translation, also known as CAT tools, is a translation programme that allows a
translator to use computer software to help them with translations.
Déjà vu X3 is an example of a CAT
tool that aids translators when translating certain documents, it is especially
useful for texts that will be modified over time, for example, car manuals due
to the specific features the programme uses for example, term bases,
translation memories and other automatic functions.
Term base (TB)
term base is simple, I explored features of creating the terms in excel then
uploading the terms straight to the empty term base directly from excel, making
the process quick and hassle free.
A term base is a file that stores terms from the text
that is being translated and, much like the translation memory, is usually in
multilingual format. It also shows additional information such as Parts of
speech, Gender, Number and context.
Here is an
example of what a standard term entry looks like when editing.
With the term
base, I added a definition section as some terms were difficult to interpret.
For example, La Rada Suprema didn’t
have a literal translation so under the definition section, I explained that,
from research is seemed it was a government in Ukraine, making it useful when
it came to translating the text. It was also useful for any Hungarian that was
in the text, or places in Ukraine or Hungary that I had not heard of, so that I
knew why they had not been translated.
creating the term base, I struggled to locate specific terms in the text so I attempted
to use the lexicon feature in the project. The project lexicon shows a list of
all the phrases or words in the translation. However, this created too many
terms and I found it difficult to narrow down to about 6 to 8 terms so I abandoned
translating, the term is already in the memory and translates automatically
when working on the segment the term is in, making sure the translation is an
exact match as it already recognises the term. As well as this, it also
identifies terms in the segment under the one being worked on and will
automatically translate the term ready for the translator to use. This supports
the task by ensuring the term is used effectively in the segment and that there
are not different translations throughout the text.
Translation Memory (TM)
memory is a file that stores the segments from the translation that have
already been translated. The translation memory keeps the source text and the
target language in translation units which makes it clear for the human
translator to use.
In déjà vu x3, the translation memory is multilingual and
multidirectional. Meaning that you can change the source and the target texts,
see the TM from start to finish and filter the text.
to set up filters in the translation memory however I did not have extensive
knowledge of SQL statements and therefore this was difficult to do. It is very
easy to add a new segment into the translation memory window through the edit
tab which is similar to the term base however you can only see one target
segment at a time in the translation memory. An extra feature that can be very
useful to translators is that the translation interface can sort translation memories
alphabetically, making it clear for the translator to see the TM they are
translation memory helps mostly when modifying a translation. Due to the
translation memory storing certain segments, if they are then in another
translation, for example the second modified text, it will identify if there
are certain segments in the new translation that are the same. Because of this,
the translation memory only supports the translation task when translating the
second text and is not as effective as the term base for the translation task.
I added the
translation memory after I had translated the first text. I did this by firstly
creating a blank translation memory, then under project, there is a button that
is called ‘add to translation memory’ and this puts the whole text into the
translation memory. This ensured that the segments in the translation memory
were the correct translations making it very useful when translating the second
merging the translation memory with my partners, the newly merged translation
memory would not open automatically and we would have go to ‘tools’ and ‘repair’
so it would open which lead to déjà vu crashing multiple times and slowed down
the whole translation process.
Project/ Editing environment
Déjà vu x3 has a row of tabs that are similar to many Microsoft
Office programmes. It has File, Home, Project, Lexicon, Insert, View and
Review. This makes it useful for translators who often use Microsoft office programmes
as the similarities will make accessing certain features easier. This is due to
the smart view feature which allows everything to be easily accessible like a
word document, making it quick to import, export, pre-translate, create new
projects or translation memories and term bases.
It is useful
that you can open multiple documents on Déjà vu x3 so that I could work on the
project, term base and translation memory at once. It also allows you to work
on different projects that use the same term bases and translation memories
which was very useful when working on the modified text.
The main body
of the text if very clear to see, with the source text on the left and the
target text on the right with the text split up into segments. I was surprised
to see that some of the segments were not split by sentences but more so in the
middle of sentences, this means that I had to be aware of the grammar I was
using in each segment so that the sentence would make sense as a whole.
importing the second text to déjà vu, I looked at the features available when
pre-translating the new text. At first, I only selected ‘repair fuzzy matches’
and lock guaranteed matches, which only pre-translated a few segments. This was
because the second text contained a lot of different sentences and a different
structure to the first text. However, I then also selected ‘assemble from
portions’ and ‘use DeepMiner statistical extraction’ as this attempts to
translate from other fragments in both texts. Deepminer takes information from the translation project and the
translation memory database to find terms and sentences that it thinks will be
useful in your empty segments and this will then often repair fuzzy matches and
will make the translation even more precise
Here is the pre-translated text using the ‘DeepMiner’
selection, all the fragments that are in blue and are underlined were
translated through this selection, as you can see it does not translate
sentences but more so single words or phrases as well as terms from the term
base. Unfortunately, this was not as effective as I would have liked as when I
went to then translate the segments, unless I used the exact words from the
DeepMiner settings, a little red exclamation mark came up to say the
translation was not correct. And in some segments the words needed to change to
fit the context and the sentence structure.
I then tried
to use auto propagate as this tends to translate all segments that are the same
as the one that has just been translated and will never propagate into a fuzzy match,
however because the modified text was so different this also did not help with
the blank segments.
matches and exact matches were mainly in the second text, especially the fuzzy
matches. This is due to the work of the translation memory. I found it very
difficult to repair the fuzzy matches I had in the modified text. One segment
in particular was very interesting to try and repair as it showed that the segment
was only 84% translated and even when certain words were changed in the
segment, there was still a 16% fuzzy match. Fuzzy matches generally repair with
the help of the translation memory as it will delete the incorrect part of the
translation and suggest a better translation from the memory
did not happen with the fuzzy matches, even when I pre-translated the text
using the DeepMiner feature.
of Déjà vu X3
a feature that will automatically suggest terms or phrases with the help of the
term base and translation memory that it deems suitable for the segment you are
working on. This often happens when translating and is seen above the cursor in
a small, white text box and when pressing enter, the phrase or sentence auto
write has suggested, will automatically enter the segment
I found that AutoWrite was one of the best features when
working on the text as the suggestions made the process easier and quicker to
will check to see if the segments you have translated are correct. This is done
by pressing control then the down arrow on the keyboard. At times, you can get
inconsistent translations that are shown through exclamation marks. However, I found
that the DeepMiner setting showed that most of my segments that DeepMiner had
translated parts of, created inconsistent translations because in some segments
I did not use the specific words or phrases even though the translation was
correct. Therefore, I had disable auto check, which could hinder the accuracy
of the rest of the translation.
Format tags are often hyperlinks that link to a photo or
web page that usually add extra context to the text however déjà vu x3 does not
recognise format tags when translating.
It is very
important in déjà vu x3 to keep the tags in the same place that they are in the
sentence from the source text, this is because it could hinder the structure of
the translation and the tags would then be misplaced when viewing the exported
file, creating, possibly and unreadable translation. Therefore, it is vital to
check tags that have been placed, this can be done by going to ‘review’ and ‘check
tags’ or there is also the shortcut of shift and f8.
Even though I
did not have any tags in my first translation, I ensured that I understood how
to use them before the second translation in case there were some in that
section, however, there were none in the second translation also.
The state of the
segments is indicated by the coloured bars shown to the left of the translated
text. A list of what all the colours indicate can be seen at the top of the
project in a drop-down menu. Ones that were frequently used in my translation
were dark green, light green, orange, dark blue and the black tick.
Here is a
screenshot of the segment statuses and what each colour indicates.
To conclude, Déjà vu X3 is very
useful for translators who want to translate texts quickly and efficiently
especially through the help of the term base and the translation memory. As well
as this, other features such as auto write or pre-translate are useful for
modified texts and even though it is quick to open or load projects, the
programme is prone to crashing and I often had to use the repair tool with the
merged translation memory. Overall the translation process was easy to do due
to the features that would not be possible for a human translator to do in a
short period of time.