Word count issues – Part I June 17, 2008
Posted by Jill (@bonnjill) in Business practices, Tools, Translation Sites.trackback
I’m amazed to learn that one of my translation agency clients, which specializes in financial translations (note: I do not translate the heavy financial stuff. They have me translate magazine articles and the like), relies on the Trados analysis feature to perform its word counts, because Trados does not count stand-alone numbers. Numbers often have to be localized (commas changed to decimal points and vice versa), so if you have a document with a lot of numeric information with decimals or tables of pure numbers, Trados short-changes you. I feel that if I have to look at the number and/or physically alter it, then I should of course be compensated for that work.
Trados is a translation environment (CAT) tool and not a word count tool. Marita Marcano and I wrote an article about word count tools in the ATA Chronicle (“What’s in a Word?, p. 32 of the August 2006 issue) and presented a session at last year’s ATA conference on word count tools with Clove Lynch, who was representing the Localization Industry Standards Association (LISA). LISA is working on coming up with a word counting standard, so hopefully this will change things soon. However, as it stands now, each tool has its own counting algorithm it uses to perform its count. There is no standardization between tools. We found PractiCount and Total Assistant to be the most accurate tools in our comparison. Word was a disaster (for reasons I will go into tomorrow for those of you who are not aware of the problem).
It seems to be a well-known fact among my colleagues that Trados does not count numbers. I don’t understand why some agencies insist on using Trados to do their word counts. They are not only shorting the translators, who are typing in the numbers, but are shorting themselves because they bill their clients based on these incorrect word counts.
Update: Client education does indeed work. I just received my client’s response: “I will definitely forward your feedback on the Trados Tool. It is the only tool we have to make a wordcount!!” I referred them to my article as well.
Bear in mind that Trados not only does not count as words numbers, (at least when they are alone… I’m not sure whether it counts them or not within segments): a more serious problem is that Trados does not open as segments numbers, so any formatting chage has to be done outside Workbench.
When the work to be done involve changing from a numeric format to another (e.g., US 10,000.21 to Italian 10.000,21), a good solution is to use the wildcard functions of MS Word (or, otherwise, use some regular expression tool).
If you are interested, you can find a detailed discussion of this and instructions in my article “How to use wildcard and format searches in MSWord to make sure all your numbers are formatted correctly” (http://aboutranslation.blogspot.com/2005/05/how-to-use-wildcard-and-format.html).
It does count numbers if they are within segments. But your point about not opening the numbers-only segments is a very valid one. That is why I hate working in TagEditor – I can’t touch certain numbers. I like using Trados with Word because you can easily alter numbers that are jumped over (such as dates or numbers in a table).
I’m pretty sure I read the article in the ATA Chronicle and really enjoyed it, but it is worth reading again.
Jill,
Another point to remember is that Trados doesn’t count spaces, so any line counts based on Trados character counts will always be underreported (and consequently undercharged). You’d have thought they’d have fixed that by now, but I guess user-friendliness never really was a feature of Trados products….
Talking of line counts, it’s also worth remembering that Word doesn’t count characters/words in text frames.
Interesting comment about that other financial translation provider. Just goes to show that size doesn’t necessarily mean quality 🙂
Great post, thank you! It seems that everything other than tools like AnyCount and PractiCount (please make a Linux version!) have issues. I used to use WordFast’s wordcount feature, but then it didn’t match the MSWord count that the client used, and for all of its positive qualities, OpenOffice.org is similar–the word count is almost always higher than Word’s. Let’s hope that LISA comes through!!
A standard for word counts would definitely be great. I just hope that they remember to include the “double-byte character” count — essential in my language pair (Japanese to English), and so often left out by word-count tools.
In my opinion, MS Word is the de facto standard right now, but it also has issues (for example, it counts bullets in bullet lists as “words,” which is pretty odd IMO). Incidentally, Word 2007 finally counts words in text boxes.
Corrine — I have an online word count tool that you could use from Linux. It doesn’t handle MS Office documents, unfortunately, but you could save your Office document as XML or HTML and try that.
Too bad Trados isn’t really comptatible with Word 2007. The speed of Visual Basic makes Trados essentially unusable with it.
I have used CompleteWordCount, which is an add-in for Word 2003, for a number of years because it automatically counts text in Text Boxes, etc.:
http://www.shaunakelly.com/word/CompleteWordCount/index.html
Ryan,
Thanks for the info about Word 2007 and frames. However, we don’t use that version of Word, neither do any of our (corporate) clients. We’re all still happily using Word 2000 or 2003, and none of us see any reason to change. We all need relatively simple, uncomplicated word processors, not lifestyle accessories 🙂
I should point out that the same applies to Windows Vista, which simply doesn’t exist as far as we or our clients are concerned.
The French versions of Microsoft Word have always used the English algorithm designed to take account of the “apostrophe + s” in English. So the elided articles or pronouns in French are not considered to be individual words by the French versions of MS Word (the same problem obviously exists with the English versions of MS Word). This means that MS Word’s counts on French text have always been an average of 20% below the actual figure – which is absolutely ridiculous!
Example:
“Je te l’ai dit l’autre jour” is 8 words but the both the French and English versions of MS Word give 6.
RT