Velior's Corporate Blog about Translation and Translation Industry

This post offers a few best practices for translating uneditable text. The approach to handling this type of text depends on the source files:

1. Many uneditable texts are scanned copies of poor quality. A good example is medical records written in Russian. Russia-based healthcare professionals often develop illegible handwriting, because they handle too much paperwork daily in the environment that is still largely paper-based. For this reason, their written opinions or prescriptions are often Greek to the people outside of medical profession, including translators. Ideally, we prefer to use a healthcare professional to “decipher” such records, instead of translating directly from the scans. This eliminates hours of guesswork otherwise spent by a translator.
2. Another common uneditable text type is a PDF file printed from a desktop publishing software or MS Word. In such cases, we ask the client to provide the underlying editable file such as INDD with Adobe InDesign, because handling this editable file requires much less effort than translating the uneditable PDF. Typically, the underlying file is available and can be easily provided by the client. A common reason for the client to initially provide a PDF file instead of an INDD file is simply being unaware that the latter can be translated more easily, leading to reduced costs.
3. Where the items 1 and 2 above are not applicable, we resort to the optical character recognition (OCR), recreating the uneditable content in a separate text file. Creating an editable file enables our standard translation workflow, which is built around processing bilingual files. Where any complex formatting is required, we add the cost of OCR to the total amount billed to the client: this service is called “recreation of formatting.” With smaller or simpler files, we may do the OCR free-of-charge or bypass this step by translating directly from the source file.
4. Sometimes, the client would complete an automatic OCR at their end and then request translation of the resulting editable file. With the files that include complex formatting such as tables or figures, the automatic OCR tends to provide poor results, and the OCRed file may still need many manual adjustments. In such cases, we may ask the client to compensate the “recreation of formatting” anyway, despite the theoretically editable source file.

This post discusses Velior’s policy on the translation memories created by someone else. It applies to the situations where a translation job comes together with an existing TM, i.e. another translator worked on it previously, and it is new to us. In such cases, we prefer to calculate all words as new despite the TM leverage, or at least increase the standard 100% and fuzzy match rate. Here is why:

1. Most importantly, we cannot be sure about the quality of the existing translations. This represents a risk of spending uncompensated time on reviewing and correcting them. We cannot afford to take such risk free-of-charge.
2. Our opinion of the external TMs is largely influenced by Velior’s internal quality policy. For example, in our own work, we minimize the risk of mistranslations and omissions by using an independent editor in each job and eliminate the risk of terminology inconsistency through an automated QA. Because of this policy, we expect the same level of quality from the external TMs, and where they fail to meet these expectations, we have to spend additional time on editing them.
3. While considering a job involving an external TM, you also need to evaluate the TM quality. This means spending additional uncompensated time. If the job is assigned to us after all, I don’t have any objection to this. If the job is however placed somewhere else, I personally feel disappointed because of the time wasted in such an unproductive way. Reading someone else’s irrelevant old translations, that is.
4. Whenever you re-use a fuzzy match in a new segment, you have to review or at least read the existing portion of the translation. The typical TM-related rates are however geared towards paying for translating the new text only. With the external TMs, the effort invested in reviewing the existing portion is therefore either not compensated at all or compensated only partially.
5. If you encounter any inconsistency issues in the provided TM, you have to check the entire current job for such issues, including the 100% matches. This additional work is not covered by the typical TM-related rates. Simply put, you are paid to only review the existing translations briefly, but not to correct any errors. The provided TM is naturally expected to be correct, otherwise why provide it all?
6. Quite a few external TMs we worked with included translations that were completely unusable due to the technical issues such as a wrong encoding, damaged special characters, or incorrectly aligned translations (probably aligned by a non-native speaker). Such problems normally don’t surface until you delve deeper into the TM in the course of translation—already after accepting the job. You end up re-translating them at your expense.
7. The writing style used in the previous translations is often different from your own, but you are expected to follow that style to maintain consistency. This normally requires an additional effort, e.g. you need to specifically check that you did not use your own style unconsciously. Again, this is not compensated by the standard TM-related rates.

As many other professions of today, translation is tightly interconnected with the Internet. Some of the ways translators use it in their daily work include researching the subject matter to understand the original text and finding the commonly used target language terminology. But which websites can be considered trusted sources of reference materials? I’ve recently worked on an English to Russian translation concerning innovative disease treatment methods and referred to the Internet extensively to find terminology that is not in the dictionaries yet. I was however frustrated with the low efficiency of my research, because the search engines continuously yielded tons of unusable results. This post summarizes my general opinion on this matter.

The average quality of the texts I scan through while searching for the reference materials on the Internet is decreasing. The quality is diluted, because the useful content is just a fraction of the total content volume added daily. I can think of two potential reasons behind this:

1. As I use the search engines to research Russian terms, I increasingly stumble upon the automatically generated texts. A major source of such content is the machine translation (MT). The webmasters simply feed English (or other language) content into the MT engines and publish the resulting Russian text without any editing. Another significant source is the software that generates “unique” content by taking any text and replacing as many words as possible by synonyms. As seen by the search engine crawlers, this content is indeed unique, but for the human users and especially translators, who look for the commonly used terminology, it is of no practical use.
2. The abundant human translations also contribute to diluting the content quality. While the Internet is certainly home to many well-crafted translations, a huge portion of translated content gives rise to concern. In particular, I find it hard to trust the terminology. For this reason, I rarely use translations as the reference materials. The odds of a literal or incorrect translation are just too high for me to take the chances. I’d rather invest more time in finding texts written by the subject matter experts using their native language.

In the light of these observations, I am often surprised by the criticism of Wikipedia. I do realize that it’s not ideal, just as nearly any other crowdsourced initiative. But, as a whole, it might be one of the best general knowledge resources we’ve got on the Internet, simply because there is no other website that provides the same amount of reliable content under one roof. The additional benefit is the well-structured, easy-to-use GUI. Naturally, the translators should not use it as their only source, but at least it’s a good place to start-you can develop a basic understanding of the subject matter and then continue researching by using the available links or keywords.

You can read more about how the Internet helps translators in this post.

While reading The Associate by Jonh Grisham, I came across a thought that resonated with some of my professional experiences as a translator, despite the different settings. The main character, a young law school graduate who has just started at the bottom of a large law firm, works in the corporate office basement reviewing documents the entire day. He soon realizes that, although the client is being billed $300 per hour, this work is completely meaningless and will never be used. He ends up thinking: “At that moment, his father was either in his office counseling a client through a problem or in a courtroom mixing it up with another lawyer. Regardless, he was with real people in real conversations, and life was anything but dull.”

Now, why does this situation resonate with me? Just as the main character, I sometimes wonder whether my work will be used at all and feel like there is little connection between what I do and the end users. Below are a few general translation industry observations explaining why I feel so occasionally:

1. Clients sometimes view translation as a mandatory, but actually useless provision rather than an enabler and business asset. A great article on this has been written by Terena Bell. As a consumer, I am strongly opposed to this approach, because, even though I understand English and German, I still prefer products that come with documentation and interface correctly translated into my native language. Let’s use the currently popular HD media players as an example. With these players, all user interaction is typically happening through a graphic user interface (GUI) on a TV screen. If I buy a media player with a well-localized GUI or read a review that praises excellent translation, I will likely become a loyal customer of this company, which makes me a potential buyer of their next generation products and a supporter who might recommend them to others. On the contrary, when I encounter a faulty GUI, I might lose the respect for the manufacturer and will certainly discourage those, who do not understand the original language and will have to use the localized version, from buying the product.
2. Many translators collaborate mainly with agencies and don’t get to know or work with the end users of their translations. Also, most translation assignments are handled over the Internet these days, taking the personal touch out of the equation. For me, working with the end client directly is a completely different experience. When you meet the client in person and discuss why they need your services and what their expectations are, you feel that your work is important to them. Knowing exactly how the client will benefit from your services helps you feel more connected to the real world and “real people.”
3. With about 90% of our work, there is no feedback whatsoever from the end users. This sometimes makes me think that no one even reads those translations and increases the feeling of being detached from the real world.
4. Quite a few translations, in my experience, don’t make it to the end users, because the project is being canceled halfway through. For example, a few web-sites Velior had translated in the past never made it to the web. In such cases, the translator’s work is again useless. One reason could be that the clients are just unsure whether they need a translation at all, and after a second thought prefer to cancel the request, even if the translation is already at the editing or desktop publishing step.
5. Some translations are requested for reasons other than making them available to a target audience. For example, legal documents are often translated for compliance purposes only, not for reading. Just as Grisham’s character’ work assignment, they simply need to be done, but are of no practical use.

The good news though is that with most translations we do at Velior I feel that our translation is important to the client and the effort we put into delivering high quality will be appreciated. This is why I specialize in and enjoy medical translation such as medical device manuals: their users are always supposed to read the documentation to make the best use of the products, because other people’s lives might depend on this. Knowing that quality is important to them motivates me to continuously improve and excel in what I do.

Recently, I have covered the topic of outsourcing typesetting work in translation jobs, mentioning that the original translator should be preferably used at the proofreading step. In this post, I would like to discuss this recommendation in more detail.

Place of Proofreading in Translation Workflow

Due to the role distribution in this industry, translation vendors sometimes have little or no control over the typeset translation. In a common multilingual project scenario, a translation agency requests translations from its single-language vendors, inserts them into the final files, typesets these files, and arranges proofreading. Here is a schematic view of this process:

Source text preparation (DTP specialist) => Translation (SLVs) => Typesetting (DTP specialist) => Proofreading (?)

Why Proofread at All?

Is the proofreading step at all necessary? The answer is yes, absolutely. Without proofreading, the typeset files may contain so much layout issues that even a top-notch translation will not mitigate the pain. The next logical question is who should be responsible for the proofreading. I think the choice depends on the workflow design, agency’s flexibility, and certainly budget. The potential proofreaders are listed in the order of preference below:

1. Original translator
2. Original editor
3. Any person with linguistic background, ideally a native speaker of the target language
4. DTP specialist

Ideally, this task should be assigned to the original translator. Generally, this translator possesses more knowledge about the translation than anyone else. They will identify issues more easily and efficiently, because they already know the potential pitfalls. Even the original editor is less efficient in this role, because of the reduced involvement in the translation as compared to the translator. Below are a few examples from our experience in support of this approach.

Issues Eliminated through Proofreading

1. Twice in my career, I have seen agencies deliver large (over 100 pages) manuals to the end client without any typesetting and proofreading (the latter could have helped bring the layout issues to the agency’s attention before delivery). Both times, the end clients were furious about the layout, which was basically non-existent. I have been always using these examples to discourage our customers from skipping the proofreading step.
2. With many language combinations, English to Russian translation being one of them, the translated text is usually longer than the source one. Inserted into the final files and typeset, it often does not fit into the available space, which results in cropping or overlapping. Such issues are hard to identify, unless you know exactly which translation is supposed to be in this or that place of the document or proofread the entire document carefully (requires knowledge of the target language).
3. Certain issues such as different word order in translation due to split source sentences cannot be dealt with directly in the translation files. These should be checked and corrected in the typeset final files.
4. A DTP specialist may forget to extract the text from images for translation. Such untranslated text is routinely identified during proofreading.
5. A DTP specialist may also skip numbers, especially in tables, while they often do need localisation such as changing a decimal point to decimal comma (e.g. 0.01 to 0,01) or changing thousand separators (e.g. 50,000 to 50 000). Again, proofreading helps eliminate such issues easily.

For more information on proofreading, you can read our article “Proofreading, Editing, Review, Independent Review.”

Tight deadlines are quite common in the translation industry. The reasons behind them certainly vary and are generally justified such as an urgent need for translation because a truck is detained at the customs due to untranslated bill of lading. However, my experience with English to Russian translation suggests that deadlines sometimes might be unreasonably short. For instance, there have been situations when we were given just 2 days to translate about 6,000 words (while the norm is about 2,000 words per day) and then waited for weeks before the typeset or reviewed files arrived. Another example is website localisation: given the complexity of the process, especially when multiple languages are involved, it is quite lengthy. From time to time, I check with the sites we translated recently to see how our translation was implemented, and from what I have seen so far, the average time between translation delivery and implementation in the CMS is three months. Meanwhile, the time provided for the translation step of the process is typically just a few days or weeks, i.e. disproportionately short. Based on this experience, I strongly believe that reasonable deadlines are usually possible, and both clients and translators can benefit from them. This post is intended to discuss some of the reasons in support of this belief.

1. For me, quality has always been the most important reason. Translation is not a pizza delivery that can be sped up by simply asking the delivery man to ride faster and straight to your home, skipping other clients on the route. A translator can hardly shift into a higher gear without sacrificing quality. In particular, pressuring translators to deliver faster than they normally would leads to incorrect understanding of the source text, less attention to detail, reduced editing and proofreading time, which all add up resulting in lower quality. If you want a high-quality translation, it is recommended to let the translator establish a deadline they find reasonable to do a good job.
2. We normally add a certain percentage of time to each quoted deadline to account for potential delay risks such as a translator getting sick, power outage, or hardware failure. Obviously, when such risk occurs while you are translating urgently (and therefore have no “just-in-case” time), you are bound to miss the deadline. And no matter how pressing the deadline was, this will still cause damage to your reputation. In other words, the risk of late delivery, hence reputation damage, is much higher with the tight deadlines.
3. Velior is usually likely to provide discounts for jobs with longer deadlines, as they increase our flexibility by enabling long-term planning and handling other jobs within the same time frame. The results are reduced downtime and higher efficiency which can be normally reflected in a discount to recognise and encourage the client’s flexibility.
4. As a company, we also need to develop new translation talent, including graduates and interns. Involving them in actual jobs at the translation step (of course such translations are subject to thorough editing) is an important part of their individual development plan. Tight deadlines greatly reduce such on-the-job training opportunities, as these translators need more time than the seasoned ones. This hinders talent development in the translation industry to a certain extent.

In addition to a reasonable deadline, a translation can also benefit from a higher price. You can read why in this post.

In a recent post, I explained how a higher price might have a positive impact on translation quality. However, the current trend in the translation industry is the price reduction. This topic is now hot in the blogs and translation forums (e.g. see posts by Wendell Ricketts or Jill Sommer), as translators are increasingly pressured to lower prices. Here is my two cents worth:

Our Situation

Indeed, more customers are now interested in discounts. Some also insist on applying a standard minimum rate to each translation instead of negotiating the prices individually, depending on the difficulty or deadline. As a translator, this is certainly frustrating for me. I do believe though in the supply and demand rules of the free market. If the buyers have an increasing demand for cheap, lower-quality translations, they should be able to buy them. I think this is a natural trend in the sagging economy and therefore see no reasons to resist it specifically. If the end users accept this quality, why pay more? Otherwise, the company that buys and uses such translations will lose some of the business due to low quality, review the lessons learned, and proceed with more caution next time.

Recent Quality Issues

Unsurprisingly, we are now receiving more requests to re-do someone else’s work after it was rejected by the client. One of them was a large, almost 100K-word translation. Receiving the client’s feedback on this one is something I would never want to experience as a supplier. Personally, if some day I am forced to always deliver such work and face the risk of similar embarrassment, because I cannot get any better-paid projects and therefore have to translate more to maintain profitability, I’d rather quit this business.

We have recently translated a medical user manual from English to Russian. This job involved a translation memory created by someone else. In many 100% matches, the word “haptic” (element supporting intraocular lens) was translated as “sclera lens.” While this is an outrageous error itself, it is the fact that it still resides in the TM, that is more alarming. This error was once made by a translator and then missed by an editor and proofreader (or were there any?). It might have been spotted by the end users, but no one cared about correcting the TM, hence about the quality of any future jobs. Or, this manual might have never been read, so high quality is just unnecessary.

The other day, I proofread a Russian to English translation made for a large global client. While I had my share of bad translations as an editor over the years and developed a thick skin, this nightmare penetrated my low sensitivity threshold easily. It was literal and unprofessional beyond readability, with the meaning of many sentences escaping me completely. I guess the only reason why such translations make it so far in the supply chain is the low price.

Our Policy

We did not decrease the prices and are unlikely to do so. As the cost of living in Russia continues to increase drastically each year, our rates are bound to increase only. Regardless of the current trend, we will continue to pursue quality and serve our clients to the best of our ability.

This is the 2nd part of the post. Part 1.

Reasons Not to Charge Full Rate

3. Translators who prefer full compensation also maintain that they invest a lot in translation memory software and related training to make it possible to process 100% matches and repetitions with less effort. Because of this personal investment, it is they who should reap the full benefit, not the client. As a translator myself, I think that they are correct to some degree, but the reality dictates otherwise:

• This logic could apply when the TM tools were just introduced to the market. Now, they are an inherent part of the translation industry. A translator is often forced to buy them to work on the specific types of jobs, because otherwise they will not be able to accept such jobs at all. It is not therefore a genuine investment in the sense that you put your own money into optimising your own work, but simply a mandatory requisite just like a PC or phone. Trying to keep all benefits of this software to yourself by labeling it as your investment means resisting progress in the industry where many peers and clients use it routinely. It is now a common technology, and, as such, its benefits should be shared by all parties to the translation process. For example, many clients put additional effort into writing consistently and should share the resulting savings as a compensation of this effort. (I understand though that some translators do, and will continue to, charge full rate for individual and completely justified reasons.)
• The clients become increasingly aware of both repetitions in their documents and the translation memory capabilities, therefore expecting discounts, as they do not see why they should pay the same price for identical text. And I find this fair, because it is a healthy consequence of the technological progress.

What if Client Prefers Not to Pay

1. Generally, Velior requests that 100% matches and repetitions be compensated. When a client asks to automatically process them without checking, there is often a risk of mistranslations, and we do not want any flaws in our translation, even if they happen through no fault of ours. For this reason, such requests are honoured on an exception basis only.
2. In these situations, we ask the client to extract the 100% matches and repetitions into a separate file, which includes only a single occurrence of each match/repetition. After the translation, the client can use this file to assemble a final translation. Again, there is always a high risk of mistranslations. Normally, such assembled translation should be proofread. Otherwise, the quality will likely decrease.
3. Translation agencies also use the “Perfect Match” technology to insert 100% matches only when the current context is identical to the context, in which they were originally translated. While this technology indeed brings client-side efficiency and cost-effectiveness, it often has drawbacks from a translator’s perspective. The translator normally needs to have control over the entire text, while this technology lets them work just on parts of the text. They do not see a “big picture” and also cannot change the existing translation (either because it is blocked, or they have little motivation to do so, as this work is not compensated). This reduces the translator’s efficiency and quality.

Notes

1. Generally, Velior does not provide discounts for any matches with the translation memory created by someone else, because we cannot be sure about the quality of those translations. They may contain errors that will reappear in the current translation and eventually become our responsibility. Such jobs are subject to negotiation on a case-by-case basis.
2. The match and repetition discounts do not apply to editing or proofreading, because these services are designed to check the entire text, hence matches and repetitions give little or no time savings.
3. Interestingly, our main language combination, English to Russian, yields much more repetitions than the Russian to English translation, which can be mainly attributed to more complex and diverse grammar and punctuation rules in the Russian language.

This post discusses Velior’s approach to charging for 100% matches against translation memory and internal repetitions. Generally, there are three opinions on this matter. Two of them are opposite: some believe that these matches and repetitions should be compensated fully just as any other words, while others maintain that no compensation is necessary, since processing these words does not require any specific effort. The third opinion, shared by Velior, is neutral: a reduced rate should apply to these words.

Reasons to Charge for 100% Matches and Repetitions

1. The context where a 100% match or repetition occurs may be different from the context of the original translation. It is therefore important to check whether all such words fit the new context. The time spent on checking should be compensated. Processing 100% matches and repetitions without compensation and hence without checking is justified only when the client is absolutely positive that the context is always identical, e.g. in a data sheet or parts catalogue. However, in our experience, such texts are rather the exception than the rule: they account for about 3% of our workload (translation from Russian to English, English to Russian, German to Russian).
2. 100% matches and repetitions often require the same amount of typesetting work as any other words. Even when 90% of your text are repetitions, you might end up spending just as much time on formatting them as if they all were new words. And unless this type of work is compensated separately, it should be included into the repetition rate.

Reasons Not to Charge Full Rate

1. Because the amount of work associated with 100% matches and repetitions is normally reduced as compared to the new words, it does not make business sense to charge full rate for them. You do not have to translate them from scratch, but just check whether the previous translation fits the new context, and it usually does. Truth be told, sometimes such words do require adjustment or even complete re-translation. For instance, an English to Russian translator may render the word “position” as “должность” when it means job position. This translation however will be of no use as a 100% match in the context where “position” means simply location and should be translated as “положение.” Yet, as I mentioned above, this happens to a very low percentage of 100% matches and repetitions, and a discount is therefore completely justified.
2. Translators who prefer full compensation for 100% matches and repetitions often refer to various examples from other industries to explain that a discount for repetitive work is unnatural. For instance, in the construction industry, you will not push the construction crew, which is building your new home, for a discount on the basis that one room is an exact match of the other. While this idea might have some truth to it theoretically, in real life, it is the actual amount of work/time/effort associated with the repetitive task that is important here. The new room might be indeed identical, but the amount of work required to finish it will still be the same. In contrast, the amount of work associated with 100% matches and repetitions is usually much lower as compared to the full-rate new words. Normally, the translation memory software inserts them automatically, and you only check whether the previous translation fits the new context. It is therefore fair to charge a reduced rate reflecting the actual time spent on checking instead of charging the full rate as if these words required the same amount of effort as the new ones.

This post is continued here.

In February, 2010, one of our regular clients enlisted Velior’s help for an urgent translation that was intended for their English-speaking partner: it was assigned on Friday evening and expected before noon on the next day. Please follow the link to read this case study.