Volume 4, No. 2 
April 2000

  Suzanne Falcone



An Amazing Tribe
by Gabe Bokor
Index 1997-2000
  Translator Profiles
Reflections on a Translator’s Life
by Susanna Greiss
  The Profession
The Bottom Line
by Fire Ant & Worker Bee
Sorry Guys, You Can't Win
by Danilo Nogueira
Un Secreto Bien Guardado
by Daniela Camozzi y Daniela Rodrigues Gesualdi
  Translators and Computers
XML and the Translator
by Alan K. Melby, Ph.D.
  Genealogical Translation
Translating for the German Genealogy Market
by Ann C. Sherwin
Lexicographical considerations in creating an online bilingual lexicon for students from a Chinese background
by Christopher Greaves and Han Yang, Ph.D.
  Science & Technology
A Translator’s Guide to Organic Chemical Nomenclature XIX
by Chester E. Claff, Jr., Ph.D.
  Caught in the Web
Web Surfing for Fun and Profit
by Cathy Flick, Ph.D.
Translators’ On-Line Resources
by Gabe Bokor
  Translators’ Tools
More Translation Memory Tools
by Suzanne Assénat-Falcone
Translators’ Emporium
Translators’ Events
Call for Papers and Editorial Policies
Translation Journal
Translation Aid

More Translation Memory Tools

(not many more, but good ones)

by Suzanne Assénat-Falcone
fter using "the client's software" for a number of years, i.e. sublicensed programs such as IBM TM/2, Star Transit or Alpnet's proprietary TSS/Joust, we finally decided to get our own. Following recent and alluring offers by TM providers, we are now the proud owners not of one, but of TWO other programs. And the good news (for us) is that those two are among the best three programs I have tried (Joust is available only for Alpnet's contractors). Among other good things, those "client-independent" programs provide really user-friendly translation memory and contextual search features, import any files you ask them to and re-export them as ready-to-go documents with all formatting and graphics in place. Maybe the other programs can do that too, but all I've ever had to deliver with them was a project to deliver "as is," and I have always had mixed feelings about their memory.

A word of warning: again, do not expect an expert opinion here. This is just a few comments jotted down by the end user, i.e. a stressed-out translator with two little kids and an obnoxious cat who may not have the time to explore all options in her programs. And of course, giving a personal opinion may mean that I am forgetting a lot of things and exaggerating a lot of others.

Atril DéjàVu: very, very good for a "classic" design

DéjàVu (V2.3, http://www.atril.com) is still designed according to the traditional tabular form of MT programs (see Michael Benis' explanations about this choice in this Journal).

An advantage of DéjàVu for translators is that it can process projects from other TM tools. This can be very useful when working for several clients using several different TM programs. In theory, DVI can cope with Trados files and IBM TM/2 untranslated segments (which seems to require a number of manipulations, according to more experienced colleagues).

DVI can import a number of other formats, including Word, RTF, HTML and plain text, but also PowerPoint, FrameMaker, PageMaker, Quark Xpress, help files (.cnt), C++ etc. So far I can confirm that it works like a breeze with text (including RTF and HTML), but I have been unable to open PowerPoint 95 (it turns out that DV will only accept PowerPoint 97 and later versions—the problem seems to be the same with Trados).

With the formats that have tested successfully, DVI provides an excellent end result as all the formatting of the source text is restored in the target file. Using DéjàVu for Web pages also produces excellent results. Basically, the major difference from Word is the hyperlinks, and these are translated properly, i.e. click here for more information will be translated, but not the "underlying" URL. Combining DéjàVu with FrontPage turned out to be a highly productive solution to translate HTML files.

Creating a memory database is easy when you have at least one source file and the corresponding translation. When you just get started on your own, you may find it takes a little time to chop up both texts into sentences/segments and make a two-column table of source and target as a text file. But then one day, someone points out that the Database Maintenance feature includes an "Align" option that allows importing source and target documents as plain text files and aligning them as pairs of source/target segments. It also takes some time, but not half as much!

Similarly, the terminology base can be enriched either manually with single items, or using the same two columns format (plain text). A convenient terminology check function is also provided, for translators who keep forgetting their terms or, like the undersigned, don't want to clog their display with the term base window.

As for adding to these two databases when working in DéjàVu, it's also a fast and easy process. Adding a single segment to the memory database requires one stroke of two keys, while a term can be singled out (double click) in both source and target and added to the terminology database. These functions are also available through a right click, which displays a nice long menu, including memory features and tag processing. There is also an "AutoSend" function to send all segments to the memory database by pressing Ctrl+(. (But why do you have to specially remember to enable it?)

However, updating translation memories after finishing the job may not be very convenient, according to the way you work. Translation memory programs seem to be designed with the idea that you also want to revise your text in the program. Because this is hard on the eyes and prevents the translator from thinking beyond the limits of individual segments, it is definitely better to print out or export the translation to revise it. Then you can get back to the Terminology Maintenance tool and realign your source and target. That's a bit tedious again, but it could be worse. Interestingly, the translation memory can be edited like any table (the add/suppress function is similar to Excel's add/suppress rows), so that you can clear it of any text that won't be needed in the future.

Those who didn't have the time to discover AutoSend had to remember to store their translation into the memory base after completing each segment (Ctrl+F12, easy when you remember to use it). But when you have a bird's brain like mine (well, I don't mean you, just a general "you," of course), you can make your way through a page and suddenly bump into a sentence you did a moment ago and didn't remember to add—then you have to get back up the text, find the segment, add it and return where you were. You may also decide to continue, in which case, obviously, your next translation of the same sentence may be slightly different. Not to worry—there is an interesting function to find duplicate source segments with different translations. But when you know better, you certainly prefer AutoSend.

Once you have assembled you project, i.e. put together a translation memory, a terminology database and your source texts, all you have to worry about is translating successive segments. To begin with, you can use the pretranslation function, which does a good job on a whole. Michael mentions an AutoProcessing function that basically assembles translations for you (see his article for further detail). Pity the corresponding "Edit/AutoProcessing enabled" menu entry mentioned in the help file doesn't appear in my version. Maybe they mean AutoText? But then the Help file says F2, and the menu says F8... Until I have found out, I still have the "Pretranslate" feature to work on the remaining text, and the easy Alt+A keystroke to translate individual segments.

Navigating in DéjàVu is not very easy and requires a lot of mouse work until you find out how to jump to the next segment (Alt+¯ ). If you don't find this solution accidentally (or if you have just been informed that there IS a manual on the Atril Web site, after working with DV for a year, like me!), you will have to double click on successive segments, which, as pointed out by others, is eventually painful for the wrist and thumb, and is also not particularly fast. However, there are plenty of keyboard shortcuts, although they could bundle more functions. In a recent discussion in the Foreign Language forum, one of the forum members proposed "an interim solution" to the lack of convenient keyboard shortcuts in DV: a keyboard programming software called KeyGo (http://www.gdgsysteme.com) that can be used to combine a bunch of functions into one keystroke in DéjàVu (heartfelt thanks to Tony Roder for the tip—but Atril has also referred me to the appendix of the User's Manual for more keystrokes in "plain" DéjàVu).

Looking at the really bright side of things, DéjàVu handles two of my pet peeves with TM tools relatively well: format tags and table headings.

The format of the source text is "registered" as tags which appear as a number in brackets, e.g. {1235}. These tags are quite inobtrusive, and they are updated automatically in full matches, i.e. the number will change in the next repeat (tags are numbered sequentially throughout a project). All tags are checked when exporting the file, so don't forget any or you're in for some boring time. The only snag is that they are not inserted automatically: all it takes is a right mouse click, but you have to remember to do it.

The other problem is the way TM programs handle table headings, because the order of the terms isn't the same in all languages (Maximum pressure = pression maximale), so when a table is chopped into individual terms there can be some disorder in the end. Therefore it would be interesting to keep the headings whole. DVI provides two ways to help solve this problem:

(1) Segments can be split or joined, which is useful if table headings are split up into successive segments:










can become:







You can then join successive segments into coherent items, i.e. "Maximum pressure" in one segment instead of two, etc.

(2) The source text is accessible for modifications. Imagine the table above is imported as follows:







You can then go to the "Maximum" segment and type "pressure" in the source window, delete the corresponding "pressure" and join the empty segment with the previous one.

So, all in all, this is a good classical TM tool, which would just require some more "keyboard navigation." Obviously, it would take some more time to delve into the advanced functions (term learning, SQL), and I haven't gone that far within the past year. Refer to Atril's Web site for further info. They have a really fine information/help section there, too. But, as recent e-mails showed, you have to be really inquisitive and ask them questions (they are very responsive and helpful), otherwise you will discover the user's manual after a year of exploration on your own!

Trados Translator's Workbench: best in class?

Let me confess that I once hated Trados, but then I had a good excuse. Because the program produces RTF files that can be processed in Word, it has been extensively used by translation agencies to send out "pretranslated" files to their translators. After a few lines of a patchwork of mixed blue source text and black target text such as:

{0>Die Abmessungen des Portalausschnitts im Wagenkasten sind in der Zeichnung "Meßblatt Seitenwand, Endwagen," Nr. {0>2 16106.<0} dargestellt.<}0{> Die Abmessungen des Portalausschnitts im Wagenkasten sind in der Zeichnung "Meßblatt Seitenwand, Endwagen," Nr.{0>2 16106.<0} dargestellt.<0}

the translator usually turned hysterical and swore that she never wanted to hear of Trados again. That was in the days before the translator's partner bought Trados Freelance Edition...

Trados (V2, Freelance Edition, http://http://www.trados.com) has many of DéjàVu's advantages, so this review may be much shorter, but the reason may also be that I have a more superficial experience of Translator Workbench (TW). It doesn't mean that the program is less good. Au contraire, it may be the best TM program I have tried so far (although it would be really difficult to award the "best in class" prize to any one of my three favorites).

Basically, the main Translator Workbench uses Word's interface and works with .doc or RTF files. In addition, the ITP Filter Pack can import Ventura, PageMaker, Xpress, FrameMaker and HTML formats. TagEditor also accepts HTML, SGML and RTF. (A note about TagEditor: I haven't investigated much into that part of the package, apart from trying to find out how tag checking works. The interface is much less user-friendly than in Word, but the tags are reduced to a small, inobtrusive "tag icon." My partner thinks it is quite convenient to process HTML or PowerPoint (98 only) files, but it doesn't seem to be much appreciated among translators.)

Learning "only" the basic use of TW, i.e. not looking too deeply into TagEditor or the terminology manager (slightly more complicated than with DV, but still manageable), is very easy. Most functions appear as a Windows toolbar and are therefore easily accessible, while leaving all the Word commands you may want at hand (except when it suddenly and inexplicably decides to turn them off!). It takes one click to insert a fuzzy match, and only the Alt and + combination to save, record the segment into the memory base and move on to the next segment. There may still be a bit too much mouse work for my taste, though. It is theoretically possible to program keys/key combinations as with "plain" Word, but we haven't (yet?) found out how to do that. Contextual search is very easy, requiring only one right click in the memory window (by the way, it is also very easy in DéjàVu-I just forgot to mention it because I didn't need it quite at much with the files I translated with DV.)

Because you work in Word (I will only deal with TW in Word here, not with TagEditor), eye comfort is considerably improved as compared to the average translation memory program. Segments open as two highlighted lines of text (one for the source, one for the file), and you can define your display zoom as you prefer. The characters used for the memory window are a bit small, but it is not much of a problem if you insert the sentence before revising it. The format commands of the source text are inserted as large, invasive tags, which may also be a problem if a "source tag" has changed in an otherwise identical sentence, e.g.:

<ps "Body" 3>{0>The chapters in this manual are organized by the treatment modalities available in the Xxxx<:cs "Superscript" 1>3<:/cs> system.<}80{>Les chapitres de ce manuel sont organisées en fonction des modalités de traitement disponibles dans le système Xxxx<:cs "Superscript" 1>3<:/cs>:<0}

Later in the same text, you have the same sentence, but for some reason the figure in the "superscript" tag has changed: then the system flags the segment (a full match, except for the tag) as a fuzzy match, and doesn't change the tag:

<ps "Body" 3>{0>The chapters in this manual are organized by the treatment modalities available in the Xxxx<:cs "Superscript" 2>3<:/cs> system.<}80{>Les chapitres de ce manuel sont organisées en fonction des modalités de traitement disponibles dans le système Xxxx<:cs "Superscript" 1>3<:/cs>:<0}

Mismatched tags are only detected when importing the translation, and then one may spend a boring moment correcting them, so watch out for each tag. An automatic change, like in DéjàVu, or an immediate mismatch message like in Joust, would be welcome.

TW's main advantage is the easy update of the translation memory. A translated text can be revised in Word, not necessarily leaving Trados open (beware of inconsistencies between full matches then, but with the text fresh in memory, these are easily avoided). In this case, the RTF file still contains all the source text, but this is formatted as hidden text and it can therefore be easily—well— hidden, to obtain a clearer view of the translation. The edited file can then be reopened into Trados' translation memory and used to update the memory. This goes very fast, and can be repeated whenever you change a comma in the text.

And when you have a pair or source/target texts to add to the memory base, the good news is that Trados Freelance Edition is bundled with WinAlign.

A final note, but not mine: Ecco, the Translator's Friend

One of the Foreign Language forum members is so fond of his TM program that it is worth mentioning, although I never used it. Ecco has the advantage of being cheaper than the other TM programs. So here is what Keith Freeman has to say about Ecco (http://members.xoom.com/PrimaTrans/index.htm):

The main things I like about Ecco are:

  • It's keyboard-based: all the most frequently used commands have keyboard shortcuts, the vast majority of the Alt command keys are mnemonics (the letter actually stands for something meaningful!).

  • It doesn't try to construct a translation for me, it leaves me free to insert the terms found in my dictionaries in whatever order I want.

  • Adding terms to dictionaries is very fast and simple.

  • I can see as much or as little of the surrounding text as I want.

  • It's highly configurable. [Hey, Keith, you could have given more detail about this one!]

  • The sentence breaks are fully adjustable.

  • It has some of the useful features found in e.g. Word: an equivalent of Word's Autocorrect (smarter than the latter in that it recognizes more symbols as word delimiters), ShiftCase (Word's Shift+F3) and a ShiftDate and ShiftNumber (for decimals).

  • The Word macros supplied enable me to export the text to Ecco then re-import the translation, preserving most if not all of the formatting (all I have to do by hand is fix underlined/etc. words/phrases within a sentence).

  • I can decide which old projects to include in the translation memory used for each project.

  • It's fast, even on my not-exactly-state-of-the-art system (100 MHz, 32 MB, Windows 95).

With all that, though, it doesn't do fuzzy matching (not yet, says Keith).