Poll: My lowest threshold to consider a 'fuzzy match' is:
Thread poster: ProZ.com Staff
ProZ.com Staff
Local time: 11:50
SITE STAFF
Jul 31, 2007

This forum topic is for the discussion of the poll question "My lowest threshold to consider a 'fuzzy match' is:".

This poll was originally submitted by Richard Jenkins

View the poll here

A forum topic will appear each time a new poll is run. For more information, see: http://proz.com/topic/33629


Direct link Reply with quote
 

megane_wang  Identity Verified
Spain
Local time: 20:50
English to Spanish
+ ...
A 90% may be too low, or very high... Jul 31, 2007

Note that I almost always work in technical translations. If we just move onto marketing materials... some of this can be different.

When this 10% means that a phrase is negative, or a word takes a different meaning, a previous entry that "fuzzily" matches a new one may be simply useless in Spanish and even worse in Catalan. As a result, stating that a 90% match means 90% less job is false (you need to translate the whole thing to make it sure that you don't skip an important detail and get nonsense, and you, or a colleague, have to proofread the result).

To be honest, though, from a 70-80% fuzzy match I still may get a suggestion I can use for consistency reasons. But that almost never happens below that.

If "fuzzy matches" are only to be considered an excuse to reduce my bills, then a 99% is not a match at all (CAT tools should be used to get better quality translations, not to cut the translator's income through "Trados scales" and dont-know-what)

This is why I created my "Trados" rate. What I never tell is... is this rate is above or below any other rate?

Ruth @ MW


Direct link Reply with quote
 
MikeTrans
Germany
Local time: 20:50
Member (2005)
Italian to German
+ ...
Fuzzy matches are not that productive, because... Jul 31, 2007

...the sentences are too long most of the time. What I want to retrieve are logical parts of expressions within a sentence and so, I don't care about the fuzzy percentage.

Trados, DVX, SDLX etc.. are all not able to make longer parts of sentences available without having to start a manual concordance search.

Trados...
has the "automatic concordance". But this tool also only considers the word quantity (percentage) and not the occurences of words IN A ROW, which is more important. It's however the best feature in the hole program.

DVX is not able to show short sentences against long sentences in the TM, which clearly shows that the fuzzy percentage is not the only important criteria.

SDLX:
lacks of autotext and has bad terminology assistance but a very reliable concordance, but also fails to make sentence portions available without manual search.

The best program in relation to fuzzy match is:

Fusion...
but this program really cannot be made worse, as far as term insertions are concerned and plenty of bugs, not worth a normal, critical workflow.

My conclusion:

Presently, there is no serious CAT tool on the market that really helps to increase productivity: In the time necessary to make a concordance search or compare fuzzy matches (or whatelse you like to call them), I write half of all this nice reply.

good work!

Mike


Direct link Reply with quote
 

Jussi Rosti  Identity Verified
Finland
Local time: 21:50
Member (2005)
English to Finnish
+ ...
Surprised Jul 31, 2007

by how low fuzzy matches people accept! I accept lowest 85%, but I see that well less than 20% of the translators have marked 80-100% on the poll.

Maybe it's different with other languages? In Finnish, which is strongly inflectional language, lower matches are virtually useless, since as the grammar of the sentence changes, many words change in it, and the correct translation for 75% match in source is probably about 30-40% match in target. Editing this match is typically much more time consuming that re-translation (since careful checking against copy-paste type errors takes more attention than not to make those errors in the first place).

I can imagine that in less inflectional languages the situation may be more favourable for the translator. Still...

[Edited at 2007-07-31 15:18]


Direct link Reply with quote
 

Luciana E. Lovatto  Identity Verified
Local time: 15:50
English to Spanish
90% is for me the lowest threshold to accept a fuzzy match Jul 31, 2007

Although in practice, things are different because generally clients consider a 60% a fuzzy match. I totally disagree with this policy since it takes more time to compare what matches and what doesn’t than making a simple copy source and doing the corresponding translation from scratch. But unlikely, sometimes I can’t change reality; I just have to adapt myself to today’s market.

Direct link Reply with quote
 

Jerzy Czopik  Identity Verified
Germany
Local time: 20:50
Member (2003)
Polish to German
+ ...
I work permanently with 30% fuzzy match threshold Jul 31, 2007

in SDL Trados and 60% (or 50%, depending on job) in Transit Satellite.
When working with technical text an explanation in brackets in identical sentences may lower the match value so far, that you have something about 40% and just need to delete part of the sentence.
OTOH, when the match coming from the TM is worthless I simply don't use it. However, would I have set my threshold higher, I could possibly lose valuable informations.

Jerzy


Direct link Reply with quote
 

Fernando D. Walker  Identity Verified

Local time: 15:50
English to Spanish
+ ...
Sometimes it can be a problem... Jul 31, 2007

Contrary to one may think, fuzzy matches sometimes are not a good help. From my point of view and my experience translating with CAT tools, in some cases, it is necessary to translate a whole sentence again although it is a 90% fuzzy match. I agree with some of the colleagues in that a word, in some contexts, has not the same meaning, so you must retranslate it or adapt it to the subject you are translating.

Direct link Reply with quote
 

Vito Smolej
Germany
Local time: 20:50
Member (2004)
English to Slovenian
+ ...
A fuzzy match is the mother's little helper Jul 31, 2007

... a vocabulary with the context browser thrown in. In a lot of cases it answers my ID's mulling over the question :"How should I phrase THIS now..."

I am usually at 65-70%.

Please note that 65% in Slavic languages may be equivalent to 85+ in English.

[Edited at 2007-07-31 18:02]


Direct link Reply with quote
 

Luciana E. Lovatto  Identity Verified
Local time: 15:50
English to Spanish
Me again! Jul 31, 2007

Personally, I think CAT tools should only be used by translators and they shouldn’t be used to make the word count, because when you translate you have to check everything even 100% matches and repetitions. For instance, there are cases when the phrases are no well segmented or the TM has mistakes or doesn’t meet the criterion set in the client’s guide. That’s why I think it’s difficult to apply a rate based on fuzzy matches and if I had to set a lowest threshold for a fuzzy match I would say 90%.

Direct link Reply with quote
 

David Russi  Identity Verified
United States
Local time: 12:50
English to Spanish
+ ...
Depends on the text and the TM Jul 31, 2007

When a TM contains lots of formatting and other codes, the text in 90% matches can be virtually identical.

Sometimes, if the client made many global changes (for example, changed the name of a product,), low fuzzies can be close to 100 percent matches (except ofr the one change involved).

Of course, other times 90% matches can be useless.


Direct link Reply with quote
 

Tomás Cano Binder, BA, CT  Identity Verified
Spain
Local time: 20:50
Member (2005)
English to Spanish
+ ...
Absolutely right! Aug 1, 2007

David Russi wrote:

When a TM contains lots of formatting and other codes, the text in 90% matches can be virtually identical.

Sometimes, if the client made many global changes (for example, changed the name of a product,), low fuzzies can be close to 100 percent matches (except ofr the one change involved).

Of course, other times 90% matches can be useless.


Yes, I entirely agree with this David. Sometimes a 90% match is useless, and in other cases a 60% match can help.


Direct link Reply with quote
 

William [Bill] Gray  Identity Verified
Norway
Local time: 20:50
Member (2006)
English
+ ...
What are you using the statistic for? Aug 1, 2007

If it is to help you with the JOB of translating, then I concur 100% with Jerzy, set the limit as low as possible, so that you get maximum information from you carefully collected TMs (I know Jerzy is a fan of "Big Mama" TMs! ).

But if it is to help with PRICING your translations, then anything under 100% should not be discounted, since it requires your checking/proofing/editing time, even if you get some help on the side. (That's why you forked out so much money for a good CAT, isn't it? Don't give it all away a second time with "fuzzy matching".)

Thanks to Richard for launching us into this discussion.


Direct link Reply with quote
 


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

Moderator(s) of this forum
Jared Tabor[Call to this topic]

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

Poll: My lowest threshold to consider a 'fuzzy match' is:

Advanced search






LSP.expert
You’re a freelance translator? LSP.expert helps you manage your daily translation jobs. It’s easy, fast and secure.

How about you start tracking translation jobs and sending invoices in minutes? You can also manage your clients and generate reports about your business activities. So you always keep a clear view on your planning, AND you get a free 30 day trial period!

More info »
Across v6.3
Translation Toolkit and Sales Potential under One Roof

Apart from features that enable you to translate more efficiently, the new Across Translator Edition v6.3 comprises your crossMarket membership. The new online network for Across users assists you in exploring new sales potential and generating revenue.

More info »



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