100% text repetitions: To review or not to review
Written by ForeignExchange Translations on Wednesday, February 04, 2009
The adoption of translation memory (known as "TM") technology continues to soar among drug and device companies.
One of the questions that more and more clients ask is "Why do we need to pay for 100% matches? Shouldn't these sentences be free since they have already been translated?"
To answer this question, let's start with a look at how TM systems work.
Translation memory systems basically consist of a database in which a source-language segment (a segment is usually, but not always, a sentence) is stored together with the corresponding translated segment. This combination is called a translation memory "unit." During translation, new source segments will be compared to the database and a match value will be calculated.
A match value of 100% means that the new source segment is identical to a segment in the database, down to the last space, period, and letter. If the match value is below 100% and above a certain user-definable percentage (i.e., "fuzzy match"), there exist some differences in the segments, such as different words, plural vs. singular, or even just different numbers. If the match value falls below the user-defined percentage (typically, 75%), the text is considered "new."
Based on this description, it would be reasonable to conclude that 100% matches should, in fact, be free. This is where the wonderful art of language comes into play.
Consider, for example, the following two pairs of segments, with segments 1 and 2 stored in a TM database and segments 3 and 4 to be newly translated:
Segment 1: The blue house has been built by Mr. Jones.
Segment 2: It is displayed on the next page.
Segment 3: The green hospital has been designed by Mr. Smith.
Segment 4: It is displayed on the next page.
In this example, segments 1 and 3 would be considered as new text. They are similar, but for the purposes of translation, segment 3 needs to be translated from scratch.
Segments 2 and 4 are identical in English. However, when translated into French and German, the "It" would need to be translated differently because house is feminine and hospital is masculine in French, and, respectively, neutral and masculine in German. If the TM segment 2 were to be reused without editing, the German and French would be wrong.
Most translation vendors will charge a nominal fee for reviewing 100% matches. At ForeignExchange, our translation charges are discounted as follows:
- 10% of the per-word cost for text repetitions and 100% matches
- 25% of the per-word cost for 99%-95% fuzzy matches
- 50% of the per-word cost for 94%-75% matches
- 100% of the per-word cost for matches below 75%
Some companies insist that this quality-control step not be included in the translation process. For certain types of texts and for companies that can accept gender-agreement and other minor errors in their translations, this is absolutely appropriate.
However, for the majority of pharmaceutical and device companies that are not willing to accept these kinds of linguistic quality risks, 100% matches must be reviewed to ensure superior linguistic quality.
Confused by all of this talk about "translation memories"? Read on:
- What you need to know about translation memories is a good starting point for more information.
- Translation memory vs. glossary tries to clear up the common confusion between these two tools.
- Machine translation vs. translation memory looks at the difference in tools that aid vs. replace the human translator.
- We have produced three audio conferences on efficient TM design, leveraging and use.
Like this post? Subscribe to Medical Translation Insight via email or RSS!
We create them and they are ours. We have to accept CAT tool reductions so we should charge for our TMs and that is perfectly normal. As soon as translators wake up and realize they are being swindled they might start working to earn a living instead of continually being put upon by agencies at huge losses to the said translators.
10% seems to me to be about the minimum you could charge for 100% matches and still get a quality translation, without cheating the translator.
As a reviewer of translations into German, I often modify sentences that are perfectly fine on their own to create flow within the text. Isn't it in the client's interest to have a document that will be easy to read and understand? In the end, it is not just the content that matters but also the presentation.
In general, if you don't have any additional quality steps, I agree that 100% matches should be reviewed. But if your process, like ours, includes a 100% in-country review and you are leveraging from a previous revision of the same manual that was translated by the same linguists, it's generally a waste of time and money, in my humble opinion.
It's all good and well for a translator to say that they are being swindled out of money they deserve, but you have to be realistic and understand that clients have a budget and a schedule to work with and they have a responsibility to make sure that budget goes as far as possible. It is up to the client to decide which trade-offs he or she is willing to make based on the company's goals and objectives for quality, time and budget. It is the linguist's responsibility to point out the risks of the various options if the client does not fully understand those.
Another thing that has not been mentioned here is that the ContextTM solutions that most TM vendors offer these days go a long way towards preventing the problems as described in the article.
As a rule-of-thumb, the criteria I use to decide whether or not to ask linguists to review 100% matches are:
-Is the translation a revision of a previously translated document?
-What percentage of the text has changed compared to the previous version?
-Was the previous revision translated by the same linguist?
-How recent is the previous translation?
And in case of doubt, I am not shy to ask the linguist's opinion.
If you look at the circumstances on a case-by-case basis and understand the process and the risks, an experienced project manager should be able to make a common-sense decision. As a linguist, make sure your clients understand that they get what they pay for. Do what they ask you to do and move on to the next job. Don't secretly review the 100% matches anyway because you don't want to deliver bad quality and then complain about it afterwards in forums like this. That way your clients will never learn.
Even with matches that I created myself, I may have to modify the text to fit the text environment.
A good translation is more than the sum of individual, correctly translated sentences.
And there is another aspect to using CAT tools and providing discounts for matches. These tools come at a cost to the translator. They are useful to enhance consistency and this should rather be rewarded by higher rates than unfair discounts on fuzzy matches.
1. 100% matches need to be reviewed
2. Context-dependent pretranslation allows for increased revenue and requires nothing from the linguist so it's a fair mechanism to share the responsibility and the work (of course when one says not to review something he takes on the responsibility for this text).
From an engineering point of view nothing stops the service provider to extract freqent segments, have them translated and apply pretranslation (lock relevant segments). Then everything is clear.
In general, we do have a different rate for 100% matches and fuzzy matches so that the 100% matches are reviewed instead of translated from scratch. That's fair enough. I do not particularly like the idea of leaving 100% matches unread. It is not safe enough quality-wise.
My 2 peso cents. :)
Good translators keep up with the ever-evolving technology and realize that -as much as they love their job and want to deliver top quality all the time- their first priority is to pay the bills. I know a lot of translators who have done very well for themselves. They are happy to use translation memory tools and provide discounts. They have taken advantage of the opportunity these tools provide them to deliver consistency and better terminology management at a fraction of the time. They have increased their productivity, have been able to take on more work and are making a lot more money than I do.
Once again, the root of the argument is money - and all these comments just go to show how complex the whole issue is. In an ideal world, everyone would make their own decisions based on a full awareness of the different factors involved. Those decisions, if based on the right information for both parties, are not going to be the same every time. I know they aren't for me.
For instance, if I want to pay my bills and not hate my job (i.e. burn out), I need to earn a certain amount per hour (which might mean a different rate for each job if I get really picky about this point), but I made those decisions with some experience behind me and the knowledge of what I can and cannot do. When it comes down to it, each job has to be evaluated on its own merit.
And then there are word counting methods (mentioned a number of weeks back) that have to be taken into account - just to add another dimension of complexity.
So, in the end, some honest soul searching, calculations and limit setting are in order for everyone. What to do with 100% matches is just one of many things to take into consideration, both for the translator and the client.
Frankly, I hate it when I have to use a TM with 100% matches that I am supposed to leave alone - and that I don't get paid to review - because my perfectionism makes me want to check everything. But the client here has a point about efficiency and other factors that the translator may not be aware of. The translator needs to know what these factors are, just like the client needs to be educated about certain things as well.
It is tempting to make this into an either-or problem, but, like most problems, it just isn't.
In the three step process we apply, fuzzy matched segements require handling by all three parties involved (translator, editor, and proofreader), and our standard pricing is 66% of the new word rate. Full matches and repetitions are not processed by the translator, but the editor "validates" that they are indeed correct, since this is not guaranteed for various reasons, such as TM quality (very important, as mentioned by Gabriela...), context, etc. It is not guaranteed that any such errors are picked up in the ultimate processing step - proofreading - as this is done only on the target text. Our standard charge of 33% for full matches (and repetitions) reflects that the segment is processed by two out of the three parties involved.
It is of course feasible to exclude 100% matches and repetitions from editor processing and only include the final proofreading step, in which case we can accommodate a more "aggressive" discounting. It is a very common practice in the industry in order to optimize return on technology investments, and perfectly OK as long as it is understood that it involves some exposure to quality risks. We only recommend this approach if TM quality is known to be excellent and that a very tight process for updating and maintaining it is deployed.
It demonstrates a fundamental lack of knowledge of what the translation process entails when clients wish to pursue this course, even if rolled up under the banner of acceptable loss. Even when a claim of 100% match can be made, for the many different reasons cited here, you have to take surrounding context and subject matter into consideration. And the fact that the translator ultimately has to read through and make a judgment as to the efficacy of the existing translation in this new context indicates clearly that at the very least, you are proofreading/verifying the usability of the phrase, if not ultimately engaging again in the translation process.
If you bill on a two-tiered pricing structure (verification versus new translation), or the three-tiered structure championed above, and don't discount as much as the proponents suggest, then basing some degree of pricing differentiation on matches could be (but still probably won't be) justified.
Ultimately, the cheese maker is equally responsible for the Emmental as he is for the holes. You're paying for the results of the process, not just the final weight.
http://www.proz.com/forum/money_matters/213901-whether_to_translate_matches_over_above_95_sdl_trados_if_client_does_not_pay_for_this.html