1.
The Resource-Full
Translator (Premium Content)
| |
Last week, the Canadian government opened up TERMIUM Plus, the Translation Bureau's terminology database, with nearly four million
terms in English, French, and Spanish. My feeling is that most English <>
French translators have had access to that in some way or another for a while,
but now it's free and open to all. (The website claims rather exclusively that
"Canadians now have FREE access to TERMIUM Plus" -- but I did not have to scan my passport to gain
access.) I'm envious -- I wish we had such a fabulous resource for my language
combination.
Here is another resource that has turned from so-so to fabulous overnight:
The TAUS Data Association's search engine. Most of you will remember that I reported
on this in the past (and if you don't remember me talking about it, chances are
that you have heard about it elsewhere).
Here is a quick repeat for the small handful among you who are not aware
of what TAUS, the Translation Automation User Society, does:
The TAUS Data
Association or TDA is, like its name says, an association of mostly large
corporate translation buyers who originally came together to pool their
translation memory data to better train their machine translation engines. To
make statistical machine translation work, you need a lot of high-quality data,
and even industry giants like Oracle, Microsoft, and Adobe do not have enough
data on their own to get the results they hoped to achieve.
So they decided to
open the data up to the public -- not as TM data, mind you, but as a
terminology resource. If you want to get to the data as TM data, you can become
a TDA member, contribute your own data, and download some other data for your
own use.
But since this is
financially out of reach for many of us, we can at least use it as a
terminology resource.
This is where it stood awhile ago,
but there were some caveats that kept it from being particularly useful as a terminology
tool. The greatest weakness was that you could not see where the data came
from. And data without intelligent metadata is often noisy and does not help
much. Well, this has changed in the latest incarnation of the TDA search
engine. Aside from the obvious choices (language combination and broad industry
sectors), it is now also possible to filter the data by owner (originating
company) and content type (user documentation, software, websites, etc.) The
companies that so far have contributed to the total of almost one billion words
are ABBYY, Adobe, Avocent, Dell, eBay, EMC, Intel, McAfee, PTC, Sun, and
Sybase. Lionbridge, SDL, and Moravia have also contributed their own translated
website and marketing materials (as far as I can tell) and ProMT has
contributed training material from its machine translation engine (such as
idiomatic expressions). Other resources that are included are materials from
the United Nations and the European Union. (And here is one gripe: you can access
the EU materials only through a search from "British English," and
while this is of course technically correct, it becomes a nuisance when you
have to run several queries instead of one. It would be helpful to have the
regional variations as well as an overarching English, French, or Portuguese, as
for Arabic in the present version.) While this list is impressive, it is only a
small part of what you will eventually find in this database.
So, the first huge improvement is
the much more intelligent filtering -- and if you choose not to filter by
owner, you can still see where the data comes from by hovering over a little
information icon to the right of the language pair. Right next to that icon you
find an X icon that allows you to report problems, and also to the right is a
reverse icon if the translation originally was in the opposite direction.
The next improvement is that the
engine now includes some subsegmenting capabilities so that it is able to
identify the matching term in the target segment and highlight it. And not only
that, it will also list likely translated terms with a probability rating on
the top of the search window. Pretty cool.
There is one thing that I find maybe
even cooler, though: You can download a "widget," a little Java-based application that lets you do all
these searches right from your desktop. If you close the widget it will still
remember your last settings so you don't have to modify them again when you
reopen it, and the search is blazingly fast.
(The only thing that I still miss is
the ability to use IntelliWebSearch
to automatically copy the search term into the little Java application. I have
no doubt that Michael Farrell, the developer behind IntelliWebSearch,
will figure out a script to do this at some point. If you think that IntelliWebSearch
is a great tool but you feel a little overwhelmed by all those customizable
searches, Michael is offering online training courses,
and he has promised all Tool Kit readers a 20% discount if they enter 1D54BBAF
as the discount coupon.)
I was really struck this past week when I realized that I had done most
of my translations with the aid of resources like Linguee or the TAUS
Search rather than specialized dictionaries. It was not that my projects
last week were so generic, but both of these large-scale tools provide enough intelligent
information and data to make them highly usable even for very specialized
searches. Why is this relevant? Because I think that we have entered a new era of
data availability that is going to and has already changed the way we work. So
far we have talked about large-scale data accessibility as a more or less
theoretical fact -- it still took a lot of effort to gather it. Now there are suddenly
a number of goldmines at our fingertips. And that's new.
(Admittedly, I was lucky that my projects were
in the realm of software translation, an area that is still highly favored by
TAUS, and they went from English to German, the only language combination that Linguee
covers at this point, but both of these things will change.)
|
2. Dealing with Mixed
HTML and XML Content, Part II
| |
I knew it! There had to be a follow-up article on this. (Last time I
wrote about this, I had to have three follow-up articles.)
For those who missed the article last week, it was about the increasingly
ubiquitous XML format that contains encoded HTML code. You can download a copy
of the article here.
One comment came from Gábor Ugray (the developer who contributed the
-gray to Kilgray) who recommended PrepTags, a tool
developed by Galibert Sylvain, a French translator in Thailand. Here is how Galibert
described his tool:
How do you translate HTML, PHP, XML, ASP,
Javascript, SQL, PO, and many other formats? This is a regular headache for
translators and agencies alike. Ideally, a translator simply translates the
text to be translated without touching the code, using his favorite CAT
software, like Wordfast, Trados, Déjà Vu, MemoQ and
others.
This is done through file preparation. The file
is converted to RTF and the code is protected so the CAT tool will recognize
that it should not be translated. But how do you prepare files? This is where PrepTags
comes in.
I downloaded and tested the tool and it's helpful, very helpful indeed, and
it does just what you'll want it to do -- if you are well-versed with regular
expressions, that is. For everyone else it's rather technical, even
though the author tries to dissipate some of the fear in his help articles.
One particular problem I had with PrepTags is that it uses
"greedy" regular expressions, which means that it looks for the longest
possible solution in its searches. This means that if you look for anything
between < and > in this text <code>here</code> and
<code>there</code>, everything would be selected (<code>here</code> and
<code>there</code>).
In a "lazy" search, on the other hand, where the shortest possible
solution is searched for, only the four individual instances of the tags would
be located (<code>here</code>
and <code>there</code>).
This in turn means that you have to write rather specific regular expressions
in PrepTags.
MS Word uses
"lazy" regular expressions, a feature used by Tortoise Tagger, a free little
Word macro (I assume that it would not work with Word 2007, but I
am not sure) that allows you to define macros to format any text-based format
into an RTF document with Trados styles that most TEnTs ignore. (This is
exactly what PrepTags does, only that PrepTags allows for more
efficient batch-processing of many files at once.)
So, writing the regular expressions in Tortoise Tagger is
arguably easier than in PrepTags, but . . . when I wrote about Tortoise
before, I declared it to be a tool that is not for the faint of heart, and I
would certainly stand by that.
So, it's time for TEnTs to be able to handle files with mixed XML and
HTML content as a distinct format by using a pre-configured filter so we won't
need a Ph.D. in computer science to translate these files adequately.
Anyone out there? Anyone? (I promise some big-time coverage to the tool
vendor who develops a more-or-less preconfigured method of dealing with these
files!)
|
| ADVERTISEMENT |
Why pay a high price
for a CAT tool and keep fighting with it every day? Make your life
simpler and easier by making the right choice! Heartsome Translation
Studio is what you need and deserve:
·
based on open standards and cross-platform
·
the most customizable CAT tool in the market
·
gives you the choice to deliver in XLIFF and TMX or tagged RTF and TXT
TM formats
And as Jost puts it
in the Tool Kit: "There are many obvious benefits to Heartsome -- platform-independence,
a fairly small footprint on your system, a good choice of underlying database
servers . . . and a very pleasant 'feel.' . . . It is certainly an interesting
and powerful tool."
Enjoy the spirit of freedom. Make today your
independence day! Buy now at www.heartsome.net.
|
3. L15s
| |
I hope you noticed my clever heading: I expanded the site name L10NWorks
into its full form, LocalizationWorks, counted the characters between the first
and last letter, and added that back in as a number! Sounds sort of d6d
(deranged)? I know! So are all these numeronym constructions. In my opinion, "words"
like L10n, i18n, and G11n are among the saddest and most uncommunicative
"inventions" that our industry has come up with.
That said, L10NWorks calls
itself
a vendor-neutral
platform where localization professionals can share their expertise, as well as
tools and information with other peers in the industry.
China-based CSOFT's
engineers have created this site as a venue to bring together localization
engineers, translators, and industry stakeholders from around the world in an
attempt to grow as a community from each other's knowledge.
It comes in the form of a three-pronged portal. The first part is called
Localization 101, a collection of links and syndication of third-party
blogs dealing with translation, localization, and languages. Then there is a Forum
part, rather self-explanatory and with the potential to become really
interesting, but so far it's been very quiet. And lastly there is Tools.
This is what interested me the most.
There are some three dozen tools that deal with encoding, conversion,
quality assurance, extraction, comparison, and some language-related issues. There
is also a fairly lengthy list of links to open-source TEnTs, text editors, etc.
Some of the tools are open-source tools that are simply relisted and linked on the
site; others are other freeware tools, such as the ones from Maxprograms; but most have actually been
developed by CSOFT's staff. You can download some of the internally developed
tools, but the more typical method is to upload your files (for instance, in
the case of the many QA tools, you are asked to upload your translation
memories in Trados, Déjà Vu, SDLX, or Idiom WorldServer
format) to be processed. You can then download the corrected file as well as a
log file containing a list of the fixes. I tried that with several files but
ran into a number of errors -- but those may simply be early gaffes. Either way
I would advise CSOFT to find another way to handle these QA and conversion
procedures since folks will not be willing to upload their TMs to a server of
an unknown third-party company willy-nilly.
Some of the features of the L10NWorks tools can be carried out by
tools that will most likely be in your arsenal of tools anyway (such as
character set detection or codepage conversions -- all things that are done by
text editors -- or certain file format conversions). However, other applications
are handy, such as the HTML entity converter (which gives you the HTML code for
special characters), the helpful list of open-source TEnTs, the tools from
Maxprograms (XLIFF or TMX validator), or tools like the Xliff RoundTrip Tool,
which allows you to convert an XML file to the translation format XLIFF and
back. All in all, it's a nice collection of some tools, some of which might
need some better strategic planning, but that may still happen.
|
4. Google Translator Toolkit, the Next Incarnation (Premium Content)
| |
Much has been said about the Google Translator Toolkit in
this and other publications, and almost everyone has remarked on the odd fact
that English was the only supported source language, with a rather small number
of languages supported as target languages. We all expected this to change,
and, boy, has it ever. Yesterday, Google released a new version with 37 source languages (all the ones you would
expect, plus possibly less-expected languages like Belarusian, Croatian, and Yiddish
-- please don't send letters protesting that I think less of these languages --
I don't) and more than 400 target languages and regional variations (as an
example here are some "K" entries: Kabyle, Kachin, Kalaallisut,
Kalmyk, Kannada, Kanuri, Kara-Kalpak, Karachay-Balkar, Karelian, Karen,
Kashmiri, Kashubian, Kazakh, Khasi, Khmer, Kikuyu, Kimbundu, and Kinyarwanda). The
interface now is also translated into 36 languages (you can find a complete
list of languages right here). What does all this mean? For one thing, it means that Google is
relatively serious about its efforts with this, despite the fact that it has
not caught on nearly the way they had hoped (but this is just my subjective
impression based on occasional searches in the public TMs). It also means that
some languages that most of us have never heard of have some kind of support.
This is what the official Google blog posting says about this:
At Google, we're
focusing on how Translator Toolkit can help preserve and revitalize
small and minority languages. Minority languages, also called regional,
indigenous, heritage or threatened languages, are languages spoken by the
minority people in one locale in a sovereign state or country. Were these
endangered languages to become extinct, it would mean an immeasurable loss of
knowledge, culture and way of life to minority people worldwide.
For this project we
worked with Dr. Te Taka Keegan, a Māori language activist and senior lecturer
in computer science at the University of Waikato who spent much of his career
on how technology can assist in minority language revitalization. (...)
Dr. Keegan found
that tools such as Translator Toolkit can help minority languages in several
ways:
·
Translation memories and glossaries, when shared across members of a
language community, can help unify the language's written form, increasing
translation speed and quality of documents published in that language and
preserving the language in the long run.
·
Because computer-aided translation can improve translation speed and
quality, translators become more productive. When automatic translation is
available, as it is for 87 of Google Translator Toolkit's 345 languages,
it increases speed further by producing instant translations that people can
use as a starting point for their work. And at Google, we use these human
translations to improve the translation algorithm of Google Translate
over time, creating a virtuous cycle that benefits both human translators and
machine translation.
·
Online presence of small languages keeps languages relevant in the age
of the Internet and globalization, encouraging minority language use by
children, who are ultimately responsible for bringing the language to future
generations.
Languages provide
identity, pride, a sense of belonging and spiritual guidance to minority
language communities. We hope that by giving both majority and minority
language speakers around the world the tools to make online content accessible
in their language, we will enable more people to share their culture and
knowledge with others worldwide.
I don't think you can find much wrong with those statements. What you
can also see, though, is that for the present, many of the languages supported
are just as much or as little supported as they would be by any other TEnT. There
are no existing translation memories or machine translation engines (only for
87 of the target languages), but the difference is of course that
(theoretically) the public translation memories will grow over time and will at
some point then help the machine translation efforts also.
You can't say that Google is not open about that
goal -- to use our translations to better its own machine translation -- and
that is something that you need to be aware of when you use this tool.
|
5. Just Released!
| |
This morning I received this note from the maker of ApSIC Xbench:
Just a quick mail to let
you know that we released an update to ApSIC Xbench. The biggest
new feature is probably the new support for regular expressions (or Microsoft
Word wildcards if you prefer), which we think will be useful for QA, as it
should provide a lot more flexibility and accuracy for checklist items. Also
the search engine in v2.8 is faster than in v.2.7 (a pleasant side effect of
having to look again into the search engine algorithms to add support for
regular expressions).
I did not have time to look at this latest release, but I have written
about Xbench in the past:
Originally developed
as a tool to index and search the so-called Microsoft glossaries, this contained
translation memories of the translation of the user interface of most MS
products. (These are now accessible only through a paid membership of MSDN.) It
now supports a huge number of bilingual file formats, including tab-delimited
text files, TMX memories, Trados text TMs, Trados MultiTerm
XML files, Trados .ttx files, uncleaned Trados Word files,
SDLX .itd files, IBM Translation Manager files, Star
Transit files, Wordfast TMs and glossaries, some XLIFF files, TBX
files, and the Mac OS X glossaries. (By the way, these last glossaries can
be downloaded here
-- you do need a Mac to get the .ad files out of the .dmg image.)
You will clearly
note that there are two different kinds of bilingual files in the list above:
reference files (TMs and glossaries) and actual translation work files. The
reason for this is that Xbench not only allows you to search reference
files from within any Windows application (just highlight a word and
press the Ctrl+Alt+Insert key
combination and within milliseconds the tool searches potentially hundreds of
thousands of lines and shows you the appropriate matches), but it also allows
you to import Transit, SDLX, IBM TM, or Trados
files to perform QA checks such as consistent usage of terminology,
non-translated segments, numeric consistency, etc.
It's a very well liked tool by its users. Though I mostly use it for its
glossary features, I know that many -- especially project managers -- swear by
its quality assurance features. And I did mention that it's free, right?
Another tool that was newly released and that I have not yet had a
chance to look at is the 2.0 version of the open-source TEnT OmegaT.
Here are some of its new features (I will do my best to review them fully
for the next newsletter):
- Loading and
indexing processes have been "completely rewritten" to allow for
larger projects and "large" translation memories (the sample in
the announcement was 63 MB, 20,000 entries -- I am not sure that this
would be considered a large TM inother contexts).
- Better support
for right-to-left languages
- Stemming is
available for a number of languages to improve search results
- Google
Translate can optionally be
used to machine-pretranslate
There are various other new features as well, but these are the ones
that stood out to me.
SDL released its Service Pack 1 for Trados Studio last
week -- no doubt you have heard about it if you are using Trados Studio.
Just one hint: Don't think that you can do the installation during a quick
coffee break -- it takes its own sweet time.
Madeleine Lenker pointed me to another software release, perhaps not quite as current but certainly
equally or more important, is GlossMaster.
This is a highly advanced terminology tool to build up terminology for the 2500
most common terms used in open-source software in Afrikaans, Akan, Amharic,
Kinyarwanda, Lingala, Luganda, Mandinka, Somali, Songhay, Swahili, Wolof,
Yoruba, and Zulu. It's worthwhile to have a look at this tool even if you don't
translate into any of these languages. It demonstrates a love for detail and
accuracy that matches or surpasses the most complex corporate tool setup.
A few weeks ago I was talking to a representative from MultiCorpora, the
makers of MultiTrans, who are supplying the official translation tool to
the African Union. I thought it was ironic that the only languages requested by
the African Union are European languages and Arabic, with Swahili in the
pipeline in the years to come.
It's good that there are a number of projects which
focus specifically on the localization of African languages. These include ANLOC, the African Network for
Localization (this is the project that GlossMaster has grown out of), and
Translate.org.za, a non-profit organization
focused on the localization of open-source software into the languages of South
Africa. The latter has released and maintains tools such as Pootle, Virtaal and the Translate
Toolkit, tools that are well known in the world of open-source translation.
|
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. © 2009 International
Writers' Group
| |
