38
Computer-Assisted Translation Tools
The Translator’s Tool Box - © International Writers’ Group, LLC 283
Integration of Machine Translation
The relationship between professional translators and machine translation is
uneasy at best. Between the perception that machine translation is a potential
job killer and the widespread disdain for post-editing of sub-par machine
translation output, many translators wish they had never had to deal with it in
the first place. While this book is not the right place to argue the pros and
cons of machine translation in its many shapes and forms, I would at least like
to look at the different ways that machine translation is integrated in today’s
translation environment tools and what the usefulness of that integration
might be.
First I would like quote Jaap van der Meer (in MultiLingual 71, 2005), a strong
proponent of machine translation:
Disdain on the side of the professional translators for the hilarious and stupid MT
mistakes gave birth to a new variant of MT called translation memory (TM). TM
started off as a lower-level feature of commercial MT systems (...). But the
success of TM came with dedicated products such as IBM TM/2 and Trados. The
marketing message was tuned in to what the professional translation industry
wanted to hear: "Forget about MT; it doesn’t work well. Instead, use our TM
product because it leaves you in full control of the process."
The message worked well. Within a period of 10 to 15 years, TM products have
found their way to the workstations of more than 50,000 translators in the world.
But the message had also caused a "cognitive disorder" in the translation industry,
namely that TM is good and MT is evil, foregoing the fact that TM is just a new
variant of MT (...). The damage is done, however, and it will take years to convince
the community of business translators that post-editing fuzzy matches from TM
databases is, in fact, not different from post-editing fuzzy matches from any other
MT system.
Mind you, I don’t completely agree with everything van der Meer states. I
particularly disagree with his assertion "that post-editing fuzzy matches from
TM databases is, in fact, not different from post-editing fuzzy matches from
any other MT system." There is in fact a fundamental difference between the
work of post-editing MT translations and fuzzy matches, and often this is not
fairly represented in the MT community. Provided that your translation
memory is in good shape, editing a fuzzy match means altering an inherently
correct segment (correct as a translation for the earlier source segment) to
match your current source segment. Typically this involves changing a couple
34
Computer-Assisted Translation Tools
284
The Translator’s Tool Box - © International Writers’ Group, LLC
of terms, which can be done easily. This is not necessarily so with MT, though,
which is not inherently correct. It can be, but it does not have to be. If you
work in my language combination (English>German), you will quickly find
that more often than not there are fundamental changes you will need to
make to bring the translation to the required quality level.
Still, van der Meer points out some uneasy truths: translation memory is in
fact a lower-level feature of machine translation; it is used as such within
machine translation programs to the present day; and, yes, we have certainly
been influenced by the marketing van der Meer points to.
Up through 2011, most TEnTs came with only a connector to Google
Translate, but Google’s decision to start charging for its machine translation
service (if it’s integrated into a third-party tool) made most tool vendors look
for other solutions alongside Google Translate, resulting in a very large
number of easily accessible possibilities.
Here are the different offerings of most TEnTs (as of November 2015; please
note that to actually use most of these MT tools you’ll need a license key).
TEnT
Integrated MT Engines
Across
Google Translate, Reverso, Moses, Asia Online
Cafetran
Google Translate, Microsoft Translator, MyMemory
Déjà Vu
Google Translate, Microsoft Translator, itranslate4.eu, PROMT,
SYSTRAN, Asia Online, MyMemory
Fluency
Google Translate, Microsoft Translator, SYSTRAN, MyMemory
Google
Translator
Toolkit
Google Translate
34
Computer-Assisted Translation Tools
The Translator’s Tool Box - © International Writers’ Group, LLC 285
Heartsome
Google Translate, Microsoft Translator
Lilt
MT engine developed by Lilt/University of Stanford
MateCat
Google Translate, Microsoft Translator, SYSTRAN, MT engine
developed by Translated
memoQ
Google Translate, Microsoft Translator, itranslate4.eu, SYSTRAN,
LetsMT!, Asia Online, MyMemory, KantanMT, CrossWay Gateway
Memsource
Google Translate, Microsoft Translator, SYSTRAN, tauyou,
Apertium, Asia Online, KantanMT, LetsMT, PangeanMT, Safaba
Multitrans
Google Translate, Microsoft Translator, PROMT, SYSTRAN,
MyMemory
OmegaT
Google Translate, Microsoft Translator, Apertium, Belazar
SmartCAT
ABBYY-internal MT, Google Translate, Microsoft Translator,
Yandex
Swordfish
Google Translate, Microsoft Translator, Yandex
Trados Studio
Google Translate, Microsoft Translator, itranslate4.eu, SYSTRAN,
SDL Language Cloud (formerly BeGlobal and LanguageWeaver),
SDL MT, MyMemory, Sunda Translator, Tilde, taoyou, Iconic,
KantanMT (all of these, with the exception of SDL Language
Cloud, have to be downloaded as separate apps from the SDL
OpenExchange (see www.translationzone.com/openexchange)
TEnT
Integrated MT Engines
34
Computer-Assisted Translation Tools
286
The Translator’s Tool Box - © International Writers’ Group, LLC
You can find information on these different MT engines and the language
combinations they cover on their respective websites, but two might be worth
highlighting because of their different nature:
• MyMemory is a mixture of a large collection of translation memories,
aligned, web-based data and MT via Google Translate, Microsoft Translator,
a customized Moses engine and Systran. The data is hosted and curated by
the Italian translation and technology provider Translated (see
mymemory.translated.net).
• itranslate4.eu is a European consortium of nine different machine
translation vendors: MorphoLogic, SYSTRAN, Linguatec, pwn.pl, SkyCode,
Amebis, Sunda Systems Oy, PROMT and Trident MT.
Are the many integrated MT engines helpful and are they used by professional
translators? I will leave the answer to the first question up to your preferences
(and language combination, and kinds of translation you do, and the many
likes and dislikes that you might have about this kind of technology), but the
answer to the second question is, yes, more and more translators are using
MT as one of many resources.
Wordbee
Google Translate, Microsoft Translator, Reverso
Wordfast Classic
Google Translate, Microsoft Translator, itranslate4.eu,
WorldLingo, MyMemory
Wordfast Pro
Google Translate, Microsoft Translator, WorldLingo
Wordfast
Anywhere
Google Translate, Microsoft Translator, WorldLingo
XTM
Google Translate, Asia Online
TEnT
Integrated MT Engines
VB.NET PDF - Annotate PDF with WPF PDF Viewer for VB.NET Text box. Click to add a text box to specific location on PDF page. Line color and fill can be set in properties. Copyright © <2000-2016> by <RasterEdge.com>.
how to make pdf editable form reader; how to flatten a pdf form in reader
25
Computer-Assisted Translation Tools
The Translator’s Tool Box - © International Writers’ Group, LLC 287
Some are using it if no quality TM matches are found (the tools can typically
be adjusted so that machine translation is pursued only if no match of, say,
75% or higher can be found), others are using MT as an extended dictionary
for highly specialized terms, and yet others are using it as a source for a
variety of suggestions.
Consider this example from Wordfast Classic (with machine translations from
Google Translate, Microsoft Translator, Linguatec, Systran and Trident MT—
the last three through itranslate4.eu ).
Figure 192: Wordfast Classic with machine translation matches
There is no need to argue about how "good" these matches are, but most of
them contain some material that in some kind of combination might be useful
in the actual and final translation. You as a translator will have to decide what
kind of role this information plays for you. Does it help or hinder? Is it
different, for instance, than having a lot of matches from a general TM shown?
The answer to that will most certainly depend on your language combination
(some language combinations are much more suited to a first machine
translation draft than others) as well as your project type or subject matter. It
might even be different between different projects.
Lilt’s (see page 259) approach toward machine translation might be one way
to view the productive use of MT in the (near) future: A combination of auto-
suggest with a constantly updated stream of suggestions from the MT engine
based on what you have already entered might deliver a real productivity
breakthrough.
22
Computer-Assisted Translation Tools
288
The Translator’s Tool Box - © International Writers’ Group, LLC
Quality Assurance Features
We’re at a point in the development of translation environment tools when it’s
generally understood that translation memory tools are not primarily time-
saving applications; first and foremost they are programs that allow
translators to increase the quality and consistency of their work. During the
translation process this happens (or should happen) with the use of
terminology databases and translation memories that give on-the-fly access
to previously used translation or to terminology that the client requires for the
current project.
Beyond that, virtually all translation environment tools offer quality assurance
features such as spell-checks or checks for formatting integrity. In fact, tool
vendors have recognized only rather recently that there is demand for more
far-reaching quality assurance features.
Though many think of Wordfast Classic as a rather lightweight tool, it offered
advanced quality assurance features long before most other tools did.
Wordfast Classic offers a total of about 15 quality assurance checks, starting
with the verification whether the same translated term or phrase was used in
the terminology database.
Figure 193: Setting which terminology database is to be used for Wordfast’s terminology check
12
Computer-Assisted Translation Tools
The Translator’s Tool Box - © International Writers’ Group, LLC 289
The now defunct SDLX offered its quality assurance checks as the major
improvement when it released its version 2005. At that point it was probably
the most comprehensive solution.
Figure 194: The QA Check module in SDLX: the Inconsistencies tab
It’s important to realize that this feature is not equally useful in all languages.
Terminology checks in languages with heavy conjugation or declination, or
agglutinative languages such as Turkish or Finnish, will typically find a lot of
"translation errors" that are really just different forms of the correct term. A
strategy to counter that is to enter various term pairs to cover the different word
forms.
14
Computer-Assisted Translation Tools
290
The Translator’s Tool Box - © International Writers’ Group, LLC
However, Trados (whose owner also owned SDLX) versions 7.1 and above
included a larger set of QA features than any of its competitors.
Figure 195: The QA Checker module in Trados Studio
One particularly helpful aspect of the Trados QA Checker is the fact that you
can load and save a profile (under QA Check Profiles), enabling all members
of one translation team to use the same QA procedures.
In a sense it is not a surprise that Trados offers such encompassing QA
features. For a number of years, several tool vendors have been offering a
variety of quality assurance tools that specifically provide quality assurance
for Trados files. So when Trados got ready to offer it themselves, they could
just pick and choose the most helpful features.
19
Computer-Assisted Translation Tools
The Translator’s Tool Box - © International Writers’ Group, LLC 291
The only tool to rival Trados’ QA features at the moment is memoQ:
Figure 196: A sampling of memoQ’s quality checks
For even more comprehensive standalone quality assurance tools, see page
300.
Quality Assessment
In 2011 and 2012 respectively, two different organizations started to work on
creating quality assurance models that can be used to evaluate translation:
TAUS (see taus.net) worked on the DQF or Dynamic Quality Framework, and
the EU-funded QTLaunchPad (see qt21.eu/launchpad) created the MQM, or
Multidimensional Quality Metrics. Both of these models are remarkably similar,
so similar, in fact, that they are now being merged into one model. The
similarity between the two is not a coincidence. Both organizations looked at
quality assessments that were used throughout translation and localization
and distilled them into one large scheme.
For both of these efforts, the main goal was to provide a framework to assess
the quality of machine translation. The models are helpful in that regard, but
they also provide relevant and interesting tools to assess "human" translation.
18
Computer-Assisted Translation Tools
292
The Translator’s Tool Box - © International Writers’ Group, LLC
As translators, we know that quality assessment is an important task, but one
that is often contentious and fraught with difficulty. Traditional methods were
highly subjective and involved reviewers reading translated texts and marking
"errors," but reviewers often disagreed on their assessments. In response,
many translation companies invested in developing formalized metrics for
assigning errors to different types (e.g., terminology, spelling,
mistranslations), counting them, and determining how serious they were. In
the 1990s, these efforts led to the creation of widely used specifications such
as SAE J2450 (mainly used for the automobile industry) and the LISA QA
Model. Unfortunately, these models have not been updated over time and are
presented as "one-size-fits-all" models were do not reflect the needs of a
rapidly diversifying translation industry. This is where MQM and DQF come into
play. Here it’s possible to mold the quality assessment frameworks into text
type-specific models in which the reviewer decides which categories are
included and what severity is assigned to the various categories.
Documents you may be interested
Documents you may be interested
