specific to your webpage and will not be used
as the standard translation for any other webpage, even if the original content
of that webpage is the same.
Well, that's actually not quite true since they will be used by Microsoft
to train its machine translation engine, so in a roundabout way they will be
used by others as well.
So, yes, Microsoft will benefit by getting very
high quality data, but it seems to me that the benefit given in exchange is
tangible enough to make this an interesting barter (plus, at some point, the
webpage owner can also download the translated data as a TMX file).
|
| ADVERTISEMENT |
|
SDL Trados Studio 2009 Service Pack
2 is here!
Celebrate the launch with a free 30 day trial for SDL Trados Studio
2009 SP2. Plus, enjoy exclusive discounts of up to 25% when you buy online.
Hurry, offer ends 31st March.
To
download your trial or buy online visit -- www.sdl.com/toolkit_03.
|
2.
@coöperate
|
|
(You notice that I spared you from the all-too-common wordplays with
"Twitter," "tweeting," and "tweets" for this
heading? You're welcome.)
I would like to officially eat some crow and admit that I was wrong when
I wrote some rather ditzy comments about my by-now favorite social network
application (and really the only one I use): Twitter.
Twitter is useful.
If used right.
Here are the things that I -- well, actually Jeromobot and I --
use it for: We have about 200 people whom we follow -- all from within the
language industry -- and we spend about 20 minutes or so per day to glance
through the 500 or 600 tweets from these people and maybe add a couple of
tweets ourselves. In those 20 minutes daily, we (OK, I'm going to drop the
"we") have learned more about new developments, new tools, ongoing
discussions, and outlooks on language and translation than I could otherwise have
done in three or four times the amount of time.
It took me awhile to narrow down the people I follow to the ones who truly
have something useful to say (not the "Oh-I-feel-like-taking-a-long-hot-bath-after-a-long-day-of-translation"
kind of folks), and it has been useful to have learned how to use the Twitter
tools.
The hash tag (#) is particularly useful when used properly. If it's
placed before a word (note that the word can't contain special characters or
spaces) it makes it searchable and hyperlinked and therefore relates it to
other tweets that also have the same hash tag. Every couple of days I make a
search for #tl8, the most commonly hash tagged word for translation-related
tweets, to see whether I'm missing folks I should be following.
The hash tag also becomes important when you are at an event you are
tweeting about and you want to make sure that other people see it, or so that
you can see what other participants say about the event at the same time. Also,
it makes it a fantastic tool to follow and interact at conferences and
other events without actually being there. For instance, take a look at the
discussions on last week's TCD conference at #tcd11. It will be very
interesting to follow the ELIA conference in Istanbul -- #NDlst -- on April 15 to see
what Jeff Chin and Michael Galvez will have to say about the Google
Translation Toolkit. (By the way, I was promised that my legal fees would
be paid if I sued Google for the copyright violation -- you can read all about
it on Twitter!)
The @ symbol before a name denotes a specific person with a Twitter
account, and using it within Twitter turns that person's name into a
hyperlink that can be used to jump to that person's page where you can see all
his or her tweets and subscribe (or not). @Jeromobot is my
Twitter page, and if you look at my followers you get a pretty good idea
of what I think is useful.
Oh, and if you don't want to sign up to Twitter but still receive
tweets from a specific person, you can subscribe to the corresponding RSS feed,
such as this one for Jeromobot, or for an event or a specific
topic, such as this one for the ELIA conference.
Overall, Twitter is a good tool for building community, and since
many within our industry have turned to Twitter, the community will only
grow stronger. To reinforce the theme from the introduction: that is something
we can only benefit from.
(An added benefit: There are also language jokes
on Twitter. How about this one from @hansfens: "Why do the French
only ever eat 1 egg? Because 1 egg is always un œuf. (Read aloud.)")
|
| 3.
Trados Starter | |
In my last newsletter, I wrote about the new SP2 version of Trados
Studio, and in particularly about the Starter Edition. I had been
briefed by some SDL folks about it beforehand and thought I had a good
impression of what it is and what it is supposed to do, but I was surprised to
see some features described quite differently on the feature list on SDL's
website. Cool, I thought, they changed that, and I wrote:
In
some discussion groups, questions were raised about whether this move with the Starter
Edition was a response to Lionbridge's upcoming Translation Workspace offering. But the
more I looked into it, the more I am convinced that it's not. For instance, it
is neither possible to connect to server-based TMs nor to open packages that
clients might send you, so it has little in common with what the Lionbridge
offering intends to do: be a cog in the larger workflow picture.
It turns out that SDL's website was erroneous (it's now corrected). So, yes, it is possible to
"connect to server-based TMs or to open packages that clients might send
you," which makes the Starter Edition, very much like Lionbridge's
offering, a "cog in the larger workflow picture."
|
4.
Searching, Zooming, and Wheeling
| |
Here is a cool little trick for Firefox users. We all know that Firefox
(and most of its competing browsers) has a little search bar in the upper right-hand
corner. Presently the default search engine is Google (I would not be
surprised if that would change anytime soon), but if you click on the little
arrow to the left of the search it gives you a bunch of other choices of search
engines or other searchable sites as well. You can add to this list or delete
existing entries by selecting the Manage Search Engines option at the
bottom of the list -- most other browsers have something similar.
I only recently learned that if websites use a
certain code in combination with an XML code file on their server, Firefox
automatically recognizes them as searchable. When it encounters that, the
little down-arrow to the left of the search field gets a blue shading. If you
open the search list, the currently opened website is already listed, waiting
to be entered as one of your default search engines. Cool, huh?
And here is another trick -- really, it's more of a bug fix. Many of us
are used to being able to change the zoom in an MS Office document (and
many other applications) by pressing the Ctrl
key and turning the wheel on the mouse. It works great -- only not in some
combinations of Windows Vista/7 and Word 2003/2007.
The fix? Turn the mouse wheel REALLY hard and fast. Not very elegant, but it
works.
|
| 5. TermWiki | |
Awhile back I mentioned the efforts of CSOFT on the development of TermWiki,which is now officially released. I took some
time this week to talk to some folks at CSOFT in order to better understand the
underlying concepts of the product, and I was impressed.
Why another terminology tool? I'm not sure this is the right question,
and I imagine that the folks from CSOFT would respond: What do you mean another
tool? Well, they may not quite say it like that, but they did look at the
existing tool landscape and essentially found two different kinds of
terminology tools within TEnTs: the first group essentially supports mere
glossaries (i.e., bi- or multilingual lists of term pairs) and really frustrates
terminologists because it essentially focuses only on the translation aspect
rather than the whole document creation and terminology maintenance life cycle;
the second group provides full-fledged terminology tools (think SDL
MultiTerm or TermStar) that are so complex they are either not used
at all or too little.
So the goal was to create a tool that provides real ease of use (to
translators, authors, and terminologists) while providing enough complexity and
customizability to make it useful for all parties involved, including the
different levels of access rights that are necessary for the different roles within
different groups. The platform that seemed most suitable was a Wiki: it's very
customizable, easy to use, and role-centric, but at the same time allows for a
great deal of collaboration. Thus the idea of TermWiki was born, and the
development efforts were helped by a large CSOFT client whose commitment to a
new terminology solution acted as a catalyst.
What is now offered are two different versions: an open-source version
that is downloadable from CSOFT's site, and a paid version (with a
per-user/annum fee) that includes certain additional features, including
notifications as well as (the very important) import and export of data. Aside
from that, the normal (un-technical) mortal might be hard-pressed to install
and run the Linux-based open-source tool, so the paid version also comes in a
hosted format (if that is desired).
So how does all this fit into a translation workflow? Well, it doesn't,
and that's really what sort of convinced me. TermWiki (in its current
state) is a platform that allows you to create and maintain corporate
terminology with all the sophistication or simplicity you choose to have. And rather
than bringing the TermWiki interface directly into the translation
process, it allows you to export the necessary (and at any point very current)
terminology data into a TBX or other kind of exchange file, which then can be
imported into the terminology component of your TEnT and be as interactive with
you as your tool allows it to be.
The LSP project managers' job suddenly becomes much easier because they receive
only a link that allows them to export a data set that is project-specific
rather than meddling with 50+MB large terminology that may or may not be
current but for sure contain lots of unnecessary information. Translators might
never actually see the TermWiki, but they benefit from the fresh data export.
And the clients just need to know how to generate correct links (and, of course,
how to maintain the data).
I think this sounds really clever, and as I am
writing this I'm wondering whether this might not be something that TEnT
vendors would be well advised to look at for their own use, in particular
because of the open-source component. |
| ADVERTISEMENT |
|
Having project management and tools interoperability in focus, memoQ
4.0 is just the perfect tool for language service providers.
Find out
more on our new website at kilgray.com
|
| 6. Error Recocknition (Premium
Content) | |
When I posted the title "Eror Recognition" for the first part of
this article in an earlier newsletter, a reader wrote to ask whether I knew that
I had it misspelled. I did.
The article was about our inability to see our own typos when editing
our own work. I'm interested in this topic not only because I have that problem
a lot in my day-to-day work, but also because I have long believed that we essentially
read words very much like Japanese kanjis or Chinese characters. We don't
decipher them stroke by stroke -- we read them as whole images, very much like
we look at words. We experience this when we stare at a certain word and know intuitively
that something is wrong -- typically we are right in our sense of wrongness, but
we just can't see the forest for the trees. We need to force ourselves to look
at it differently so the image dissolves and the letters appear.
One way to do this is to change fonts -- particularly if you change into
a weird font that will first throw you off -- to force your mind to actually
read. Another reader suggested that reading aloud is a really good way of
finding those errors, or, for that matter, having the computer read to you with
a text-to-speech program. (Someone else suggested reading backward -- I'm not
sure how that would work. . . . ) But there are other ways to trick your brain
within Word (some of these also work in other applications):
Switch to Draft font (make sure that you are in the Draft
or Normal view of your document and select Tools> Options>
View> Draft font in Word 2003 and below or Office button> Word Options>
Use draft font in Draft and Outline views in Word
2007). This will give you a
plain text view of your document without actually changing anything. Change the color of the font by highlighting the text and using the
appropriate button on the toolbar (Word 2003 and below) or the button on
the Home ribbon (Word 2007). Change the color of the background: You can either change that by
printing the document out on colored paper -- but that's not very
environmentally friendly -- or you can use the old DOS/WordPerfect look
(blue background, white letters) by selecting Tools> Options>
General> Blue background, white
text. Unfortunately, this has
been kicked out for Word 2007, but you can manually change the
background color in Word 2007 under Page Layout> Page Color
(you can then hover over the different colors and it will show you the
different options immediately).
|
7. A Love Story (continued)
| |
This is an intermittent series about characters I really like. I have a
passion for exploring different writing systems, character sets, and fonts, and
this one
was particularly meaningful to me -- it made me feel at home.
If you have a Javascript-enabled browser, hold your cursor over the
character for a definition.
|
| The Last Word on the Tool Kit |
|
If you would like to promote this newsletter by placing a link on your website, I will in turn mention your website in a future edition of the Tool Kit. Just paste the code you find here into the HTML code of your webpage, and the little icon that is displayed on that page with a link to my website will be displayed.
Last week this
reader added a link:
www.word4wordtranslation.com
© 2010 International Writers' Group | |
