Fuzzy match translation rate

[Kuma-kid]

Hey Yakkers,

Been a while since being on board the good ship Honyaku as I have been off selling other wares. I am back yakking and wanted to know if someone could help me with some rates I have been asked to submit:

Words Translation 50-74% Match
Words Translation 75-84% Match
Words Translation 85-94% Match
Words Translation 95-99% Match

I have submitted a standard word rate and offered 100%  matches at 30% of standard.

Can anyone give me some guidance (or simply offer reasonable rates) for the match levels listed above? 

Not so critical that I make so much as I am getting back in the swing of things. 

Thanks,

Bart

[Fred Uleman]

I am puzzled by this discounts-for-matches thing.

Almost all of the words in the stuff I translate are in my dictionaries. I have 100% matches in most cases. Should I be offering discounts for the words that are in the dictionary? Does this depend upon who paid for the dictionary?

- -- --- ---- ----- ---- --- -- -
Fred Uleman

[Christopher Frederick]

> I am puzzled by this discounts-for-matches thing.

I could be wrong, but I think that this is mainly relevant to
translators working with shared translation memories. Agencies that
pre-process all of the files to be translated and then farm out the work
to multiple vendors, for example, may very well discount matches so that
they don’t have to pay for them twice.

All that being said, I’m not sure how common these discounts are in
the absence of shared translation memories…

— Chris

[Christopher Blakeslee]

Fred

A 100% match basically means the appearance of the exact same sentence that was translated before and is in the translation memory, including the order of the words. Even in that case, the context could very well demand a substantially different translation. Depending on the algorithm used by the CAT tool, the skill and speed of the translator, and the different phrases within the sentence, an 80% match could either be very helpful, useless, or even have a negative impact on translation speed, as you waste time fixing the match rather than starting from scratch. This is a bigger problem when working from Asian languages without spaces to delineate words. Obviously, the quicker you can graduate from having to work on projects using match discounts, the better, but we all have to start somewhere.

Bart,

I started in this business long enough ago that I have no experience with having to give discounts for matches, so sorry I can't help, other than to say  give as little discount as possible and maybe zero discount at 80% or lower. But I feel your pain...

Best regards,

++++++++++++

Chris Blakeslee

--
You received this message because you are subscribed to the Google Groups "Honyaku E<>J translation list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to honyaku+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Susan Murata]

I would surmise that this is only relevant to large-volumes of repetitive text such as might be found in operation manuals that need to be updated.

 

I say this not being a CAT user, nor a translator of high-volume repetitive text.

 

It does seem to me a highly mechanical way to translate and not at all suited to my kind of scatter-brained approach….

 

Susan Murata

[Hào Anh Lê 黎英豪]

Website translation tends to use this sort of thing… Hào Anh Lê

[Richard Waddle]

Kuma,

Depends on currency (and translation memory integrity score), but my rates on the off occasion that I slum it, as it were, in the CAT tiered rates arena would be approximately as follows (just did a weighted average of my schedule to align with your buckets).
Words Translation 50-74% Match: 0.3499
Words Translation 75-84% Match: 0.5512
Words Translation 85-94% Match: 0.8212
Words Translation 95-99% Match: full rate

For Powerpoint, I apply a factor of 1.03372 (based on my past "time spent vs. Wrod" records).

HTH and Regard
R. Waddle

[Christopher Frederick]

Richard,

Words Translation 50-74% Match: 0.3499
Words Translation 75-84% Match: 0.5512
Words Translation 85-94% Match: 0.8212
Words Translation 95-99% Match: full rate

Shouldn’t this rate table be inverted so that lower match percentages correspond to higher compensation? i.e.

• (Presumably) 0% base rate for a 100% match
• 34.99% base rate for a 95–99% match
• 55.12% base rate for a 85–94% match
• 82.12% base rate for a 75–84% match
• Full rate for a 50–74% match
• Full rate for any matches less than 50%

I think “slumming it in the CAT tiered rates arena” is a good way to describe the situation. :)

— Chris

[Fred Uleman]

Should probably have indicated that I was being facetious. I understand the question but question the concept. You invest money in a translation memory system and then invest time in the learning curve, all so you can give clients a discount?

[Christopher Frederick]

> You invest money in a translation memory system and then invest time
> in the learning curve, all so you can give clients a discount?

Or perhaps it would be more accurate to say “…all so your clients
can expect you to give them a discount.” :/

— Chris

[Rene]

On Thu, Jan 12, 2017 at 1:27 PM, Christopher Frederick <hon...@frederick.io> wrote:

All that being said, I’m not sure how common these discounts are in the absence of shared translation memories…

— Chris

If the client does not send a TM, the job consists of "no match" segments plus possibly a number of "repetitions" (identical segments). Usually there is a match rate for repetitions, so yes, there are also match rates for jobs without TM.

Rene von Rentzell

[Rene]

There seem to be some misunderstandings here.

1) Match rates are only relevant for jobs done with CAT tools.

2) Of course "matches" do not refer to individual words, that would be ridiculous. Matches refer to SENTENCES or PARAGRAPHS (depending on the settings). I.e. if a sentence is completely identical with one in the database, it is a 100% match. (101% match if the context is also identical -- yes in the TM world we do have 101% :) ).

3) A match rate for a 50-74% match is nonsensical. Matches at that level take as much time as a new translation.

Rene von Rentzell

--

You received this message because you are subscribed to the Google Groups "Honyaku E<>J translation list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to honyaku+unsubscribe@googlegroups.com.

[Rene]

On Thu, Jan 12, 2017 at 6:56 PM, Fred Uleman <ful...@jpnres.com> wrote:

I understand the question but question the concept. You invest money in a translation memory system and then invest time in the learning curve, all so you can give clients a discount?

1) These days, some TM systems are free, and others are dirt cheap, so not much of an "investment".

2) If there are plenty of identical or near identical segments in large jobs, in the pre-TM days clients would ask about discounts; with a TM analysis one can quantify these things.

Rene von Rentzell

[Michele Miller]

I expect there was a big wink at the end of Fred's post.

I don't go in for this type of translation, but it's becoming more prevalent. I recently noticed an article about translation start-up WritePath (based in Taiwan) giving clients discounts for fuzzy matches etc (OP: See their website). When asked how the company could keep its pricing so low (client pays US8-10 cents/word) and still source quality translators, the CEO said: "… even professional translators may be willing to accept a slightly lower rate in exchange for a steady flow of work or to keep them busy when they are in between their bigger projects". Naruhodo. So I wrote to them to ask how much the 'slightly lower rate' would be, taking into account repetitive text etc.

Here is their reply:

 

Repeated terms will be paid the first time it shows up, and glossaries are deducted. Exact matches for a segment are also deducted from the final word count.

For example:

I will be reading a book today tonight at seven is a total of 10 words.

If we have a glossary for book, the payable word is 9

If we have an exact-match sentence, the payable word is 2 (only 20% will be counted as payable).

However, we will not deduct more the 50% of the entire file.

If there's a file where every single word and every segment is in the glossary and/or are exact matches, the most we will deduct is 50%.

 

Ewww.

 

Michele Miller,

Sydney

 

 

From: hon...@googlegroups.com [mailto:hon...@googlegroups.com] On Behalf Of Susan Murata
Sent: Thursday, January 12, 2017 4:36 PM
To: hon...@googlegroups.com
Subject: RE: Fuzzy match translation rate

 

I would surmise that this is only relevant to large-volumes of repetitive text such as might be found in operation manuals that need to be updated.

 

I say this not being a CAT user, nor a translator of high-volume repetitive text.

 

It does seem to me a highly mechanical way to translate and not at all suited to my kind of scatter-brained approach….

 

Susan Murata

 

From: hon...@googlegroups.com [mailto:hon...@googlegroups.com] On Behalf Of Christopher Blakeslee
Sent: Thursday, January 12, 2017 2:32 PM
To: hon...@googlegroups.com
Subject: Re: Fuzzy match translation rate

 

Fred

 

A 100% match basically means the appearance of the exact same sentence that was translated before and is in the translation memory, including the order of the words. <snip>

 

Bart,

I started in this business long enough ago that I have no experience with having to give discounts for matches, so sorry I can't help, other than to say  give as little discount as possible and maybe zero discount at 80% or lower. But I feel your pain...

Chris Blakeslee

 

 

On Jan 12, 2017, at 12:27 PM, Christopher Frederick <hon...@frederick.io> wrote:

I am puzzled by this discounts-for-matches thing.

.

---

This email has been checked for viruses by Avast antivirus software. www.avast.com

[Christopher Frederick]

Rene von Rentzell:

> If the client does not send a TM, the job consists of "no match"
> segments plus possibly a number of "repetitions" (identical segments).
> Usually there is a match rate for repetitions, so yes, there are also
> match rates for jobs without TM.

Ahhh that makes sense. You may be asked to provide a discount for
repetitions, but not necessarily for 100% matches of previous
translations you’ve done.

I haven’t yet encountered this myself, but in a situation like the one
described above is it common for clients to request the project’s TM
in addition to the translated document?

Michele Miller:

> I expect there was a big wink at the end of Fred's post.

I expect the same. ;)

— Chris

[Fred Uleman]

Thanks for the discussion so far. Hope it is helping the OP with his question.

I particularly enjoyed the juxtaposition of Rene's:

"Of course "matches" do not refer to individual words, that would be ridiculous."

and Michele's quote from the Taiwanese company:
"For example:

"I will be reading a book today tonight at seven is a total of 10 words.

"If we have a glossary for book, the payable word is 9."

The smaller the segment (e.g., individual words such as "book"), the less meaningful the "match" is. We all know the alphabet. We all know the individual words. It is how we fit these things together that adds value. Even at the sentence or paragraph level, sentences and paragraphs that are not fit to flow well do not make a good translation (as we all know). That has to remain our focus. Not "are the words right?" but "does it work?"

[Rene]

On Fri, Jan 13, 2017 at 6:21 AM, Christopher Frederick <hon...@frederick.io> wrote:

I haven’t yet encountered this myself, but in a situation like the one described above is it common for clients to request the project’s TM in addition to the translated document?

It happens. However in that case the TM only contains the project. If you mean they are asking you to send a personal TM, containing other jobs, no. That does not happen, for many obvious reasons.

Rene von Rentzell

[Rene]

On Fri, Jan 13, 2017 at 8:43 AM, Fred Uleman <ful...@jpnres.com> wrote:

and Michele's quote from the Taiwanese company:

That Taiwan company is nuts. I have never heard of such a way of counting. Their idea of "match" has nothing to do with the way it is used in the CAT scene.

Rene von Rentzell

[Brian Watson]

Is it common?

The scenario I have encountered involves clients providing a TM for use with my tool. My work on the project updates that TM, specific to the project or client, and I include it in my deliverables.

Using CAT tools saves me time. I'm happy to use it for the right jobs.

Brian Watson

[Rene]

On Fri, Jan 13, 2017 at 4:25 PM, Brian Watson <brian....@gmail.com> wrote:

The scenario I have encountered involves clients providing a TM for use with my tool. My work on the project updates that TM, specific to the project or client, and I include it in my deliverables.

That is the standard procedure. And frankly, reading that in 2017 there are still people who do not use some sort of CAT system is like finding a cargo cult living in your basement, LOL

Rene von Rentzell

[Kuma-kid]

Love it!! Will pass this on to SDL (purveyor of Trados and translating robots who also make mean margaritas to soothe the savage client)

[Kuma-kid]

Thanks Chris and while I agree with your advice my principles of tech-ceptance and client brown nosing require me to respond reasonably. I know I saw LSPs charging clients 30-50% for 100% matches. I think offering one or two added rates would make them warm and fuzzy.

[Kuma-kid]

Thanks Susan. It is brain dead but I need that flavor to spice up my revenue basket. All these haiku translations are akin to eating fois grax all day long ...

[Kuma-kid]

Thanks Waddle-san! What is HTH?

substantive post reply to Christopher's below.

[Kuma-kid]

As for slumming it I can only say that one man's corrugated tin pan is another fiddler's roof.

The point here is not to earn less but present rates that allow me to make the $1million per hour that I am used to making (ha ha). So, my rates are ones vetted to allow my hourly income to measure up by ACTUALLY translating the given clients material NOT those I present at quote time. I will quote a chicken sam 'wich if needed to win the job with the caveat that I will revise rates after receiving client material. Even those revised rates are subject to change if the client changes their material. Seeing as most clients are so disorganized they always change or fail to meet my conditions, I rarely find myself in a bind or charging rates that earn less ... in fact I dare say on these things I earn more. Occasionally I will outsmart myself and  しょうがないね！

As a translator, LSP salesman, LSP PM, law firm translation coordinator, eDiscovery reviewer/manager/consultant over the last 20 years I have learned to roll with the zaniness of the LSPs and clients who are just trying to hold on to their project budgets or sales wins. The wise money gives 'em what they want with escape conditions that are reasonable enough not to raise eyebrows. 

I appreciate, and like :), the rates you have been so generous in providing to me. I disagree about offering anything for free like 0% discount for exact match. I believe it is referenced in a post below but there is "fitting" to be done. I call this the SAT name-game. The US SAT test awards 12.5% of the score simply for the brain dead function of filling out one's name correctly. Little work for a large percentage of score disproportionate to that trig alogorithm problem based on 3 polish lawyers riding a NYC skyscraper elevator and getting off and on while serving ham sandwiches or something. Ya know? Just reading the match, confirming  match and shrugging while I grab another cup of Joe takes enough time. I see LSPs charging up to 50% for repeats and see no reason why I cannot charge at least 33% for them.

Appreciate your post and further discussion. 

Cheers,

Kuma Kid

[Kuma-kid]

Thanks Fred - nope, its so you can get these stingy clients into your income stream ... no 石庭　stone left un-silently contemplated.

[Kuma-kid]

Roger, Rene. Thanks.

[Kuma-kid]

ありがとうございました！

On Thursday, January 12, 2017 at 10:12:55 AM UTC-8, Rene wrote:

To unsubscribe from this group and stop receiving emails from it, send an email to honyaku+u...@googlegroups.com.

[Kuma-kid]

Ｌｏｖｅｌｙ　ｉｎｄｅｅｄ．　Ｔｈａｎｋｓ！

[Kuma-kid]

Agreed Fred but I'd take it further. It must be fitted and one must be able to fit it, get paid (in this case 50%) and to have finished the work in half the time of a normal translation. I do a lot of patents where fitting can be easier than literary or marketing pieces, although errors are costly and damning. I would say that CAT does not meet business case specs for use in many types of content. Go read a Toyota Prius manual if you can wihtout screaming and realize that as poor as it is it gets you what you need when you flat a tire in the desert and need to use their Goo injection system. This is CAT at its most applicable. Believe me, though, Toyota did not sell the car basd on CAT but, rather, on Primo Yakker delicacy output.

[Kuma-kid]

Holy Sheepshank ... let me check downstairs

[Robin Boyd]

I've been intrigued by this billing practice for a while. It sounds reasonable in theory and I may have to start offering it in order to keep my clients at some point. However, I'm not sure of the logistics of it. Is there a feature in Trados that gives you a report on the number of characters in each type of fuzzy match so that you can calculate rates such as this? I use software called Felix, which can read *.tmx files, but doesn't have such a feature.

[Geoff Trousselot]

> On Jan 15, 2017, at 4:47, Robin Boyd <robin...@gmail.com> wrote:
>
> I've been intrigued by this billing practice for a while. It sounds reasonable in theory and I may have to start offering it in order to keep my clients at some point. However, I'm not sure of the logistics of it. Is there a feature in Trados that gives you a report on the number of characters in each type of fuzzy match so that you can calculate rates such as this? I use software called Felix, which can read *.tmx files, but doesn't have such a feature.

Normally, the translator would receive a project order that lists the number of characters for each matching categories with the amount to be paid for each category. The key points to look for would be the total amount awarded, and the amount for 100 % matches. Then the total number of characters and the number of 100% match characters.

If total amount of characters is 60,000 and 100% matches is only 4,000 of those char, then erroneous quoting would lead to either not getting the job or getting totally screwed.

Usually rates are at the bottom of the pear-shaped range of industry rates, so it's a different type of translation. It's about speed vs quality, where quality is what is acceptable to the client and not you personally.
Sometimes, if proofreading is half the rate, that can be more profitable, even if 60% needs rewriting. I only know Trados and Felix, and Felix is not the tool for such jobs.

Personally, I work too many hours a week to be able to tolerate the personal compromise of quality, and the mechanized approach. There is no time to get involved in the project, and work becomes an issue of concentration and endurance.

Geoffrey Trousselot.

[Geoffrey Trousselot]

Normally, the translator would receive a project order that lists the number of characters for each matching categories with the amount to be paid for each category. The key points to look for would be the total amount awarded, and the amount for 100 % matches. Then the total number of characters and the number of 100% match characters.

Sorry, I somehow was thinking 100% of full rate, or 100% new translation. The translator's mind. Only caring about what needs to be done. Apologies for confusion.

Geoff 

[Robin Boyd]

Thanks Geoffrey

 

 I only know Trados and Felix, and Felix is not the tool for such jobs.

I haven't used Trados, but that was definitely the impression I got about Felix. I've found it to be a helpful tool, but I don't think it would work well for this kind of job, which is one of the main reasons I've always refused a rate scale like this up to now. 

 

Personally, I work too many hours a week to be able to tolerate the personal compromise of quality, and the mechanized approach. There is no time to get involved in the project, and work becomes an issue of concentration and endurance.

I agree, and that is definitely the case for me right now, but I can foresee a future where I might not have a choice. I was absolutely horrified to receive a request like this from my main client, which provides around 80% of total work at the moment. If I were to lose them, I would be in trouble. 

[Andy]

Hi all,

 

Having used Trados, Wordbee, and other CAT tools for 15 or so years, I would say the benefits of this approach outweigh the disadvantages for both translator and client by a wide margin. This is especially the case when using browser-based systems.

 

Among other things, you can communicate very quickly (real time in some cases) with other translators, editors, managers, and subject matter experts who are working on the same project. There are references at your finger tips, and many that don’t require even that (a good setup will have most terms, phrases, various matches appear automatically). You can also easily search through memories, in some cases hundreds of millions of translations of the same type.

 

Everyone involved also knows in advance the charge for a translation, making it easy to accept or reject. It is also easy to split jobs up on the fly in case it turns out too big for one person.

 

Used properly, these systems do not inhibit creativity, although they make standardization—a key to successful technical translations—much, much easier.

 

I must admit when I get text written on paper and have to type the translation out in a Word document, there is a nice bit of nostalgia. But after page 1, I go to the copy machine, OCR it, and use the CAT tool.

 

Andy

San Francisco, CA

[Rene]

On Sun, Jan 15, 2017 at 4:47 AM, Robin Boyd <robin...@gmail.com> wrote:

I've been intrigued by this billing practice for a while. It sounds reasonable in theory and I may have to start offering it in order to keep my clients at some point. However, I'm not sure of the logistics of it. Is there a feature in Trados that gives you a report on the number of characters in each type of fuzzy match so that you can calculate rates such as this? I use software called Felix, which can read *.tmx files, but doesn't have such a feature.

Analysis are made for source files (e.g. Word). TMX files are memory files. I thought all CAT programs have an analysis feature, I am surprised that "Felix" lacks that.

And I don´t really see what the big deal is. In the pre-CAT days, clients would redact files or haggle for a discount. With a CAT analysis, you can quantify what you are looking at.

And of course it everybody tries to get a better deal -- it is human nature and the nature of a market place. What is the big surprise?

Rene von Rentzell

[B. Hyman]

Robin,

You might want to try using this tool.

http://ginstrom.com/AnalyzeAssist/

From the website:

Analyze Assist analyzes source documents against your translation memories, and reports the numbers of matches (exact and fuzzy) that it finds. You can use this information for estimation and scheduling purposes.

HTH,

Brian Hyman

On 2017/01/15 4:47, Robin Boyd wrote:

I've been intrigued by this billing practice for a while. It sounds reasonable in theory and I may have to start offering it in order to keep my clients at some point. However, I'm not sure of the logistics of it. Is there a feature in Trados that gives you a report on the number of characters in each type of fuzzy match so that you can calculate rates such as this? I use software called Felix, which can read *.tmx files, but doesn't have such a feature.