Some Fundamental Questions in Translation

By Ibrahima Diallo

Since we all use language and there are many different languages in the world, most people understand, at a superficial level, what translation is: You have two parties that do not share a common language, so you need some kind of intermediary to make communication possible.

However, most people outside of the linguistic field do not give much thought to the core issues of translation.

In the following paragraphs, I am going to present some of those issues, in the form of questions and answers. I will try to present each topic in a simple way, without going into the intricacies of the debate. Each of the topics discussed below is only the tip of the iceberg in a vast ocean of translation theory, where multiple competing – and often conflicting – views exist. The presentation below is a summary of my own opinions, based on more than 30 years of practice in the fields of translation, translator training, translation management, the interface between translation and language technology, as well as the books I read on each topic throughout the years.

Q: Is everything translatable? In other words, if we take a statement in any language, is it always possible to say the same thing in any other language?

A: In principle yes, but with some caveats. Linguists consider all languages to be equivalent as communication tools. They can all describe reality and convey information to the same degree. They are all rule-based and capable of integrating new concepts and ideas. However, we must take into account the fact that language is an aspect of culture. Each language evolved in a specific natural and cultural environment, making it best equipped to describe things that exist in that environment. When you translate, you are not just transferring a message from a source language to a target language. You are also transferring meaning from one culture to another. If you have sufficient knowledge of two languages and their cultures, you should be able to somehow transfer meaning, but this can be harder the more “foreign” the two cultures happen to be from each other. In some cases, the statement you are trying to translate may be about something that is completely alien to the target culture. In that case you may have to resort to some explanations that are not necessarily stated in the original language, for example by adding translator’s notes, which is prohibited in some institutional contexts, such as the United Nations. Translators generally try to avoid adding their own comments to document as much as possible. But there are some cases where translation cannot be done without some explanation. Imagine (as an extreme example) trying to translate computer repair instructions to the language of a tribe that lives in – say – the Amazon forest. As stated previously, such a language would be just as capable of describing its environment as any other language. But in this case, its users may be unfamiliar with the concept of a computer. So, unless the translator has the luxury to first explain to his/her audience what a computer is and what it does, he/she might as well throw in the towel and declare the goal unattainable.

Fortunately, most translation situations are not that extreme, but cultural/environmental differences are a constant source of struggle for translators. They are always trying to find a way to transfer meaning across cultures in a manner that is accessible to speakers of the target language and sounds as natural as possible to them, without introducing explanations that go beyond the intent of the speaker or author. I once read about a translator who was translating from English to Arabic a text that mentioned a dog race in passing. Faced with the fact that dog races are alien to the Arab culture, the translator opted instead to substitute dog racing with camel racing. I have also seen movie subtitles in Arabic where curse words and sexual references were significantly “softened,” or where references to alcohol were replaced with soft drinks. Many translators would object to such alterations and consider the translations inaccurate or even objectionable. But we can also agree that making the translated discourse more acceptable, less shocking, and therefore more accessible/entertaining for the target audience is an understandable goal. It is all a question of striking the right balance for each specific case.

In cases where the translator sees a need to introduce these types of cultural adaptations, with the aim of helping the author’s message achieve a more positive impact, it is important to reach out and provide adequate explanation to the owners of the material being translated; to seek their approval, making sure intellectual property rights are respected in the process (for example when subtitling movies).

Another “untranslatability” problem comes from differences in the semantic scope of words from one language to another, which can cause ambiguity, leaving the translator with the responsibility of arbitrating meaning. Arbitration comes with the risk of picking an option that was not intended by the author. For example, a simple verb in English like “to clear” can be used with many different meanings, and it is highly unlikely that another language will have a word that covers all those meanings. This can sometimes result in the word “clear” being used in a sentence where the context does not tell us whether it means “clear” as in delete, remove, get rid of … or “clear” as in validate, authorize, etc. …

Puns are another type of language use that is notoriously hard to translate, because they are based on one word meaning different things or two words sounding alike. There is a silly joke in English which goes like this: Why is 6 afraid of 7? Because 7, 8, 9. This joke would be impossible to translate into, say, French or Spanish because, even though those two languages are culturally quite close to English, the pun based on “8” sounding like “ate” cannot be reproduced.

In the majority of cases, it is difficult, or even impossible to come up with a perfect translation of a given statement, especially between culturally distant languages. Translation is therefore a constant negotiation and – yes even compromise – with meaning. A translator, assuming she/he has a perfect understanding of the original statement – which is not always guaranteed due to factors such as the natural ambiguity of language – must pick among possible renditions in the target language, one that she/he thinks is the best, because oftentimes there is no guarantee of a perfect equivalent, and even if there was one, time constraints are usually such that the translator cannot afford to spend too much time searching for it.

Q: How do you assess the quality of a translation?

A: There is no perfect method to assess the quality of a translation. There is a broad array of quality assessment tools and techniques used by different groups of translators. They are usually based on an attempted categorization and weighting of translation errors (meaning, use of appropriate terminology, grammar, typos, appropriate use of idiomatic expressions, style, etc.) but in the final analysis, quality is a question of judgement. In most institutional contexts, written translations, especially when they are done by junior translators, or translators who are not familiar with the institution’s work, are subject to revision by a more seasoned translator. In other cases, various peer-to-peer review methods are used to ensure translations are scrutinized by another pair of eyes, so as to catch possible errors. It is proven that when we re-read our own writing, confirmation bias makes it very difficult to spot errors. This is why scrutiny by a second pair of eyes is always useful when doable. For this reason, beginner translators working directly with clients, should, as much as possible, find ways to associate with other translators in order ensure proper review of their work, and to build that review process into their cost structure and delivery deadlines.

One of the most important elements of quality is accuracy (also referred to as fidelity), meaning the degree to which the translator faithfully transmitted the message from the original statement or text. This requires that a reviewer – whether the review is done by the original translator after completion of a draft (self-revision) or by a different person – make a sentence-by sentence comparison between the source and target texts to make sure they say the same thing.

This ties into the question of understanding what the speaker/author is trying to say, i.e. the author’s “intent.” It also ties into word-choice by the translator, especially the use of appropriate terminology.

The notion of “intent” in translation is a tricky one. A translator is always searching for a balance. On the one hand, he/she wants to avoid taking the surface level meaning and running away with it (literal translation) which can miss the core message or produce strange-sounding sentences in the target language that baffle the intended audience. On the other hand, a deep search for hidden intent can lead to excessive interpretation or “re-telling the story,” where the translator gets tricked by her/his own worldview. Many ancient translations, such Arabic translations of Greek philosophy and translations of biblical texts from their original languages (Ancient Greek, Hebrew, Aramaic) suffer from this flaw.

Terminology is also a crucial component of a good translation. Each field of knowledge, especially highly specialized knowledge, has its own technical terms that convey precise meaning. Terminology can also be localized, so that different groups of speakers using the same language in the same field can have their own local terms. Using the wrong term, for example in the medical field, can lead to negative consequences. This is why translation requires the translator to carefully check and select appropriate terms; do as much as research as deadlines allow; and keep an open mind at all times, questioning whether her/his interpretation is generally consistent with the rest of the statement and the context in which it was made. If the content of the source document is highly specialized, the translator should carefully consider whether she/he understands the material enough to provide a good translation. In case of doubt, the safest option – professionally and ethically – is to decline the assignment, unless she/he can find a more qualified translator to review the translation.

Terminology choice must also take into account the target audience of the translation (if known), since different regional or professional groups of users of the same language can have different terminological preferences. Terminology verification, with its associated research, is an aspect of translation that usually takes up a large proportion of the translator’s working time.

Another component of quality in translation is consistency. Internal consistency requires that each sentence in the source text be understood, not as an isolated unit, but in the context of what comes before and after it. The aim is to make sure different parts of the source document come together in the target text to form a coherent whole, while taking care not to excessively smooth over actual incoherences or ambiguity in the source text, which can be intentional sometimes, such as in the case of documents born out of arduous negotiation on contentious matters.

External consistency means that, when translating related documents, such as successive annual reports of the same entity, or different but related documents on the same topic, the translator must do so in a way that preserves their unity and continuity as much as possible. This includes spelling proper names the same way, consistently using the same names for the same entities, phenomena, events, etc., as well as using the same translation whenever citations or chunks of text are repeated in different places. This is one of the areas where computer-assisted translation (CAT) tools are most useful. They can help find similar chunks of text in different places and compare their translations, or “recycle” previously translated sentences from the same or related documents.

Quality also includes the element of “fluidity,” which means producing a translation that sounds natural in the target language. Here the translator strives to translate meaning, not sentence structure. He/she must pay attention to word order, verb tenses, prepositions and other structural elements of the source sentence, but without getting bogged down in such details or dragging them carelessly onto the target language. The indiscriminate transfer of structural elements can produce a translation that sounds unnatural to the users of the target language, thus affecting their ability to understand the translation.

Once I was reviewing a document translated into Arabic by a junior translator. The author was describing situations of extreme poverty around the world and the need to increase humanitarian aid. The source text contained the sentence “It is time for those who have the most to step up in order to help those who have the least.”  The translator somehow thought he needed to use the equivalents of the words “have,” “most” and “least” in the target language, and ended up with a literal translation that made no sense. I reached out to him and was able to convince him that the author simply meant that the richest of the rich should step up and help the poorest of the poor. By taking the meaning and abstracting it away from the actual words chosen by the author, we were able to come up with a much smoother expression in the target language that would make sense to the reader, without altering the meaning.

The quality criteria described above are especially applicable to written translation. But some of them can apply also to interpretation (simultaneous or consecutive). Interpreters usually do not have the luxury of doing quality assessment, given the immediate delivery of their work. However, interpretation quality can be assessed after the fact, especially in the context of training, but not before the product reaches the target audience. Institutional interpreters use various methods to mitigate quality risks, including efforts to broaden their knowledge about the specific topics on their agenda, and receiving advance copies of speeches to be delivered in meetings whenever possible.

Q: What ethical issues are there in translation?

A: As far as I know, there is no universal code of ethics for translators. However, translators generally agree that they have ethical obligations resulting from the nature of their work. A number of those obligations are uncontroversial. Firstly, translators must strive to provide their employers and clients with the highest quality product they can achieve and abstain from taking on a project they are not qualified to do. They also have the ethical obligation of remaining neutral vis-à-vis the contents of the statement or document being translated, so as not to allow their personal views and beliefs to affect their translation in any way. Confidentiality is also a core responsibility of all translators. They must not reveal any non-public information that comes to their knowledge through their work.

Beyond these rather obvious rules, there are others that fall into a kind of grey area, depending on circumstances. The first one is whether a translator should report criminal or suspicious activity to the relevant authorities if it comes to her/his attention during a translation. In this case, the translator would have to weigh the potential harm from said activity against his/her confidentiality obligations and choose a course of action that, in his/her view, would cause the least harm to society in general.

The other issue is that of achieving a balance, in specific cases, between the obligation to produce an accurate, neutral translation and any potential harm that might result from doing so.

I was once called on to provide interpretation for a delegation from an Islamic NGO that was visiting my country to do humanitarian work. The NGO was providing cataract surgery for underprivileged inhabitants of the capital city. Journalists from the national radio station came to interview the NGO delegation in charge of the operation. They asked them questions about the details of the operation. Among the questions was whether their services covered only Muslims or would include qualified patients of all religions (since the population of the capital was highly mixed in terms of religious affiliation).

After I translated the question, the head of delegation answered using a deliberately ambiguous word in Arabic (mathaahib), which could be understood both as signifying a religion (faith, creed, religious persuasion), or denominations of a given religion. I could not find a word in the target language with such nuances, so I simply translated “We serve patients of all religions.” After the journalists left, the head of delegation called me back to admonish me for mistranslating his answer. He said that by using the word “mathaahib,” he did not mean all religions, but only denominations of the Islamic faith. I told him I felt a moral obligation not to convey on national radio that the NGO was here to serve only Muslims, since such a statement could damage inter-faith relations in the country. The head of delegation then decided to remove me from the translation team.

All this is to say that in certain situations, translators and interpreters can find themselves in the midst of ethical conundrums where conflicting ethical obligations need to be taken into account. In those cases, they must exercise their own judgement to select a course of action that they believe will lead to lesser harm.

Q: What is the future of the translation profession in an era of increasingly powerful artificial intelligence?

A: My personal opinion on this matter, as someone who has worked on the user requirements side of language technology for many years, is as follows:

The evolution of language technologies is already reshaping the translation landscape and will certainly continue to do so in the coming years. Traditional computer-assisted translation (CAT) tools, which include several components (text editors, terminology databases, etc.) as well as statistical or neural machine translation “engines,” have so far served as a translator’s workhorse.

However, in addition to being a useful tool, the emerging technology of large language models (LLMs), which are “trained” on massive amounts of linguistic data, is now transforming machines into potential competitors for the translator.

Without getting into the nitty-gritty of the debate on artificial intelligence in translation, I believe the new technologies, in particular LLM-based translation tools, will be playing a prominent role in the coming years, and will force translators to adapt to a new reality, whether we like it or not. There are many cases where the target audience (or the client) only needs a broad understanding of the content of a source text. In such cases, automated systems will be the default solution, especially when dealing with large quantities of text. Automated systems are faster and much more cost-effective than human translation.

However, machine translation is still far from producing quality translations that can rival a good human translator because they work by ingesting large quantities of linguistic data and lack the intuitive understanding and flexibility of a human translator. Compared to traditional “neural” machine translation engines that are currently used in existing CAT tools, LLM-based translation engines have the ability to “digest” vastly larger quantities of linguistic material, which seems to provide them with a somewhat better understanding of context. This could give future LLM-based systems an advantage over traditional tools, especially once they are optimized for the specific purpose of translating, not just general use.

However, LLMs are known to occasionally “hallucinate,” producing incorrect or nonsensical sentences. Due to multiple factors, including the enormous amounts of material needed to train LLMs, it seems unlikely that this problem will be solved anytime soon, especially in language pairs where lesser amounts of data are available. Furthermore, research seems to indicate that the linguistic quality of an LLM’s output gets worse and worse when the system is fed with machine-produced content. The further the loop of machine-produced data feeding other machines continues, the more degraded the output becomes, until it devolves into pure gibberish in a phenomenon called “model collapse.”

Generally speaking, for any type of document requiring a high-quality translation (i.e. documents where unchecked translation mistakes can have grave consequences), human translators will be needed for the foreseeable future.

It is, therefore, safe to assume that professional translators should be able to enjoy job security for quite some time, especially if they do their work with due diligence and know how to exploit the strengths of technology tools, while using their professional judgment to make better decisions in terms of resolving ambiguity or minding cultural sensitivity. This is even more applicable in the case of translators who work with less common languages where enough quality data is not yet available to train automated systems.

Q: What advice would you give to a junior translator who wants to enhance his/her skills?

A: The first step on the path to becoming a good translator is – obviously – knowing languages. Prior to becoming a professional translator, most newcomers go through a translation training program. Others simply transition to translation from related (or sometimes unrelated) fields, using their knowledge of languages. I would hope that any would-be translator is, first and foremost, someone who has a passion for languages, who has already acquired above-average knowledge of at least two languages, and who is willing to embark on a journey of life-long learning.

Every language is a vast domain of knowledge, with an ever-expanding vocabulary; grammatical rules with myriad exceptions and special cases; idiomatic expressions that evolve with culture; and meanings that can shift from one generation of speakers to the next. No one can claim to know everything about any language, even the one we were born into, let alone ones we learned later in life. Continuous learning is therefore crucial in order to better oneself at understanding and being able to transfer the nuances of meaning.

The second pillar is practicing and a willingness to seek and receive feedback from others. A junior translator must have her/his work reviewed by more seasoned translators, and be confident, yet humble enough, to subject her/his work to scrutiny by others who bring a different perspective and who may disagree with his/her choices. Having one’s translations critically examined by other translators is, in my opinion, the best learning tool a translator can have. Due to confirmation bias, it is much more difficult, when reading our own translations, to spot errors or elements of context that escaped our attention the first time.

If, for any reason, a translation cannot be revised by someone else, the translator must then do “self-revision” as a separate step during which he/she carefully analyses his/her her own translation, comparing the source and target texts sentence-by-sentence with a critical eye. Every sentence must be read and re-interpreted if necessary, not as a separate unit, but as part of a whole document, including what the translator knows about the author’s views, the context of the document, the events and history surrounding its creation, the other stakeholders involved and the topic being discussed.

In order to achieve this holistic view, the translator must devote time to doing research and gaining as much knowledge as possible about the topic and circumstances of the text being translated. Today’s translator is privileged in having at her/his fingertips technology tools that can provide her/him not only with portions of previously translated texts on the same topic, but also the entirety of world knowledge. Learning to properly use those tools while avoiding their pitfalls (such as insufficient scrutiny of machine translation) is an indispensable skill for a good translator.

Finally, we as translators should always keep in mind that language is not just about words, but also about the phenomena being discussed and their interconnectedness, which can only be understood using our knowledge of the world. In other words, translation is ultimately about understanding people, events, things and how they work. The translator cannot and should not pretend to be an expert in all fields, but he/she should exercise maximum curiosity, being always on a quest to widen his/her understanding of the world; calling on experts as necessary and following current events in a variety of areas outside of linguistics, from science and technology to literature, philosophy, and legal, economic, social and political issues. We do not always have the luxury to choose the content of what we translate, but we can choose to be informed readers who can navigate general topics, and who use their judgement to determine which translation assignments they should accept.

Recommended books on the topic of translation

Translation is, unsurprisingly, the subject of an untold number of books, academic papers, essays and other types of documents, especially since it seamlessly permeates other areas of knowledge, such as linguistics, language technology, psycholinguistics, semantics, complexity theory and many others. Below is a sample of resources that came to my attention, which can hopefully point the interested reader to additional sources on this vast topic. For the books and articles, only the title and author’s name are provided, since the rest can easily be found through an online search. The list includes websites and online journals which provide more up-to-date information as new theories and discoveries emerge in the field of translation. There are also a number or emerging YouTubers channels that look at language and translation from different perspectives and provide constantly updated and visually appealing content.     

I- Books

1. The Interpretive Theory of Translation, by Danica Seleskovitch and Marianne Lederer
2. Translation Quality Assessment: Past and Present, by Juliane House
3. Exploring Translation Theories, by Anthony Pym
4. Introducing Translation Studies: Theories and Applications, by Jeremy Munday.
5. Revising and editing for translators, by Brian Mossop
6. Translating as a Purposeful Activity: Functionalist Approaches Explained (Translation Theories Explored), by Christiane Nord
7. Translation Theory and Development Studies: A Complexity Theory Approach, by Kobus Marais
8. A Classification of Errors in Translation and Revision, by Gyde Hansen
9. Translation Revision and Post-editing: Industry Practices and Cognitive Processes, edited by  Maarit Koponen, Brian Mossop, Isabelle S. Robert, and Giovanna Scocchera

II- Articles

1. Translation quality assessment: a critical methodological review, by Chao Han, published in The Translator Online Journal
2. A Deeper Look into Metrics for Translation Quality Assessment (TQA): A Case Study, by Mateo Martínez, published in Miscellanea: A Journal of English and American Studies.
3. Towards a dynamic quality evaluation model for translation, by Sharon O’Brien, published in The Journal of Specialized Translation
4. The ethical dimension of translation revision. An empirical study, by Alexander Künzli, published in the The Journal of Specialised Translation
5. Retrospection methods in translator training and translation research, by Gyde Hansen, published in The Journal of Specialised Translation
6. Teaching and Learning Translation Revision: Some Suggestions Based on Evidence from Think-aloud Protocol Study, by Alexander Künzli, published in Current Trends in Translation Teaching and Learning, Helsinki University, 2006

III- Online resources

1. The websites of the various associations of translators and interpreters: Provide information on each association, articles and essays, including translation ethics codes, links to additional resources, information on professional conferences/meetings and how to participate in them, as well as career advice and opportunities
2. The American Translators Association (ATA) (https://www.atanet.org)
3. The International Association of Conference Translators (http://www.aitc.ch)
4. The International Association of Conference Interpreters (https://aiic.org)
5. The International Association of Professional Translators and Interpreters (https://www.iapti.org/)

• The United Nations Official Documents System (ODS): Contains a repository of all United Nations official documents in 6 languages (Arabic, Chinese, English, French, Russian, Spanish), and can be used for self-training. (https://documents.un.org)
• The Translator Online Journal: Is part of the Taylor & Francis Online of Peer-reviewed Journals and Contains academic articles on all aspects of translation, which can be accessed for free (https://www.tandfonline.com/journals/rtrn20)
•  The Journal of Specialised Translation: Contains free videos and downloadable PDF articles on various aspects of translation (https://www.jostrans.org)

IV- YouTube channels (can be found using the YouTube search bar)

1. The Anthony Pym Channel: By Professor Anthony Pym, who is a Professor of Translation and Intercultural Studies at the Virgili University in Tarragona, Spain. Discusses concepts of translation theory and teaching, including issues of artificial intelligence
2. The Freelanceverse Channel: By Adrian Probst. A channel for aspiring and established freelancers with focus on translation and language. Discusses many language-related topics including, the translation industry, freelancing, language entrepreneurship, etc.
3. The Langfocus Channel: By Paul Jorgensen. Discusses languages of the world, their characteristics, comparisons, history, etc.
4. The Human1011 Channel: Contains discussions on interesting aspects of language, including grammar, pronunciation, etc. as well as other topics.

About the author:

Ibrahima Diallo is a retired linguist who worked as an English teacher before joining the United Nations, where he worked for 25 years as a translator, reviser and language service manager. Before his retirement, he was Chief of the United Nations Verbatim Reporting Service.

His main working languages are Arabic, English and French. He also speaks Spanish and African languages.