What's in a word
Michael Scholand
Technical Translator, Medical Translator, Software Localizer, Terminology Management Expert, Translation Technologist, Publisher
If I am to be honest, I have to acknowledge that I was very much tempted to entitle this post “F******* words”. Why? Because I am a translator. And because translators are paid by the word. And that makes me angry every time I think about it. Learn why.
What’s a word anyway?
Easy to answer, says my friend the linguist “A word is the smallest element that may be uttered in isolation with semantic or pragmatic content”. Wonderful definition, say I, but the trouble starts when you try to count the little buggers, because they come in all shapes and sizes. Let’s see a typical example of a text that we as a technical translation company translate by the million every year. Beware, this is not easy going, although it might be of use as bedtime reading if you have trouble falling asleep. In any case, you might have to read it repeatedly before it reveals its intricate meanings. The text is part of a car maker’s spare part catalogue.
3PT RR S/BELT(X2)+P/SEAT+CHILD RESTRAINT ANCHOR
FR POWER ST (LEATHER) + ST BELT REMINDER + A/BAG
LICENCE PLATE BRACKET+OVER FENDER,OVER FENDER (B)
RR TV MONI+O/GUIDE FRENCH+ENGLISH/FRENCH+AUDIO SYS
In addition to the semantic difficulties the text entails, how you count the words of such a text is of vital importance because it has a dramatic effect on what we as a translation company get paid for rendering it into another language. Well, according to MS word, this text has 26 words. Now, if you go about to put a space after all the characters in the text that are squeezed as separators between words (plus signs, commas) or used as elements that mark omissions (e.g. the slash in “O/Guide” for “operator’s guide”) the word count changes dramatically: 50 words!
This is just one, but very telling, example of how word counts have no relation whatsoever to the work that is necessary to translate the text. But the arbitrary use of separators or other characters in between words is only one part of the problem.
Why do they all count differently?
If you have ever wondered why word counts differ between different translation service providers, that might have to do with certain phenomena that appear in written language:
- hyphenated words: computer-assisted
- contractions: do not vs. don’t
- abbreviations, acronyms, units of measurement: e.m.p., f.s.a., mg/dL
- numbers, dates, mathematical formula, etc.
These words are counted differently depending on the the word-processing application or computer-assisted translation system used for quoting and there is actually no standard as to what has to be counted which way. But hang on, it gets even more confusing.
Donaudampfschifffahrtskapit?n
If you are a learner of German, you might have stumbled upon the Danube steamship company captain as an example of how long some German words can get. This one clocks in at 29 letters, but it is not only agglutinative languages like German, Hungarian, Finnish or Farsi that allow for the creation of long words. Take the famous “Supercalifragilisticexpialidocious” (34), or the wonderful “Antidisestablishmentarianism” (28) as proof that English can produce long words too. To cut a long story short, in certain languages and especially in technical texts, compound terms abound, but when it comes to paying the translator, they are only paid as one word, if, as is commonly the case in today’s business, the word count is based on the source language.
It’s characters we need
There are other, more accurate ways to measure the effort of a translator. One of them is the character or line count. A character is a character is a character, i.e. there is not much room for interpretation. The only question is if blank spaces should count as characters, and the answer is a straight “Yes, they should!”. After all, it takes the same effort to press the spacebar on the keyboard as it does to key in any character. Guess what the Google Translate API uses as its invoicing unit: characters, not some shoddy word counting.
In fact, invoicing by characters was the prevailing method for translation invoicing in German speaking countries before the Anglo-Saxon “furlong per fortnight” word counting system started to gain ground. This was caused by some of the bigger translation companies from the UK / US gaining share in the German-speaking market and the customers wanting to be able to compare the quotes of the newcomers to the old trades in town. In any case, the move to word counts is a terrible regression, an atavism. It’s as if we were to go back to measuring liquids by kilderkins or textiles by the ell.
It seems odd that an industry has adopted a non-standardized, arbitrary, undefinable means of quantifying the service it provides to its customers. Are there some hidden vested interests, is there a conspiracy? I don’t know but I think we would all win – language providers and customers - if we left the furlongs and firkins in the past and started to use characters instead.
Nothing is perfect
Counting by characters is not the answer to everything when it comes to measure the effort that goes into translation. Both word and character prices are a means of measuring an aggregate service: project management, translation & revision, proof-of-concept, terminology management, quality assurance, etc. all these and more activities or only some of them may be needed to produce a translation. As a buyer, you should always make sure what service level you are paying for when you compare quotes.
But counting characters instead of words would mean we gain transparency, accountability and comparability in the industry. And that you could call progress, right?
Investigador | OSINT | Docente | Anotación de datos | Traductor inglés-espa?ol | Lingüista
9 年I believe the standard shall be defined by each translator and we should have options so that the translator and the client can mutually, with the same leverage, decide what suits both parties the most. Then again, we all know that does not happen most of the time. In my opinion, words and characters should be an internal reference for us. As you point out, there's much more in the process of translation than just reexpressing one language into the other. It involves project management, translation & revision, proof-of-concept, terminology management, quality assurance. Unfortunately, many customers avoid to do this, as time is considered an obscure metrics that varies across the translation industry (some translators may charge higher for the hour and some may charge lower but take longer in doing the job). If some clients had a long-term mindset, the initial long hours that took translating the first pieces of work would be lowered on later works as we would have some references and we would be familiar with the terminology. However, this has a potential big downside: the ROI for some translators. This is mainly because of an internal debate that translators have failed to resolve over time: whether to charge for 100% matches and repetitions or not. For instance, if we took 15 hours to translate a document of 4,500 words one time and then, we only took 6 hours to translate a document with the same extension (4500w) as it had many repetitions and 100% matches from the first translation, and we had a standard rate of 35USD per hour, the client would have to pay $525 for the 1st document and only $210 for the second one, while that translator could have charged $525 for both separately if s/he charged by word. I believe we have to reflect on all these issues first, as characters is a measure that most clients won't even consider for now. The IT sector should be the exception for this matter though. As you well point out, wordcount for long strings is useless and does not reflect the work to be done. In this particular case, I believe it would be more than reasonable for both, the client and the translator, to charge on a character basis.
Disambiguation Specialist
9 年How about counting all the non-space characters and dividing by the space characters to get 'translation units'. (The word 'word' is too ambiguous as you point out.) That would take care of both the expensive, small (one letter but not spaces) end of the spectrum as well as the long (28 character). For example, if the letter 'P' (for 'Power' in 'Power steering') needs translating that should count. As for the word 'Antidisestablishmentarianism' you have that in a term store, right? That's a cheapy to translate.