TRY AGAIN, FAIL AGAIN, FAIL BETTER

Making errors in translation is nothing new. Fatigue, time-pressure, an unfriendly working environment – all are perfectly legitimate explanations. What really matters is being aware of where those errors come from and finding out if and how you can prevent them. In other words, the first step is to identify what errors you make, and the second – to identify the right ways of preventing them. When it comes to prevention, electronic tools offer immense help. A machine will not miss a small error after processing hundreds of thousands of words, but humans are fallible, particularly when tired or working under time pressure (Drugan 2013: 93).

Knowing the errors that you make is not so self-evident. Regardless of whether you are an individual translator or a large organisation you need a record of the errors made with sufficient level of detail to enable you to draw useful conclusions and learn from them. You can either set up your own repository or browse the internet for examples. Ultimately, however, what works best are errors directly relevant to your everyday work, so drawing from your own experience is most effective. Some examples of errors used in this paper have been taken from the corrigenda to the EU legal acts, which are published on EUR-Lex. Others were errors encountered in every day translation work, which were identified and prevented just in time to make sure they never made it into the final product.

4.1 Accuracy¹ (omissions, additions, mistranslated, untranslated)

Where target text fails to correspond to the source text, direct comparison is key to spotting an issue. The number of possible issues is vast. Mistranslations, for example, can cover anything from a missing or added negation to inconsistent numbers. Some real-life examples include mistranslating:

(a) key descriptors: “fresh, chilled or dried” as “fresh, chilled or frozen”;

(b) names of countries: “Russia” as “People’s Republic of China”;

(c) types of legislation: “Regulation” as “Directive”;

(d) numbers and units of measurement: “7500 kg” translated as ”3.5 tonnes”;

(e) names of months: “June” as “July”;

(f) numbers: “2012” as “2021”;

(g) negations: “No fragment longer than 100 mm in length shall be allowed” as “Fragments longer than 100 mm in length shall be allowed”;

(h) logical and time determiners: “unless” as “as long as” (or German: “solange”).

It would be unjustified to assume that an experienced translator cannot tell the difference between Regulation and Directive or 2012 and 2021. Instead, when investigated, these errors frequently find their explanation in the working methods involving tools and in the failures of fatigued human eyes. In examples (a) to (d) the mistranslated words and numbers were parts

1 In this document I use a subset of MQM/DQF error categorisation system.

of longer matches inserted into the segment from either an external translation memory or an earlier segment from the same text. A match close to 90% needs some minor editing, which – if the segment is long and the necessary changes very minor – can easily go wrong, i.e. the changes required are ignored. Examples e and f demonstrate the weaknesses of the human eye which gets easily cheated when it comes to minor details, like one or two letters or digits, i.e. the change required is so minor that the human eye is tricked into thinking that no change is needed.

Especially when working on screen and under time pressure. Finally, examples g and h show how the human brain likes to play tricks on us by mixing up logical relations, ignoring negations or confusing quantifiers. It is not uncommon to see “and” translated as “or” as well as “less than”

translated as “more than”.

Additions usually include single words and phrases, while omissions can cover both words/

phrases and, surprisingly, whole segments. Real-life examples confirm once again that most errors in these two subcategories result from failures to properly edit high matches (single words/phrases) or erroneous manipulations in the CAT tool (whole untranslated segments) or in the target files (e.g. when track-changes get untidy).

4.2 Terminology (terminology)

In a perfect scenario, correct terminology is extracted, prepared and made available before the translation starts. In such a case a terminology error consists in the failure to use the right term or to use it consistently throughout the document. This type of failure is frequently caused by excessive and hasty editing or accepting translation memory matches without proper attention.

Some real-life examples:

(a) “Technically permissible maximum laden mass” translated as “maximum mass”;

(b) “ted member state” translated as “applicant member state”;

(c) “Regulatory technical standards” translated as “regulatory technical requirements”;

(d) “Waste gas streams” translated as “waste water streams”.

All four examples illustrate a frequent issue behind terminology errors, which often result not from the translator’s failure to follow the prescribed list of terms, but yet again from the fact that when segments of similar wording, with only minor differences, are scattered all over the text, a hasty revision tends to ignore the changes required in the imported high match. In those particular cases the required change concerned the terminology.

4.3 Fluency (grammar, inconsistency, punctuation, spelling)

These errors, related to the linguistic well-formedness of the text, are generally seen as less harmful since they seldom impact the sense and the usability of the text. Still, their importance should not be underestimated. As Brian Mossop points out: misspellings and typographical errors […]

suggest that the author and editor are sloppy thinkers, and that the publisher tolerates carelessness.

As a result, readers may lose confidence in the actual content of the work (Mossop 2014: 44).

Some very convincing examples of rather harmful spelling errors include: “pubic administration”

or a “daft regulation”. These two are particularly dangerous, as not all spell checkers are able to detect them.

For grammar, punctuation and spelling there usually exists a shared reference, specifying relevant rules to be followed, e.g. a grammar book or a language authority issuing guidance where doubts are raised.

4.4 Style (company style, register, clarity)

Company style, whenever defined, contains a set of typesetting and editorial rules to be followed.

The Interinstitutional style guide published by the EU’s Publications Office defines stylistic rules and conventions for texts intended for publication in the Official Journal of the European Union. These include, among others, instructions on how to handle footnotes and references, addresses, phone numbers, references to the Official Journal, official names of DGs, also rules on punctuation, capitalisation, acronyms, hyphenation, spelling, plurals, participles, numerals, inflexion, geographical names, names of countries, etc. For example, based on the Irish constitution and according to the guide, the correct official name is “Ireland” and not “Republic of Ireland”.

The guide also lists protocol names of countries (to be used when the State is referred to as a legal entity, mostly in official legal EU texts) alongside their geographical names (to be used in less formal texts, like press releases). For example, where in official contexts, like trade agreements, you must speak about the Kingdom of Spain, in less formal texts you should limit yourself to Spain. In some languages it is common to use “Holland” when informally referring to the Netherlands, but according to the guide the latter is the correct short name to be used.

Register errors are those where the level of formality is not in line with the text specifications or language conventions. For example, a formal letter from an EU Commissioner is not supposed to contain colloquialisms or contractions (isn’t instead of is not) and is supposed to address its audience with the right personal pronouns, avoiding excessive familiarity. In some languages like French, German or Polish, there are clear rules about what personal and/or possessive pronouns can be used in a formal address. Conversely, a text addressing the public should not be written in a formal, complex language, full of long sentences and intricate expressions.

Clarity errors can take all sorts of forms and are frequently on the verge of mistranslation. They frequently refer to cases where non-idiomatic expressions lead to ambiguity that is not present in the source text or where expressions used are less precise than intended in the source text. Also, these errors are frequently seen when phrases or sentences used in the source text are extensively long or contain internal references that are not clear. For example, in the following sentence: They have been subject to an appropriate pre-employment check and/or background check as well as trained in accordance with point 11.2.7 of the Annex to Implementing Regulation (EU) 2015/1998, the phrase

“in accordance with” refers to the word “trained” only but was erroneously translated as if referring to the phrase “They have been subject”.