Standard Bearers:

TM brand profiles at Lantra-L

ver the past few years, significant changes have been occurring in the translation industry. The Internet has become the terminological resource par excellence, and a hitherto isolated profession has been forging an international community through mailing lists. Translation Memory (TM) software has spread out from in-house translation departments and big agencies to the freelance community, with profound effects on how translation is undertaken and perceived. The profession has reached a transition point. For those attempting to divine how it may evolve, Internet forums provide an ideal window onto the world of the practising translator.

Chief amongst such forums is Lantra-L, where in open and unprompted discussions top-level practitioners around the world debate the issues that concern them, including how they are responding to TM. As agencies increasingly demand TM for certain jobs, possessing and mastering the software has become a question of livelihood; pros and cons of competing packages are weighed, and occasionally passions run high.

As yet, no single standard has emerged, with various brands vying for market share: by examining what TMs Lantra-L professionals are talking about (and indeed, not talking about) we discover that apart from objective views on different brands—principally Trados, DéjàVu and Wordfast—translators who have already made their choice rally around their standard and defend it with vigor. With Lantra-L contributors in the vanguard of their profession, their views and attitudes offer a glimpse of what software will be used by practitioners, taught in translation schools and demanded by clients in the coming years.

In the late nineties freelance translators recognized they had to meet and adapt to the rapid advance of the new TM technologies—as the foundation dates of users' lists (Table 1) so eloquently reveal. Especially (but not only) for technical translators, the issue is no longer "Do I need TM?," but "Which brand of TM should I buy?" As developments continue, some users will also wonder "Is the TM I have now the best for my future needs?"

Table 1: users' groups discussion lists (lists ordered by volume of messages)

Over a similar space of time, the Internet has come to acquire primary status among translators, and email discussion lists are vital nodes of information exchange. Of the multitude of lists available, Lantra-L is probably the oldest (founded in Autumn 1987) and one of the most active (membership at 1 July, 2003, 1148; subscribers in 66 countries, with average traffic approximately 800 messages per week). As a principal discussion forum for translators worldwide, it constitutes a virtual agora where neophytes (newbies in list-speak) can mix with and glean knowledge from experienced users.

Of alternative sources, CATMT (computer assisted translation—machine translation) focuses directly on our study area, but risks too narrow a view (and bias due to it being serviced by Atril, the developers of DéjàVu). The brand-specific TM users' lists are too limited in scope, dealing mostly with technical minutiae. At other professional lists based on country, language, or area of specialization (literary, legal...), while CAT will inevitably be present, it will be confined to the margins. Practitioners will thus be hard-pressed to find a more authoritative and up-to-date opinion on TM issues than through Lantra-L.

The number of members and volume of messages on user lists have been considered reliable indicators of freelancer-preferred brands—an assumption supported on Lantra-L (29 Jul 2002 03:15)¹. Brands on these lists were searched in the Lantra-L weekly archives, with hits for the period sorted in Table 2.

In order to facilitate the study and leave a more transparent record of the research process, content-rich messages (772 in all) were analyzed via computer assisted qualitative data analysis software (CAQDAS)—here, Nvivo2. Data was then systematically coded (by translator and brand), with results tabulated in numerical form, then contrasted with illustrative quotes.

Messages comprise contributions by some 134 individuals—ranging from highly experienced TM users to complete beginners. In a profession perceived as female-dominated, 68 were male (51%) and 52 female (39%), with no gender specified for the remaining 14. Data on country, language and specialization were not significant enough to warrant presentation. While TM preference was not always clear-cut—some use and value more than one, others change allegiance, while yet others have made a choice but are not entirely content—it was possible in most cases to make reliable assumptions, presented in table 2 (unclear value for 44). Incidentally, the most dramatic gender division was for Trados (15 male against 5 female). A brand/keyword out of context (KWOC) analysis, also in Table 2, produced the results consistent with the numbers gathered at Lantra-L.

Table 2: TM brands webpages (ordered by hits at Lantra-L, http://segate.sunet.se/archives/lantra-l.html, 1 Jul 2002—30 June 2003)

DéjàVu is the main force by both the number of messages and number of translators who have adopted it, followed by Trados and then Wordfast. The hierarchy is not immune to challenge, and supporters of the major brands occasionally engage in heated disputes (the so-called Lantra-L holy wars), under headings such as "Trados vs DéjàVu" with very long threads. The first shot in the latest conflict was fired on August 27, 2002 at exactly 08:47, and the ceasefire came three days later, at 21:58. In the interim, some two hundred email barbs were exchanged as a coalition of Trados and Wordfast supporters battled the perceived hegemony of a DéjàVu "clique" in the list. With the importance of TM software to translators, and the development invested in these programs by several competing purveyors, an emerging "expert" preference for a particular brand is bound to concern both users and vendors. Users in particular—for whom TM software represents a major investment in money and time in acquiring proficiency—will rush to defend their choice if there are hints of it losing status. It is therefore hardly surprising that discussions occasionally bubble over into open conflicts.

DéjàVu proponents can seem overeager to some: the Trados fraternity in particular resents the "buy DV" chorus whenever someone reports a problem, exacerbated by the fact that some of the stouter DéjàVu advocates are former Trados users themselves (29 Aug 2002 16:46). Indeed, by touting their software as vastly superior to Trados (and thus, by extension, to other brands), DéjàVu advocates can occasionally antagonize the rest. While "My DéjàVu paid for itself in one week" (29 Jul 2002 22:37) may indeed reflect some translators' reality, it can be repeated once too often: a frequent accusation is that DéjàVu is a "cult" whose zealots have hijacked Lantra-L, and tensions apparently sit simmering until only a spark is needed for all-out holy war.

Though unpleasant for combatants, such conflicts offer the observer an unparalleled opportunity. Here the most experienced users offer the closest we are likely to get to the freelancer's real take on TM, right down to a systematic comparison of which features are useful or not, and why. It may sometimes be "annoying to find exactly the same people saying the exactly the same things" war after war (30 Aug 2002 10:45), but a lot can be learned from what contributors—often witty and pithy by turns—have to say in a few lines about the issues at hand. Given that features keep being added to programs and the technology's reach keeps growing, a holy war every two years or so will serve the translation community very well.

The spark that touched off the last fracas was a message posted by a Wordfaster (and avowed opponent of the "cult") who complained Atril had sent him four unsolicited mails advertising bargain-priced DéjàVu. His complaint, headed "DV Spam", drew a 20-message flurry that soon blew out into 157 "Trados versus DéjàVu" and other exchanges. Once underway, the dispute organized itself around familiar lines, with Trados supporters challenged to show what Trados could do that DéjàVu could not (29 Aug 2002 12:39).

The defense concentrated first on filters: "T-Window for clipboard allows you to translate in *any* Windows application that supports the clipboard—no need for filters" (29 Aug 2002 11:38). "Tag Editor can handle any xml file you care to throw at it, just supply the DTD. Can DV? No it just has a generic SGML filter (which Trados has had since version 1). Executables? Can DV handle exe files or dlls? No? Trados can." (29 Aug 2002 04:33). The DéjàVu camp's response ranged from dismissive—"I'll certainly keep it in mind if ever I am given a file format that DV can't handle (BTW—what formats can't DV handle?)" (29 Aug 2002 09:54)—to appreciative: "now that is good and useful information. it *should* be included in any comparison" (29 Aug 2002 10:03).

Next came terminology management: "Multiterm is far more flexible than a simple Access-based database can ever be. Fine, you might feel that the TDB is quite adequate for your needs, but that is not the point, Trados does it better" (29 Aug 2002 16:33). This is a recurring theme and, while complaints abound regarding the steep learning curve for mastering Multiterm, and some will claim a basic glossary is enough, there seems to be consensus at Lantra-L on the superiority of this particular Trados tool (2 Nov 2002 17:09; 17 Jan 2003 06:46).

While Trados lands a few blows with filters and terminology management, DéjàVu strikes back with its "assemble" feature:

Some DéjàVu opponents fail to be impressed, citing too many distracting options and too much mousework (29 Aug 2002 17:18). They will later coin the dismissive term of "DV assemblers" (26 Feb 2003 10:36). Others would take the feature seriously: "I generally prefer working in Trados, but like DV's sub-segment matching capabilities and use that for texts that will profit from it . . ." (5 Jan 2003 13:48)

Support is regarded as another DéjàVu strength over Trados.

However, for the first time, this long-held perception at Lantra-L starts to change: "I have had very good levels of support from Trados Ireland . . . despite not having a Trados maintenance contract" (30 Aug 2002 09:22); " When I have phoned [Stuggart office], have commented that I really should think about getting a contract if I use TWB a lot. They have never refused to help because I don't have one" (30 Aug 2002 11:45).

Price was another battlefront during the holy war, although the most illustrative quotes appeared early in 2003:

To which a Trados adherent replies:

DéjàVu as Trados substitute

DéjàVu is Trados-compatible, not vice-versa—a fact that would indicate the ascendancy of the latter: Trados is, unquestionably, the choice of agencies; DéjàVu just seems to be that of freelancers. The assumption seems to be that when agencies ask freelancers to do a job "with Trados", then DéjàVu (or Wordfast, or Transit, or SDLX...) will be adequate alternatives. Just how adequate, however, is a matter for debate.

At one point, a DéjàVu user will claim that "DV handles Trados files better than Trados does":

This assertion went uncontested at the time, but DéjàVu / Trados compatibility did not always fare so well. Months later someone lamented: "An agency I work for has just made a royal hash of a large web site for an important customer, and are blaming me because I worked in DejaVu, whereas the other translators worked in Trados" (17 Dec 2002 05:10). Regardless of where the fault lay, there are many instances that show that some Trados knowledge is necessary when handling its files via other packages:

In such cases it is often Trados users who go to the rescue—professional solidarity still outweighs brand loyalty, even if help occasionally comes with a dig:

One individual was blunt in his assessment: "DéjàVu is for translation; Trados, for profanation" (6 Dec 2002 19:52).

Trados plus DéjàVu

It is common for experienced TM users to be familiar with more than one brand—even if some DéjàVu cultists like to say this is due to switching from Trados. Regardless, the emerging trend is that of highly experienced users opting for more than one application: "Two better than one" was the heading of a strand commencing 17 Oct 2002. For the more discerning, the option may not be Trados versus DéjàVu but DéjàVu plus Trados. The following example shows one person coming around to the "more is more" approach:

Months later, however, she reports: "I am working in Trados now because I must track very specific information for the terminology used for this client and wonderful as DéjàVu is, this is not possible with DéjàVu" (25 Feb 2003 11:26). Subsequently, she will also complain she is reaching the limits of the DéjàVu Access based database, it being too slow to compact and repair while losing important client data in the process. A month later, a strategy combining both applications has emerged: "My solution is to pretranslate all texts with Trados (presegmenting unknown sentences), work in DéjàVu (creating Trados Workbench) projects, revise the exported file in Trados and then delete the original source file from DéjàVu and import the final translation" (18 Mar 2003 16:38).

An example of a prominent Trados user acknowledging use of both applications can be found at 5 Jan 2003 13:48.

Wordfast versus Both

Wordfast is a Word macro, a small program meant to perform—at least for Word files—almost as well as the big brands. It has gained an impressive market presence, as traffic in the Yahoogroups users' list shows, although its numbers are far less imposing on Lantra-L. Nevertheless, it is clearly positioned as the third main option, with Lantra-L members apparently seeing Trados as the agencies' preference, and DéjàVu and Wordfast as the choices of leading-edge freelancers and the average practitioner respectively.

Wordfast grew its user base through what was until recently unbeatable cost-effectiveness: it was free until October 2002. The developer, Yves Champollion, had entered into an agreement with a major agency, Logos, for the program's distribution, only to back out when he felt certain contractual conditions had not been met. Logos retained the source code and kept distributing it, Champollion began retailing a new version through his website, and the question of who has legal marketing rights is before the courts. Meanwhile, commercialization does not seem to have eroded Wordfast's market, despite availability of free or budget-priced alternatives.

Given Trados's present agency-preferred status, and that Wordfast, like Trados, employs MS Word as its user interface, Trados draws the most comparisons. For experienced Wordfasters, Trados "seems pretty familiar, like a bloated WordFast", and thus not as flexible (4 May 2003 10:42). It is deemed less user-friendly: "The learning curve for WordFast for me was about 10 minutes. Trados works on the same principle, but is a lot more convoluted" (18 Apr 2003 07:41). As with DéjàVu, Trados compatibility is a strong point:

There is also some proselytising, with floundering Trados users advised to jump ship: "Although this is probably not what you want to hear now that you've forked out your CHF 895, you could also give Wordfast (www.wordfast.net) a try" (20 Sep 2002 09:36); or "get a copy of Wordfast (easier and cheaper than Trados)" (21 Jan 2003 00:13). (For some reason, this does not seem to raise as many hackles as when it comes from DéjàVu proponents).

List members agree that Wordfast is effective for most needs, but differ on how far it can go. Some users would not hear of limitations: "in my experience, WF handles HTML, PPT and Excel as well as it does handle Word files" (20 Sep 2002 06:09) but most happy users admit the bulk of their work is in Word: "the situation could be different for people who have to translate more exotic stuff" (20 Sep 2002 09:36). The consensus seems to be that Wordfast handles Word files well, but is not as effective with other formats. Claims that Wordfast can manage Excel files from within Word are disputed: "Misleading info. It can, but the process is not pretty" (26 Feb 2003 12:43); limitations with PowerPoint and HTML are also pointed out (8 Oct 2002 16:12; 25 Jul 2002 16:51). It is repeatedly noted that results depend on how the source format has been handled.