Jul 1, 2010

Results of the June translation tools surveys

At the end of May I discovered that a new widget had been added to Google's Blogger tools, which offered me a simple way to conduct single question polls. I had been wanting such a feature for a while, so I promptly tested it with a poll on translation tools used (the one on the bottom in the graphic at the left) and followed it a few days later with a question on the number of tools used routinely when it occurred to me that they were a number of others like me who make routine use of multiple tools. As I see it, the responses to that question are also important when considering the significance of responses to the other question. The initial trend of the results remained fairly constant throughout the month, even with a large surge in responses after Alex Eames kindly mentioned the polls in his last tranfree newsletter.

Over 40% of respondents use more than one translation environment routinely. (Astute readers may note that the percentages don't add up. Google's programmers are obviously fond of new math or at least screwy rounding and truncation.)  This would include people who actually translate in multiple environments as well as those who use environments other than their favored one for project preparation and QA. My partner is a good example of this latter category; she uses "classic" Trados all the time to export TMs and terminologies provided to her, pre-segment Word/RTF and TTX files for translation in Déjà Vu X or memoQ and to perform tag QA on TTX files. Probably a few other things too. But after she learned the benefits of more modern translation environments, I would risk singing soprano if I suggest that she actually take up translating with the Trados Workbench macros or TagEditor again. Maybe I can get her to look at SDL Trados Studio 2009 one day, but first I think I'll take up something safer like mud wrestling with crocodiles.

The 8 to 12% of respondents who indicated that they use no translation tools is probably well under the true figure in our profession. My blog tends to attract readers with an interest in the technology of translation and not so many who like to contemplate flowery phrases for translating obscure poetry (too bad, really - that's more my taste, but first bills must be paid). But since one of my intentions in creating these polls was to get an idea of the technology interests of my readers and perhaps choose some future topics to reflect these interests, I'm not really concerned about an accurate representation of the full population of translators. The profession is simply too diverse and fragmented for me to care. It's interesting to see that I belong to the 1% of nutcases that use 5 or more tools - I can hear I told you so on this one already from many quarters. And I like to think I'm not a nerd. Einbildung ist auch 'ne Bildung.

In the poll on specific tools, I received some criticism for not including certain tools under a category more specific that "other". As you can see, with a grand total of 15% other, I'm probably not missing much there. When I see interesting minor tools I'll certainly mention them if I can do so with some knowledge, but I don't have the time to offer the extensive expert surveys that you'll get from a guy like Jost Zetzsche and his Toolkit newsletter or book on computer skills and tools for translators.

I probably also should have split the Trados part of the question into at least the new and old generation of SDL products, but I simply didn't think of this until a few days after the poll was running, and I decided to deal with that issue in a follow-up question another day (and I have - see the poll in the left sidebar, which will run until August 1st). This is also relevant to me, as I'll be exploring the no-longer-so-new Studio 2009 version more in the future. (Why call it 2009 still when you release the service pack that actually makes it work in 2010?) It may run like a lame pig on many machines, but I do like some of the interface elements a lot, and it's the best program on netbooks I've used as far as critical parts of screens and dialogs not usually getting cut off.

If the poll statistics even roughly approximate the real distribution of tools among commercial translators, then the perception that "everyone" uses Trados is certainly not even close to correct. The greatest share of licenses is certainly for some version of Trados, but if one considers the number of people who use it as I do primarily for some of its filters and to prepare projects for clients who ask for Trados uncleaned deliverables, then the percentage actually translating with Trados will be far under 50%. This is a point well worth considering for those who want to use the most qualified translator for a given subject: statistically, the odds are against you if you insist on Trados use. It's far better to design projects in a tool-neutral manner as far as practical and let the translators work with whatever they feel most comfortable with. The importance of using the best qualified translator for a particular subject rather than the first monkey with a particular tool in his box is one of the reasons why I like Déjà Vu X and memoQ so much: the RTF table exports from these two tools allows projects to be translated or edited by anyone with a modern word processor regardless of the source format. Every translation environment tool should follow the example set by Atril and Kilgray in this respect.

It's also interesting to note that "major" tools like Star Transit and Across only have 6% of respondents claiming use. Anyone who tries to insist on a translator using those tools is playing a very bad hand.

The main conclusion I draw from these data is that give the diversity of tool use among qualified translators, it is more important than ever to design and test our processes for true interoperability and to drop the religious insistence on the use of One True Tool, no matter what that tool may be.

But that's just my initial reaction to the data. What do you think?


  1. First impressions here, I'd say that for a relative newcomer, MemoQ is doing stunningly well in the ratings, considering how long Wordfast has been around (for example).

    That said, since you regularly blog about MemoQ, your reader profile may be skewed in their favour - but since your survey was plugged in lots of different places, I would expect that skew to be of lesser effect.

    I would like to see more people using OmegaT - simply because it's open source. I have no idea how good it is. But since it's there, it's free and a goodly percentage of people use no TEnt at all, it makes you realise that the reason for not using a TEnt is not likely to be a financial one. Is it lack of publicity or the impression (true or not) that open source stuff is for techies only)?

    It's more likely to be a technical one. If I suggested using a TEnt for a lot of the sort of work we get, I'd be better off singing soprano to the crocodiles (to mix your two images). But then a scanned PDF of a certificate doesn't usually lend itself well to this sort of app does it?

  2. OmegaT is a really nice little program for teaching purposes, but give what I remember of its handling of matches (I don't think variation in the translation of 100% matches is allowed), it isn't really fit for a lot of commercial work. If that has changed or I'm mistaken, I hope Mark P. or someone can offer a correction. But the program has evolved considerably since I first looked at it a few years ago. And there are alternatives like Anaphraseus too. So finances really aren't much of an excuse. That said, the free tools are comparatively limited in the formats they handle, which in turn limits one's project prospects. (Of course, tools like OmegaT's TOXIC improve this situation somewhat.)

    I think one barrier to TEnT use is the false perception that the main value - perhaps the only value in some minds - is for text with a lot of repetitions or matches from previous work. Such text constitutes a very small part of what I do. I depend on my tools for term management, quality control features and concordancing. As good as my memory is, it can't match push-button access to ten years of term research in context.

    I'll have to disagree about those certificates in many cases. I actually use OCR successfully in many cases to produce usable source texts for TEnT translation. It requires various cleanup tricks usually, but it's doable. However, one must be careful in many cases not to mimic the originals too closely, because in some jurisdictions this may constitute forgery (with the use of official seals as images, for example). This is a trap some are unaware of.

  3. Hello Kevin.
    Regarding the use of more than one tool - when I first started, I had to use Trados, because the one client I had insisted on it. I have since upgraded twice, first to the 2007 Suite, then to Studio, and I am absolutely satisfied with Studio (which may have to do with the fact, that I had very few problems with it, and the few there were got resolved surprisingly fast through ProZ's forum). I did look into memoQ and participated in a webinar, and I am now recommending it to everyone who has not yet bought a tool and is thinking about it. I would buy it myself, if I didn't have Studio already, but I just cannot afford to purchase two tools (even if one isn't that expensive) if I am having to use one (Trados) for one of my major clients. Besides, I am basically happy with it, so the need has not (yet?) arisen to switch.
    I would estimate that the majority of translators who use CAT tools probably only have one tool they use and maybe some kind of free version of others (e.g. Across - I have it on my PC, but don't use it, unless a client would absolutely insist). If you spent quite a bit on one tool, you'd have to really make use of another one to justify doing it again, and I know I wouldn't. Just having to constantly convert TMs back and forth is a big turn off for me for using different tools.
    Anyway, that's just what I think - maybe I am overlooking something?
    Have a great weekend, Anke :)

  4. You raise a very valid point about the nuisance of switching environments. The ergonomic difficulties of switching back and forth between classic Trados and DVX was one of the main motivators for me to become an expert on integrated workflows. Aside from the fact that the old Trados was simply less efficient in many cases, I wasted too much time in both tools getting used to the basic keyboard combinations again.

    That's why I encourage anyone who uses multiple tools regularly to check wither one of the tools offers keyboard shortcut customization. I use this feature in memoQ to mimic the DVX key combinations, and I think other packages offer this feature to some extent as well. If not, various keyboard macro programs may help, but I personally am not fond of such solutions.

    Anke, if you find a need to use some features of memoQ (such as a filter which happens to be better), you may be able to make use of the free (unlicensed) version to prepare your projects to work in SDL Trados Studio 2009. No provider makes perfect filters for everything; each of the tools I use has a superior filter for some kind of file, so I change my preparation steps between tools a lot eve if most actual translation is done in one environment.

  5. Kevin,

    My initial reaction to the data? OmegaT has 6% of the market! Not bad for a product produced entirely by volunteers, albeit including some very competent ones.

    On a more sober note: besides the self-selecting aspect of the poll, which you've already touched upon, the formulation of the question also needs to be taken into account. The results are more an indication of ownership than usage. A tool that is used only very occasionally, either in order to perform a function not available in the user's "preferred" tool or to assure compatibility with a customer's workflow, rates just as highly as a tool that might be used for 95% of the time.

    You say that you are surprised at the low figures for Transit and Across (6% each). In fact, I wouldn't be surprised if their figures (and that for OmegaT, which scored the same) are even lower in terms of actual usage. I think that we can safely say that Trados has the lion's share of the market, quite possibly well above the 50% indicated by your poll; we can also infer from the results that a significant number of CAT-tool users use more than one tool. What is very difficult to ascertain is the relative shares of the tools in terms of their use for the total work being done.

    With specific regard to OmegaT: If OmegaT really is being used by 6% of users, it would be a staggering success. And perhaps it is. But I suspect that many techie readers of your blog have tried it and found it useful for some particular purpose (perhaps translating Powerpoint or HTML files if their preferred tool is Wordfast Classic). Since OmegaT doesn't cost anything, it can quite easily acquire this status.

    Equally, to answer Alex' question, it can go out of use for exactly the same reason. If you have invested several hundred euros in a piece of software, you don't bin it as soon as something apparently fails to work. A couple of days ago, one of our devs reported that a user had tried OmegaT, not realized that tags must be preserved, produced a corrupted translated document, and for that reason simply abandoned OmegaT. That is unlikely to happen to someone who has forked out the purchase price of Trados or DV.

    There may be other factors at work, such as a possible distrust of products that don't have a commercial company behind them. Also, although OmegaT's UI has improved progressively over the years, it still doesn't look as slick as some of the commercial competition.

    I feel that we have reached something of a milestone now with OmegaT, and need to find ways of making development more continuous – which we are doing. We also need to harness the community more effectively, in the way that Wordfast has done very successfully. There's a lot of goodwill towards OmegaT, but also a certain expectation that development "just happens", as it does with commercial products.

    On the particular technical point of match handling: it's the variation of repetitions that isn't allowed (not of 100% matches). This is an issue for two reasons: firstly, it prevents the level of automation that some workflows expect, and secondly, it raises questions in the ordinary translator's mind over the validity of the underlying design. For the average OmegaT user, who is not tightly integrated into customers' worfklows (and the vast majority of translators who are, I dare say, use what their customers are using – Trados), there are superficially inelegant but perfectly workable workarounds. Most casual users of OmegaT, i.e. not its hard core of enthusiasts, will probably capitulate and buy Trados if it cements their relationship with their customer, and I suspect the situation isn't all that different for users of many other tools.



Notice to spammers: your locations are being traced and fed to the recreational target list for my new line of chemical weapon drones :-)