Hi. I’m Mikhail, CEO at Lexicon Center. Visit my LinkedIn page below to connect, chat and see more posts.

Tips for self-checking translations 💎

This is what we ask our translators to do.
– After the text is translated, leave it aside for some time
– Compare the translation against the source for major issues (meaning, terminology, etc.)
– Leave the translation aside for some more time if possible
– Read the translated text without looking at the source
The last step is very important and can’t be skipped!

✅ It helps to spot punctuation and grammar issues and see if the text flows well and sounds natural. During this step translators usually make many changes which results in a high-quality target text.
P.S. The above steps should not replace standard editing by a second person. Even the most experienced and careful translators can make minor stylistic or major meaning mistakes.

How many websites and apps does it take to translate a document❓

Here are applications and websites our translators and editors use when working on clients’ projects.

1.      Trados Studio for translation
2.      Source PDF file for seeing the text in context
3.      Style guide for following client’s preferences
4.      End client’s website in English and target (if available) languages
5.      Client’s instructions
6.      Gramota for checking tricky punctuation rules
7.      ChangeTracker for making a list of changes between translated and edited files to give feedback to the translator
8.      Verifika for running QA
9.      Orfogrammka for checking spelling, grammar, punctuation, style, semantics, and typography
10.  Kartaslov for marketing and transcreation jobs to find synonyms, word collocations, associations, etc.
11.  Google — my favorite 😊

The list may be different depending on the language pair, project, market, client, personal preferences etc.

💎 Make sure your clients don’t treat you as 𝘫𝘶𝘴𝘵 𝘢𝘯𝘰𝘵𝘩𝘦𝘳 𝘷𝘦𝘯𝘥𝘰𝘳 and are aware how much time, knowledge, and effort you put into their projects.

A great way to start a new year is to renew the corporate ATA membership

I’m sorry we joined ATA too late — in 2018. We should have done it from the very start, back in 2008.

As an American Translators Association’s Slavic Languages Division group manager, I invite ATA members working with Slavic languages to join our LinkedIn group at 

Feel free to reach out to me with any questions you may have.

P.S. ATA is open to individuals and companies from all over the world, not just the USA.

Editing, proofreading, review, revision, check… Who can tell the difference❓

Here’s what the ISO17100 standard says about the terms:
–         Check: examination of target language content carried out by the translator
–         Revision: bilingual examination of target language content against source language content for its suitability for the agreed purpose
–         Review: monolingual examination of target language content for its suitability for the agreed purpose
–         Proofreading: examination of the revised target language content and applying corrections before printing
To make things even more complicated, there are two notes in the standard:
1.      The term bilingual editing is sometimes used as a synonym for revision.
2.      The term monolingual editing is sometimes used as a synonym for review.
I have seen the above terms often used interchangeably when an LSP wants a vendor to check the translation against the source for all sorts of mistakes. Many translation agencies that are ISO17100 certified sometimes mix up the terms too.
No matter what the client may call it, most often they want you to check the target against the source for all types of issues.
It would be a good idea to ask the client about their quality expectations and if they want just a check for “real” mistakes like meaning, grammar, and punctuation or would they also expect you to improve style and consistency. In the latter case several rounds of checking may be necessary to ensure top-notch results 📃

Does your translation occupy more space than the English source? It doesn’t always have to 👉🏻

When one translates from English into French, Spanish, German, Ukrainian, Arabic, Russian, Hindi and many other languages, the text can expand by up to 35%. The problem can be even more serious, if source text is wordy.
Here’s an example from a survey that we translated.
𝘞𝘰𝘶𝘭𝘥 𝘺𝘰𝘶 𝘮𝘪𝘯𝘥 𝘵𝘦𝘭𝘭𝘪𝘯𝘨 𝘶𝘴 𝘸𝘩𝘪𝘤𝘩 𝘰𝘧 𝘵𝘩𝘦 𝘧𝘰𝘭𝘭𝘰𝘸𝘪𝘯𝘨 𝘣𝘦𝘴𝘵 𝘥𝘦𝘴𝘤𝘳𝘪𝘣𝘦𝘴 𝘵𝘩𝘦 𝘢𝘨𝘦 𝘨𝘳𝘰𝘶𝘱 𝘺𝘰𝘶 𝘣𝘦𝘭𝘰𝘯𝘨 𝘵𝘰?
65 𝘰𝘳 𝘰𝘭𝘥𝘦𝘳
The above sentence works fine in English, but you don’t have to translate every word if there’s a shorter way to express the same meaning in the target language.

Here is how our linguist translated the question. 

𝘏𝘰𝘸 𝘰𝘭𝘥 𝘢𝘳𝘦 𝘺𝘰𝘶? ✅

It may seem too straight-forward for an English speaker but is totally fine in Russian surveys. The word-for-word translation would be over polite for a Russian language audience.

If the context, style guide, text domain and purpose allow, translations should be concise and straightforward.
This saves the reader time and helps them quickly and correctly understand what the author wanted to say without having to wade through words and word combinations that carry no useful meaning.

It helps the client achieve their goal too 💎

Is 3,600 words per hour a new standard ❓

I came across a research made by an LSP.
Over the years, the company was measuring how long it takes their best vendors to edit machine translation output. The takeaway is that the quality of MT is getting better and it takes less time for translators to do editing. No surprises here ✅
According to the research, now it takes a translator slightly over 2 seconds to edit 1 word of machine translation. However, the benchmark for a perfect translation that doesn’t need any editing is 1 second per word which makes 60 words per minute or 3,600 words per hour.
Is the “good old” standard (that we use at our translation agency) of editing 1,000 words per hour gone? Will a translator be able to understand context, do research, check meaning, terminology, consistency, grammar, punctuation and style at a pace of 3,600 words per hour?
According to The Editorial Freelancers Association (US) the median pace of monolingual copyediting is between 4 and 10 pages (1,000 and 2,500 words) per hour depending on the subject (fiction take less time, medical texts take longer). And monolingual copyediting has nothing to do with translation or texts in foreign languages.

A client sent one page for translation… 📃

It may seem like a piece of cake, but here’s what steps we had to take:

  1. Export the PDF file to Word
  2. Create a Trados project with the client’s TM and termbase to ensure consistency with previous jobs
  3. Do pre-translation analysis to understand the text, its purpose, and audience
  4. Translate the file. Google search didn’t take much time and there were no queries for the client this time. PHEW!!
  5. Have a second person edit the translation
  6. Check 20 false positive issues in 7 categories produced by the Verifika QA tool
  7. Save the Trados bilingual file to a Word document
  8. Check spelling and grammar in Word (just in case Verifika missed anything)
  9. Check 2 stylistic improvements the Orfogrammka tool suggested
  10. Adjust the textboxes and font sizes because the target copy occupied more space than the original
  11. Export the Word file to the original PDF format
  12. Compare the layout of the original and target PDF files

I’m not making a mountain out of a mole hill — some of the steps took a minute or less.
I’m trying to say that your clients need to know what efforts you make to deliver top-notch translations, so they value your work and stay with you 💎

Join ATA's Slavic Languages group 💎

As a new American Translators Association’s Slavic Languages Division group manager, I invite all ATA members working with Slavic languages to join our LinkedIn group.

SLD facilitates networking and sharing of knowledge through social media, quarterly SlavFile newsletter, and blog, as well as conference activities and a certification exam practice group.

Please use the below link ⬇️

Sometimes just one sentence is enough to understand the translator’s skills ✏️

Here’s what our translator had to deal with the other day. To be able to show you this example, I changed a few words because of the NDA with our client.

✏️ Michigan State Primary Care at Outpatient Care Commerce


What can go wrong here?
1.      Michigan State is not a state name, but a short name of an educational establishment (Michigan State University).
2.      Primary care is not a type of medical care, but a primary care facility.
3.      Outpatient Care, again, is not just a type of care, but an outpatient care clinic.
4.      Commerce is not a synonym for business, but a town (Commerce township, Michigan).


So, the correct translation to another language should be something like:
Primary care facility at outpatient care clinic at Michigan State University in Commerce township.

What a great time we had at LocLunch Monterey yesterday 👏

I reminisced about 1999 when I worked in Monterey County.

Stefan Huyghe shared some great examples from his article “Bet you didn’t know that much of your English is actually Dutch”.

Caroline Crushell talked about the Irish language and accent. I love how the Irish pronounce letter R!

Thank you, Paulina Perepelkin, for sharing your optimism about future.

We discussed how crazy it is to learn Russian, Polish, Ukrainian or another Slavic language because of cases and genders resulting in a huge number of different word endings.

We even discussed Carmel’s former mayor Clint Eastwood.

One major global problem with MT 💡

Numerous machine translation engines… Statistical, rule-based, hybrid, neural MT types… Constant developments, new players, promising technologies…
Machine translation helps save time and money and is here to stay forever.
But when we talk about MT quality, we assume the source is error free, consistent in terminology, is written in a clear understandable way and does not require any Google search to understand the context and real meaning.
But is it always the case ❓

Trying to reduce costs doesn’t start with the translation stage.
Sometimes source content is created by unskilled technical writers or even non-native speakers, text in the original language is not edited by an industry expert, new document is made by combining parts of previous documents written by different authors, etc.

Are there any technologies that help MT engines fix source issues, understand meaning and translate the meaning and not words?

How long does it take to become a good translator❓

✏️ When posting jobs many LSPs require 2 or 3 years of experience.

✏️ The translation industry’s most stringent standard, ISO 17100, suggests:
–      A person has a degree in translation but no real-life translation experience
–      A person has a degree in any field + 2 years of translation experience
–      A person has 5 years of professional translation experience

👉 In my experience as a translator, editor and recruiter, a person achieves excellent translation skills after 5 to 7 years of translating and receiving regular feedback from a seasoned editor. At Lexicon Center, we only use translators and editors with at least 10 years of experience in a particular field.

I had a great time hosting the virtual side of #Locworld Silicon Valley After-Party 🎉

A huge thank you to Michelle Kuniko, Marina Gracen-Farrell, Óscar Curros, Stefan Huyghe, Paul Barlow, Joseph Holtman, Stephanie Harris, Marie Flacassier, Iti Sahai, Gary Lefman and around 70 other localization professionals who met offline 👏

Sorry not all of them are in the picture.

Does MTPE make a translator's writing style worse❓

At yesterday’s #LocTalk panel discussion by Josef Kubovský, Derick Fajardo and Igor Afanasyev I brought up a question.

❗️When a person actively does post-editing, they gradually (and unwillingly) adopt machine translation approach in their human non-MT translation jobs.

They may start copying source sentence grammar structures and word combinations that may not be characteristic of the target language. It results in unidiomatic translations that don’t sound natural.

This is what I have seen a lot.

✅For translators: be very careful to accept MTPE jobs if you have a good writing style and specialize in marketing, transcreation and other demanding fields.

✅ For LSPs: think twice before giving a particular translation job to a translator who offers MTPE.

Finding a good translator... 👉🏻

WHERE: A popular translation platform.

WHO: A translator with 15 years of experience, 90 positive reviews and 5 million translated words.

WHAT: A 500-word general translation. Deadline – 2 days.

OUTCOME: Overall, very good style and understanding. But when our editor dug deeper, he found 3 minor meaning mistakes, 2 stylistic issues, 1 unnecessary comma and 1 ambiguity. The editor pointed out the issues to the translator, and she quickly made necessary changes.


Is it just bad luck?

Do clients avoid leaving negative reviews unless there are major issues with the translation/translator?

Or do they just don’t dig deep enough?

💎 TAKEAWAY: Always use an editor, because even a very good translator can make mistakes