Wordcount "shredder" CAT tools - fooling the translator?
Thread poster: sentrodil

sentrodil
Turkey
Local time: 15:31
Sep 7, 2011

hi folks,

I don't know if anybody has become aware of some (un?)intended "shifting of the scale against the translator" policy of some CAT tools. I am certain, that they are either not intending this, or are not aware of it, but "additional features" of the analysis features of many of the new generation CAT tools initiate some feature (like the Homogeneity of memoQ, the internal repetition of Transit, and possibly the word count systems of Across and SDL Trados Studio) give such word counts that although the total no match and fuzzy words of projects look less, while the amount to be typed (and even taking into account that the "fuzzy" matches bring up some fuzzy matches) seems to be more than usual. Some of these new CAT tools let you export the parts to be translated, and if you analyze these texts with "conventional" CAT tools, the no match and fuzzy words (excluding tags, of course) comes up to more than what seems to be expected for that volume with conventional CAT tools.

So, in short:
while a project gives an analysis of 15,000 no match words, 200 fuzzy words, and 500 repetitions/100% match words in Trados 2007, the same project shows as 3000 no match words, 12,000 fuzzy words, and 500 repetitions/100% match words in "new generation" CAT tools.
Since most agencies ask for and get discounts for the fuzzy matches, this approach results in a 50% or more drop in the total rate of the project, based on the assumption that the 12,000 words are fuzzy. But ,in practice, the fuzzy matches brought up by these tools are either less than those 12000 words, or not much use, so they could be left as "no match".

anybody aware of this?? experience with this?? what is your approach to such suggestions??
"meekly accept it" ??
"fight for your RIGHT, leave if nothing's LEFT" ??
"be amazed, and feel lightly glazed" ??

waiting for your replies...


Direct link Reply with quote
 

Luca Tutino  Identity Verified
Italy
Local time: 14:31
Member (2002)
English to Italian
+ ...
Offer no discounts for fuzzies under about 95% Sep 7, 2011

You are right, the problem is there.

Also the quality of fuzzy matches is mostly overestimated. In my experience "83% fuzzy" TUs actually require what seems as an average of 99,9999% of the time needed for new TUs of the same length (rather than 17%, as the measure might seem to imply).

What to do about it? Think twice before offering any discount for fuzzies. At least do not offer it for matches under a reasonable level, variable according to the type of project. In many cases even 100% matches will not speed up your translation, if you cannot fully trust the TM.


Direct link Reply with quote
 

Gerard de Noord  Identity Verified
France
Local time: 14:31
Member (2003)
German to Dutch
+ ...
My first reaction Sep 7, 2011

My first reaction is to be always cautious about the deadlines I state. 15,700 words in all remains 8 working days, whatever the status of those words.

This 'new' method of counting is not so new at all. Maybe it has just been automated further. It's the old empty TM trick: analyse a document against an empty TM, copy source when no match, analyse the next document to obtain virtual fuzzies.

My second reaction is to calculate what I'd normally invoice and what I can invoice for the current job. I'm only interested in the bottom line. Sometimes all that creative computing results in giving a discount of, say, 30 euro.

Cheers,
Gerard


Direct link Reply with quote
 

Kay Barbara
United Kingdom
Local time: 13:31
Member (2008)
English to German
+ ...
Agree: no discounts for fuzzies under about 95% Sep 7, 2011

Luca Tutino wrote:

You are right, the problem is there.

Also the quality of fuzzy matches is mostly overestimated. In my experience "83% fuzzy" TUs actually require what seems as an average of 99,9999% of the time needed for new TUs of the same length (rather than 17%, as the measure might seem to imply).

What to do about it? Think twice before offering any discount for fuzzies. At least do not offer it for matches under a reasonable level, variable according to the type of project. In many cases even 100% matches will not speed up your translation, if you cannot fully trust the TM.


I completely agree with Luca here. Whether fuzzy matches really are a time-saver depends on your source and target language as well as the segment length.

In my specialty field (games) very short and often similar text strings (NB: similar in the source text) are commonplace, however, due to the grammatical differences between English and German (e.g. grammatical gender & cases), a 80-90% match may still require changing about 80% of the words. At the end of the day I need to charge for my services according to the time invested into any given project (obviously reflected by an appropriate word rate) but since fuzzy matches are not a reliable indicator, I do not grant discounts on the basis of such analyses.

I may give discounts for 100% matches and repetitions if they really don't require any changes, but even a 100% match may still require changes due to the above mentioned differences between languages.

I don't know about everyone else, but my impression is that the less educated a client (mostly agencies) is about translation, the stronger they try to enforce discounts.
Fortunately there are still plenty of clients out there who are more concerned about quality than haggling.


Direct link Reply with quote
 

Philippe Etienne  Identity Verified
Spain
Local time: 14:31
Member
English to French
The evil option: awareness needed from translators Sep 8, 2011

sentrodil wrote:
...I don't know if anybody has become aware of some (un?)intended "shifting of the scale against the translator" policy of some CAT tools. I am certain, that they are either not intending this, or are not aware of it, but "additional features" of the analysis features of many of the new generation CAT tools...

Well, they just don't sell this feature to translators as the option that decreases the weighted wordcount (Note: SDLX has had the feature for ages. I used to charge more compared to Trados because of that [and the fact that I hated the interface]).
Do they promote this feature to agencies? Apparently, most agencies using weighted worcounts know where this option is, and use it...I'd be curious to know the default option in CAT versions for agencies. On or off?
Let's not be naive: it is likely that CAT makers generate far more business with companies and translation agencies than with individuals like us and a few hundred euros here and there. It's only fair that their offer is geared towards companies.
But freelance translators should be business-minded too.

Why is this spreading option "evil" ? Because it adds nothing in terms of productivity, and it shakes my fragile business arrangement in terms of leadtime, deadlines, customer base, hourly income, word rates, CAT discounts and translating pleasure: it's a whole framework that needs rethinking.
sentrodil wrote:
...while a project gives an analysis of 15,000 no match words, 200 fuzzy words, and 500 repetitions/100% match words in Trados 2007, the same project shows as 3000 no match words, 12,000 fuzzy words, and 500 repetitions/100% match words in "new generation" CAT tools....

Yes, it's most visible in catalogues and similar highly repetitive texts:
Blue long flashy trousers
Green long flashy trousers
Green short flashy trousers
Old CAT tools, SDLX excluded (empty TM): 12 new words
New CAT tools and SDLX (empty TM, evil option enabled): 4 new words, 8 fuzzies
Effort: identical
assuming a 100/60/30 scheme, that's a 25% decrease in price.
And the higher the wordcount, the higher the probability to get an internal fuzzy, hence the higher the loss.
sentrodil wrote:
But ,in practice, the fuzzy matches brought up by these tools are either less than those 12000 words, or not much use, so they could be left as "no match".

Internal fuzzies are real fuzzies. It's just that they used to be to the translators' benefit. One could see this benefit as a return on investment for buying the tool (which is not known as a cheap piece of software). Now, the benefit goes to the agency. So you only buy the tool to get work from agencies who use the tool.

The only way to get some return from a CAT tool now with these options is:
- increase the base word rate,
- sharply decrease discounts on fuzzies, as said earlier.

It is most relevant with technical and repetitive translations, where CAT tools ARE useful and give internal leverage.

So beware. This whole evil option business is a serious issue, especially when you have worked a lot with Workbench with an established customer base that is gradually switching to new-gen CAT tools featuring the evil option...
Newcomers offering 0.08 euros/source word should know that with the evil option, it's rather 0.07 or less in "old currency" that they earn.

Following the same slope, I wonder when other discount schemes will come into force around features such as Autosuggest, LSC or other add-ons that only freelancers are benefiting from so far, however marginal the speed benefit may be...

Philippe


Direct link Reply with quote
 
xxxkalap
Discounts on discounts Sep 8, 2011

In all cases, the total price of a translation should always be in line with the total effort spent on this translation. If a new scheme disturbs this balance, the prices should be recalculated, and if possible, beforehand.

Philippe Etienne wrote:

Following the same slope, I wonder when other discount schemes will come into force around features such as Autosuggest, LSC or other add-ons that only freelancers are benefiting from so far, however marginal the speed benefit may be...

Philippe


The benefits of his these features are not measurable, so it shouldn't. But in my opinion they are real. Therefore, a translator benefits from knowing thoroughly his preferred CAT, and if he doesn't use these features, or if his CAT doesn't have any, he will be handicapped.


Direct link Reply with quote
 


To report site rules violations or get help, contact a site moderator:


You can also contact site staff by submitting a support request »

Wordcount "shredder" CAT tools - fooling the translator?

Advanced search







TM-Town
Manage your TMs and Terms ... and boost your translation business

Are you ready for something fresh in the industry? TM-Town is a unique new site for you -- the freelance translator -- to store, manage and share translation memories (TMs) and glossaries...and potentially meet new clients on the basis of your prior work.

More info »
PerfectIt consistency checker
Faster Checking, Greater Accuracy

PerfectIt helps deliver error-free documents. It improves consistency, ensures quality and helps to enforce style guides. It’s a powerful tool for pro users, and comes with the assurance of a 30-day money back guarantee.

More info »



Forums
  • All of ProZ.com
  • Term search
  • Jobs
  • Forums
  • Multiple search