Hello and welcome. We have drafted the following guidelines to help you deliver an excellent service with Unbabel. By aligning expectations and promoting uniform criteria, we hope to offer consistent, high quality writing, with less confusion, inconsistency and customer complaints.
Please take the time to read through these guidelines and refer back to them in your daily activities. Let’s go!
2. Tools to help you along the way
At Unbabel, we push technology to its limits and accept that even the most advanced machine translation needs a guiding human hand. That’s where our community of tens of thousands of editors comes in. To help them do the best job possible, in as little time as possible, we’ve developed a number of tools.
Smartcheck highlights words or groups of words in red (meaning something is not correct) and green (which suggests an improvement in style). Just click on the underlined word, and you’ll see a number of suggestions from Smartcheck. It’s not necessary to accept the suggestion; you can accept it (by clicking on it) or just ignore it.
Glossaries come from individual Unbabel customers who have brand terms and other linguistic specifications pre-loaded into their account. These will appear in grey in both source and target text – please ensure they have not been modified by an editor. To check, just click on the source text, and the original translation will appear.
Please only change glossary terms as a last resort (for gender and number issues, for instance).
2.3 Contextual information
Sometimes it is difficult to understand the wider context of the snippet that you might be working on. For these situations we have developed a tool you can access by clicking on the “Context” button above the text. This can be very useful in avoiding errors related to reference, gender and number.
2.4 Translation memories
Translation memories are made up of stored segments (sentences, paragraphs or elements like headlines and titles), which have previously been translated and accepted for customer usage. They are useful for ensuring consistency across jobs, and mean that customers get translated work much faster over time.
Should any of the tools appear to have issues or errors, please use the Report tool, easily found on the right hand of the editing screen, and we shall look into it as soon as possible.
3. Customer instructions
Customer instructions are vital to delivering translations that match their expectations. Calibrated on a per-customer basis, they are an essential resource for ensuring high quality work across the many different customers you may do editing for.
Register is the level of formality with which Unbabel’s customers address their audiences. Maintaining the voice across multiple languages is a huge concern for them, and we must work hard to respect requested registers, and the tone of the source texts.
For example, a text with a formal register should use polite forms in addressing the audience, appropriate vocabulary, and appropriate verbal tenses. Equally, you should not add colloquial expressions in translation.
You should also avoid formalizing the text if it is written colloquially — maintain a simple vocabulary.
Different languages have different ways to structure register, particularly in Japanese, and some customers will even specify their own register definitions. For further detailed information, check section 5.3.1.
Customer Instructions: This is a blog post introducing our new make-up range, which we want to appeal to young people, so the register should be relatively informal.
Source text (FR): Et voilà c’est ça! Tu remets le mascara dans sa petite poche vachement pratique et t’es prête pour la soirée!
✘ Translation with inappropriate register: And thus that is it. Please set the mascara back into its small, extremely functional chamber and you are prepared for the evening’s festivities.
✓ Translation with appropriate register: And there you go! You just pop the mascara back in its super practical little pocket and you’re ready for your night out!
Customer Instructions:This is an article from a French broadsheet newspaper, to be translated and published in a UK broadsheet newspaper. The register should be formal.
Source text (FR): Avec la modification génétique de primates, celle de l’homme semblait plus que jamais à portée de main.
✘ Translation with inappropriate register: With the GM of monkeys, GMing people seemed closer than ever.
✓ Translation with appropriate register: With the genetic modification of primates, the genetic modification of man appeared to be, more than ever before, at our fingertips.
3.2 Maintaining capitalization and format
Another frequent requirement from our customers is to explicitly keep capitalization and formatting exactly the same as in the source text, even in cases where you would use small capitals. In other words, this should be done exclusively when requested by a client; otherwise you should follow the English guidelines on capitalization and formats.
Customer Instructions: Keep the first letter of every word capitalized in the translation of the titles of each section.
Source text (ES):
Los 5 Hoteles que Tienes que Ver Antes de Morir
✘ Translation with bad formatting:
The 5 hotels you have to see before you die
✓ Translation with good formatting:
The 5 Hotels You Have to See Before You Die
3.3 Untranslated words
Many of our customers have brands, products and business expressions that they do not want to be translated. Pay special attention to these, checking both the Glossary and Customer Instructions to make sure you have not translated them.
Source text (FR): Rouge à lèvres Astrological Softness
Target text (EN): Astrological Softness Lipstick
4. Layout and Formatting
Unbabel developed a state-of-the-art common format (Unbabel Common Format, UCF), in order to deal with the variety of types of files we receive as input and transform them into a simple text to be edited. This is a very challenging task and, as such, there are still some labels and diacritics, which are not fully captured by our UCF. Whenever cases like this occur, we advise you to report the task and let us know about the issue. It can tackle, for instance, extra white spaces, odd sentence breaks, etc.
On customer support communication, sometimes it is necessary to include confidential information (such as addresses or passwords), shareable solely between client and agent.
The terms anonymized can be URLs, credit cards, phone and social security numbers, dates, passwords and emails.
Whenever you find a term highlighted in grey on tasks or reviews, don't edit or delete: once you submit a job, our machines will replace it with the private info and deliver the task to the receiver.
ENGLISH SPECIFIC INSTRUCTIONS
In this section, you will find some useful instructions about your language specifications.
The particular issue of English as the target language
Translation into English has a very particular issue found in few other languages. English has become the global language of communication. It is widely taught and widely spoken and, as with several other languages, such as French, there are now many different varieties, not all of which are easily understood by the native speakers of UK or US English.
The translation issue, however, is not the number of varieties, but the perception that ‘everyone speaks English’ and, therefore, the misconception that in the case of English it is not necessary to be a native speaker to translate into English.
The golden rule of translation, that the translator must translate into his/her native language, is frequently disregarded when the target language is English, partly because many non-native speakers have very good, or even excellent, English, and partly because demand for translators into English is greater than the supply of capable native English-speaking translators.
As a result, the majority of errors found relate to mistakes made by non-native English speakers. These are numerous, ranging from a literal translation with little regard for natural English to smaller errors in preposition use, sequence of tenses, and English word order.
However small the error may appear to be, any error means that the translation is below the professional standard that Unbabel requires.
5. Most frequent errors
The most frequent errors we find in translating into English can be divided into these main categories: accuracy, grammatical errors, word order, style, terminology, language variety, and named entities. Errors in any of these categories means that the target text is not fit for purpose. You can find here the basic information you need in order to avoid the most frequent errors made during editing Unbabel’s tasks.
Accuracy errors alter the actual meaning of the source text.
5.1.1 Careless omission of key words
Source text (IT): La disponibilità non è aggiornata in tempo reale.
✘ Incorrect translation: Availability is updated in real time.
✓ Correct translation: Availability is not updated in real time.
The omission of the key word ‘non’ has given the target text the contrary meaning to the source text.
5.1.2 Mistranslation of individual words and phrases
Source text (IT): l’etichetta ti verrà consegnata al momento del ritiro del pacco.
✘ Incorrect translation: The label will be delivered to you at the time of withdrawal of the package.
✓ Correct translation: The label will be delivered to you when the package is collected.
Source text (FR): Merci de faire un retour positif.
✘ Incorrect translation: Thank you to make a positive return.
✓Correct translation: Please leave positive feedback.
5.1.3 Lexical selection
Source text (IT): Alcuni prodotti prevedono penalità per la cancellazione.
✘ Incorrect translation: Some products provide penalties for cancellation.
✓ Correct translation: Some products incur penalties for cancellation.
This is an error in collocation – words that go together.
Source text (IT): Sto arrivando subito!
✘ Incorrect translation: I’m coming suddenly!
✓ Correct translation: I’m coming straightaway!
This is a ‘false friend’ error - a word in the source language that looks similar to a word in the target language, but has a different meaning.
5.1.4 Errors in idiom
188.8.131.52 Over-literal translation of formulas of courtesy
Many languages have particular formulas of courtesy when asking people to do something. Such formulas are inappropriate if translated literally into English.
Source text (FR): Nous vous invitons a contacter directement l’hôtel
✘ Incorrect translation: We invite you to contact the hotel directly.
✓Correct translation: Could you please contact the hotel directly/ We suggest you contact the hotel directly.
Source text (FR): Afin de pouvoir recevoir ce code, merci de me confirmer votre date de naissance.
✘ Incorrect translation: In order to receive this code, thank you to send me your date of birth.
✓ Correct translation: In order to receive this code, could you please send me your date of birth.
184.108.40.206 Over-literal translation of idiomatic expressions
Source text (IT): Il bilancio dell’UE, pur essendo quattro soldi,….
✘ Incorrect translation: The EU budget, while being four coins……..
✓ Correct translation: The EU budget, while being a derisory sum, …..
or: The EU budget, while worth only a few pennies….
5.2 Common grammatical errors
5.2.1 Uncountable nouns
A number of nouns in English are ‘uncountable’, that is, they do not have a plural form. This can cause problems in translation.
Source text (FR): Vous pourrez rendre vos équipements dans l’un de nos magasin.
✘ Incorrect translation: You can return your equipments to one of our shops.
✓ Correct translation: You can return your equipment to one of our shops.
Too literal a translation often leads to prepositions that are required in English, but not in the source language, being omitted.
Source text (IT): Puoi verificarlo rileggendo la politica di cancellazione.
✘ Incorrect translation: You can verify it rereading the cancellation policy.
✓ Correct translation: You can check it by reading the cancellation policy once more.
Source text (FR): Vous pouvez annuler votre réservation jusqu’19h le jour de votre arrivée.
✘ Incorrect translation: You can cancel your reservation up until 7pm the day of your arrival.
✓ Correct translation: You can cancel your reservation up until 7pm on the day of your arrival.
The use of prepositions in any language depends on the entire phrase, above all on the verb they follow. Preposition translation therefore often leads to errors.
Source text (IT): Non è arrivato a casa sua sulla spiaggia.
✘ Incorrect translation: He did not arrive to his beach house.
✓ Correct translation: He did not arrive at his beach house.
The article rules in English are complex and a literal translation will often cause errors in their usage. For example, an undefined singular noun needs to have the indefinite article whereas undefined plural nouns have no article.
Source text (FR): Jean est professeur et ses soeurs sont des médecins.
✘ Incorrect translation: Jean is teacher and his sisters are the doctors.
✓ Correct translation: Jean is a teacher and his sisters are doctors.
5.2.4 Verb tenses
Literal translation can also cause errors in English sequence of tenses.
Source text (IT): Quando riceverai l’mail di conferma, la prenotazione sarà effettiva.
✘ Incorrect translation: When you will receive the confirmation email, the reservation will be effective.
✓ Correct translation: When you receive the confirmation email, the reservation will come into effect.
Note that English makes far less use of a future tense form than other languages, frequently using either the present simple tense for an event in the future - When the postman comes, could you let me know.- or the present continuous - We are spending a month in Italy next year.
220.127.116.11 Tense with ‘since’
A very common error is incorrect tense with ‘since’, which requires the present perfect (or past perfect) tense (or present perfect continuous tense) rather than a simple tense.
Source text (IT): Il tour leader sará Luca Giovanni che guida gruppi in questa regione dal 2010.
✘ Incorrect translation: The tour leader will be Luca Giovanni who leads groups in this region since 2010.
✓ Correct translation: the tour leader will be Luca Giovanni who has led (or, has been leading) groups in this region since 2010.
The general rule, as stated in the general introduction, is to maintain the capitalisation of the source text, except in cases, such as German, where every noun is capitalised.
However, English often uses initial capitals where other languages uses lower case.
Source text (IT): il ministero di turismo
✓ Correct translation: the Ministry of Tourism
It is important to remember that in English, unlike many languages all place names and related adjectives, all country names and related adjectives, and all languages and related adjectives are capitalised.
Examples: France, the Italian people, the German language, Chinese food, Tuscan wine, Slovakian dances, Greek cooking.
Rules for English capitalisation are very clearly laid out in the Chicago Manual of Style (see online resources) using the search term ‘capitalization’.
5.3 Word order
English word order is relatively rigid, unlike many other languages, and this is possibly the largest problem encountered in translations, particularly as computerised translation tools are often unable to get it correct.
Source text (IT): Ai prodotti riservati a te verrà applicato il prezzo più basso.
✘ Incorrect translation: To the products reserved for you will be applied the lower price.
✓ Correct translation: The lower price will apply to your reserved items.
Note that in English the subject of the verb, that is, the noun or pronoun, (the…. price) precedes the verb (will apply) and adjectives precede nouns (the lower price).
English, like Italian, uses verbal past participles as adjectives (your reserved items).
The main difference between the languages is the greater flexibility of Italian, compared to English where the subject and verb must be in that order and come before the direct or indirect object.
Note also the use of the possessive form in English word order:
Source text (FR): ….pendant les heures d’ouverture de la réception…
✘ Incorrect translation: …during the opening hours of the reception…
✓ Correct translation: …during reception’s opening hours…
At Unbabel we differentiate between two registers: formal and informal.The register, that is to say the level of formality used in the text, shows how our clients address their customers and contributes to the voice of the brand itself. Register varies depending on the company, the brand, the service they offer, the customers, and the target language.
See Section 3.1 for excellent examples of appropriate informal and formal language.
In addition, it is often the case that a translated passage, while perfectly correct, sounds too formal and stilted for its context.
Here is an example of some tourism literature.
Source text (IT): Molti sono gli interessi e i percorsi che possono portare il visitatore fra i nostri sedici Comuni.
Correct but over-formal target text: The interests and paths which may bring the visitor to our sixteen Communes are myriad.
✓ Correct and more informal target text: There are so many interesting places and itineraries to attract the visitor to our sixteen little towns.
18.104.22.168 Contracted verb forms – informal
The tone of the source text should be maintained in the translation. For example, much modern marketing material, and in particular the ‘chat’ facility in customer service advice, uses informal, spoken forms – above all contracted verb forms such as can’t, won’t, shouldn’t, she’s, they’re.
However, as a general rule in translation, it is better to err on the side of formality. It is of course, important to be consistent and not mix informal and formal writing in the same text, unless this has been done deliberately in the source text.
In informative writing, informal spoken usage, such as contracted verb forms, should not be used:
Source text (IT): Lo studio si basa su ricerche sul campo.
✘ Incorrect translation: The study’s based upon field research.
✓ Correct translation: The study is based upon field research.
There is a wide division between the formal letter greeting of Dear Sir/Madam or, if addressing a particular person, Dear Dr. Morrison, and the highly informal Hi Boris.
Again, it is important to reflect the level of formality of the source text. On certain occasions it is perfectly permissible in modern English to retain the informal Ciao! of Italian or the Salut! of French, which would be totally unacceptable in a formal letter, where the correct ending is Yours sincerely or Yours faithfully.
Note that these are the correct English formulas for rendering the various formal ending formulas in other languages, such as (IT) Le invio il mio più distinto saluti (literally: I send you my most distinguished greetings).
Informal endings in English are best rendered by phrases such as Best wishes, Warm regards, or even the ubiquitous Have a nice day (which is very informal indeed).
Errors in the terminology used due to non-compliance with the client or company style guide. That is, the translation does not follow the instructions left by the client.
For example, if given instructions that Comune di Venezia (IT) should be rendered as ‘Municipality of Venice’ and the translation uses ✘ ‘Commune of Venice’.
5.6 Wrong Language Variety
English is now a global language with several varieties, but the two main varieties used are still UK English and US English. The differences relate mainly to spelling, with US English using a more simplified spelling system. Even so, this affects very few actual words.
A client will often specify which variety of English should be used. If the variety is not specified, the main point is to be consistent, so that you do, for example, have ‘archaeology’, ‘rumour’, ‘centre’ (UK) and ‘archeology’, ‘rumor’, ‘center’ (US) in the same passage.
The default machine translation variety is US English, although the machine translation is not always consistent in this respect.
5.7 Named entities
In this section, we are going to talk about errors that have to do with localisation and names.
In most cases involving money and currency the client will specify what to do. If there is no instruction from the client, you need to decide on the best course to follow.
If the reference is idiomatic and colloquial, then your policy will be to find an English equivalent.
Source text (IT): Non pensi che valga qualche centesimo?
Target text (EN): Don’t you think it’s worth a few pennies?
If, on the contrary, the text is referring to a specific, fixed amount relating to the cost of goods or services, then it should be left in the named currency.
Source text (IT): Su cento euro di R&D spesi in Europa, ne provengono dal bilancio UE tre.
Target text (EN): For every hundred euros spent on R&D in Europe, three euros come from the EU.
English differs from other languages by not, in general, using the 24 hour clock. This means that references to time need to be changed to an am/pm format
Source text (FR): Vous pouvez modifier votre reservation jusqu’ 19h le jour de votre arrivée.
✘ Incorrect translation: You can change your reservation up until 19h on the day of your arrival.
✓ Correct translation: You can change your reservation up until 7 pm on the day of your arrival.
22.214.171.124 Numbers as digits
If numbers are written as digits in the source text, this should be maintained in the translation. If they are written as words, they should be translated.
The conventions for punctuating large numbers differ between languages. English uses a period to indicate the decimal place.
Source text (IT): 295,045
Target text (EN): 295.045
English uses a comma to separate groups of thousands.
Source text (IT): 4.294.967.295,045
Target text (EN): 4,294,967,295.045
126.96.36.199 Numbers as words
If numbers are written as words in the source text, they should be translated. In English all compound numbers between twenty-one and ninety-nine are hyphenated.
Seventy-nine students gained a B grade.
Three hundred and twenty-one gained C.
When it comes to acronyms, it can often be difficult to decide how to approach translation. There are a few quick questions to ask yourself about the acronym:
- Would the acronym have an equivalent in every language?
This would be the case when translating acronyms such as (IT) PIL (Prodotto Interno Lordo)= (ENG) GDP (Gross Domestic Product) or (IT) IVA (imposta sul valore aggiunto) = (EN) VAT (value added tax)
- Is the acronym the name of a globally known institution?
For example (FR) OMS (Organisation Mondiale del la Santé) = (EN) WHO (World Health Organisation).
In all of these cases the English equivalent should be given.
- Are the answers to these two questions no? In this case, your best option is to maintain the acronym but translate what it stands for.
For example: (FR): FFBT ((Fédération Française de Ball-Trap or French Federation of Clay Pigeon Shooting).
5.7.5 Place names
Where names of countries and large cities have an obvious English equivalent, this should be used, but in general less globally known place names should be left in their original form.
For example: (IT) Val d’Aosta rather than The Valley of the Aosta.
With machine translation, serious errors can easily arise through ‘translation’ of place names.
Example: Source text (IT): Il terrremoto dell’Aquila del 2009
✘ Incorrect translation: The earthquake of the Eagle of 2009
✓ Correct translation: The L’Aquila earthquake of 2009
5.7.6 Proper names
With a few exceptions (biblical characters and certain historical characters) proper names are not translated into an English form.
Source text (FR): Bonjour Patrice, Je serais ravie de pouvoir vous aider.
✘ Incorrect translation: Hello Patrick, I would be delighted to help you.
✓ Correct translation: Hello Patrice, I would be delighted to help you.
6. Slight edits
Here at Unbabel we would like to encourage our editors to follow these instructions and not leave an original machine translated text unedited or only slightly edited as this may result in the text sounding awkward and less natural to a native speaker of the target language. We advise you to edit the text and read it over twice. Does it sound natural? Does it flow smoothly? Does it sound human? Do your best to fine tune your translation to a native-speaking public.
7. Errors in the source text
What do you do if there is an error in the source text? Use your initiative.
- If you are absolutely certain that there is an error and you know what the text was supposed to say, then translate it as if it were written correctly originally. Remember that the machine translator is unable to identify mis-spelt words or other errors in the source text
Source text (IT): Il servizio permette inolte di visualizzare se il prodotto è disponibile.
✘ Incorrect machine translation: The service permits inolte to view if the product is available.
✓ Correct translation: In addition, the service allows you to see if the product is available.
- If you find an error which makes it impossible for you to work out the meaning of the source text, report the task (Button at the top right marked Report), explaining your reasons.
8. Useful online resources
In addition to the online resources listed below, it is helpful for a translator into English and editor of English to be able to consult the following books, none of which are very expensive and which cover a wider spectrum and give more detail than online dictionaries:
New Oxford Spelling Dictionary; Oxford University Press. ISBN 978-0-19-860881-3
New Oxford Dictionary for Writers and Editors; Oxford University Press ISBN 978-0-19-861040-3
Concise Oxford English Dictionary; 12th edition, Oxford University Press ISBN 978-0-19-960108-0
A good up-to-date hard copy bilingual dictionary will also provide fuller examples, with all the possible different meanings of a term in different contexts, than an online one, as well as having tables of verb conjugations, acronyms, and other useful information.
On capitalization, punctuation, dates, numerals and just about everything else
use the appropriate search terms
English verb conjugator:
English grammar guide