Thursday, November 06, 2008

Trados: beware of wrong links

I have complained in the past of various problems Trados has with fuzzy matches. A particularly insidious one is the way Trados treats Internet URIs and other links.

For matching purposes, Trados processes URIs as if they were no different than any other text segment. By doing so, it considers two identical URIs (say, for example "") as 100% matches (which is OK), but it also consider two different URLs (for example, "" and "") as fuzzy matches.

This is wrong and dangerous.

If the two URLs are as different as and, the problem may seem trivial: a glance suffices to see that they are different, and to copy over the correct URL.

But other addresses looks much more similar, and we can easily accept the wrong link while translating quickly:

Trados, in fact, consider these two links as 97% similar.

But there is no such a think as a "similar" link: it either gets you to the correct page or file, or not, and the way Trados matches them adds to the risk of inserting a wrong link in our translations.

Changing the internal logic to treat URLs differently would be trivial, but the Trados programmers (or, more likely, their managers), cannot be bothered to spend the necessary time for a simple improvement that would ensure higher quality.


  1. Yes, this is something I have also worried about. Not only could we get a "fuzzy" web link text but we don't actually know whether the actual link - the one you see when you click Edit Hyperlink, which need not be the same at all! - is correct!

    So links have to be checked very carefully, another task I, like you, begrudge Trados making me perform and why I am still undecided as to whether to fork out for it for my company.

    Another thing with links, of course, is that when you are translating a website, in particular, the links are going to be different, probably! Meaning, if I am translating a Serbian website into English, the links will be translated, say, from: to Indeed, that is something only the customer can really determine, so that is something I carefully disclaim and merely make sure the original Serbian links are unchanged.

  2. What a fabulous thing to point out. I'll be sure to mention it at their booth today. This bothers me a lot too, but I never thought about complaining until now.

  3. What really maddens me is how stupid this is: most programs know how to deal with something that looks like an Internet address or other link. The reason Trados still can't is that probably all or most of the matching algorithms inside are over fifteen years old, and have not been improved at all since I don't know how many versions.

  4. Good to point out! I agree, changint the internal logic to treat URLs differently would be trivial.


Thank you for your comment!

Unfortunately, comment spam has grown to the point that all comments need to be moderated. All legitimate comments will be published as soon as possible.