Glossary
Automatic
Translation (Machine Translation)
General automatic translation is still
largely an utopia (although some CAT Tools are already interfaced
with MT programs, such as Power Translator). Except for some
very specific and limited niches, machine translation can currently
provide only a gist or general idea of the text, and this limitation – with all due respect
to Yves Champollion’s opinion (www.accurapid.com/journal/15mt.htm)
- is expected to remain for many years to come.
As Craciunescu, Gerding-Salas, and Stringer-O'Keeffe
pointed out in a recent article on the Translation Journal, the
term "Machine
Translation" is misleading, as MT provides a pre-translation,
a sort of raw material that a translator must completely rewrite.
I recently happened to accept an English>Italian "proofreading
job" that turned out to be an MT product. One of the gems
here was the translation of the term Adobe PostScript as "Mattone
crudo PostScript". Of course, I renegotiated the fee.
Babelfish (www.altavista.com) is an example
of a basic MT application. Quite often, the results can be quite
amusing. I tried to translate the Italian sentence "In genere si preferisce adoperare un
substrato molto spesso" (A very thick substrate is generally
preferred) and got "In kind a lot is preferred to often use
a substrate". A re-translation into Italian gave "In
natura mólto è preferito usare spesso un substrato".
However, more sophisticated MT tools are currently being used,
and a lot of research is being carried out to make them better
and better. Some automotive manufacturers apparently carry the
translation of their shop repair manuals using MT. It must be said,
however, that these manuals have been originally written using
a particular form of English, specially designed to be translated
using MT.
Ultimately, general MT tools may become more reliable in future,
and the translator might risk to become an editor or proofreader,
but not for a number of years yet.
CAT Tools
What is a CAT-Tool?
Automatic Translation (Machine Translation)
CAT is an abbreviation for "Computer Aided Translation".
A CAT Tool is a computer program which speeds up translation and
helps to ensure consistency (as far terminology and style are concerned). Although each CAT Tool works differently, most of them share
the following features:
Segmentation. The original files are divided
into segments according to a set of rules (such as splitting
the segment after a a question or exclamation mark, period,
or a semicolon, followed by a white space, unless the next
character is lowercase). These rules may fail sometimes (for
ex.: splitting "Dr. Smith said that …" into "Dr. " and " Smith
said that …") . This leads us to another feature,
the "Join/Split" function.
Join/Split. This function joins two adjacent segments, or splits
one segment into 2 segments. Joining and splitting may not always
be available.
Analysis. This function is used to get information on the degree
of repetiveness (D.O.R.) of one or more documents.
Propagation. A translated source-target pair (so called translation
unit, or TU) is propagated to all identical segments.
Translation memory (TM)
A database where all TU are stored.
Terminology database
A database where single words or word combinations (but not entire
TU's) can be stored.
Fuzzy search
This function allow to search a TM and find segments similar
but not identical (non-100% match). The degree of fuzziness
(D.O.F.) defines the degree of similarity.
Apart from these common features, CAT Tools can be classified
into 2 different categories:
- Direct Tools
- Indirect Tools.
Trados and Wordfast are 2 examples of the 1st category: translation
is carried out within the application (at least, as far as MS
Word is concerned).
On the contrary, DéjàVu and SDL have their own
table-like interface. The original documents are imported into
the interface, translated and finally exported back to their
original format.
At least with DéjàVu, this
feature offers a tremendous advantage: in the case of multi-file
projects, all the files can be imported and the translations
propagated to all files in the project at lighting speed (see
Propagation).
Ideal translation
The rationale behind the ideal translation is that "the translator
conveys ideas, not words". This is also called “Dynamic
Equivalence”. Rather than "translating literally" or
word-by-word, the translator interiorises the concepts in her/his
mind and attempts to convey the original message by reformulating
the text in the target language. Sometimes, this approach can lead
to paraphrasing, which may or may not be appropriate in the specific
context.
Since the complete meaning of a text is often deeper than the simple
apparent textual meaning, this will always be an insuperable obstacle
for a Machine Translation software.
Localization
It is much more than simple translation into another language.
For example, an icon may have a certain meaning in a given country
but mean something else, or even be offensive, in another country.
In case of a software localization, the
target language may be longer in the target language; therefore
a margin (30% in case of an En>It translation) must be planned
in advance; otherwise the poor translator will get crazy trying
to fit it all in.
Productivity Tools
Beneficial to both translators and their clients.
They include:
- CAT Tools
- Dictation software
- Search programs (serial lookup or indexing-based tools)
Translator
Evaluation Criteria
Many organizations find it difficult to identify qualified translators. How
to evaluate a translator? The following criteria sound reasonable to me.
|
| - Experience (years) |
| - |
References (clients and colleagues). However, some translators
might be unable to provide references, if their work is covered
by Non Disclosure Agreements, or if the organization requesting
the reference is a direct competitor of the organization asked
to provide the reference. |
| - Education |
| - |
Professional and Web activity, i.e. as posting
on translator's sites (such as www.proz.com; see forums and
KudoZ questions) and dedicated mailing lists (such as Langit),
or Journals. |
| - Timely delivery |
| - Absence of grammatical
errors, spelling errors, etc |
| - Lack of unjustified omissions |
| - Native fluency in the
target language |
| - A feel for current usage |
Transcreating
Neologism from "translation" + "creation".
Here the added value lies in the creativity of the target text.
The clearest examples are in the Fashion and Advertisement field.
|