Ongoing projects are fragile things: text is added, things change, new content is incorporated, while some of it remains the same. In the meantime, new translators, editors and even project managers and engineers are brought onboard. The project grows and more hands are needed.
A TM serves a very positive and helpful purpose, especially when it comes to large or never-ending projects: they help ensure consistency in wording and terminology, as project-specific glossaries can have their limitations, they help avoid overpaying for text that never changes and allow for new translators and editors to be incorporated into the process without having them retranslate text that needs to remain standard, such as copyright notices, contact information, etc.
Yet translation memories can also impose uncomfortable restrictions. Show of hands: How many times have you gotten a translation request where you were kindly asked to not even go near the 100% matches committed to memory? These purportedly perfect matches are at times anything but, and not touching them becomes a less reasonable request than might be readily apparent.
One of the greatest weaknesses of the segmentation model is that it leads CAT tools to treat text as a series of strings, single, individual, estranged from everything else, but the text never ceases to be a text. It has a context and very much depends on the other text around it, on the purpose of what it is describing. Take for instance the word “store”. Taken by itself, as a noun, it could refer to a shop, but it can also be understood as the imperative of the verb “to store”. In English they are identical, but in most other languages they won’t be. Only the context around it can clarify the issue for us.
The string, however, does not only have linguistic sense: it may also have a function, as a link in a website, as a button in a software program. The translator or editor may only see the linguistic side, while the programmer or engineer can see it for the coding underneath. A phrase is not just a phrase, a word is not just a word: it is also part of the code, and it also exists within that technical context. If changed, it can have unwanted effects in the system. This is one of the main reasons for wanting to leave 100% matches untouched.
The other reason obviously has to do with the fact that the translator won’t have to be paid for repeated content. But there is yet another reason: the client and project manager are ultimately dealing with something that they don’t necessarily understand, that is., text written in a foreign language. Project Managers often handle projects in several languages and, regardless of how much they might trust their editors and translators, they kind of have to make a leap of faith when you tell them you want to change a character in Traditional Chinese. Changing the TM might also involve telling the end client that a mistake was made by the agency’s team, which they fear the client will frown upon. “Why wasn’t this error caught earlier” is the question that will inevitably thunder down form the heavens. Nobody wants to hear their TM’s are not to be trusted.
Where does this implicit trust in a translation memory come from? Because a text that is in the TM has already supposedly gone through the hands of an expert translator and editors, finding a glaring error in the TM puts into question the previous effort that was made. Legacy is not only about what’s encoded in the program or what terminology the users might have gotten used to by the time you get the project. When you tell a client that a previous translation is faulty, it might send them into a bit of a panic about their process. Criticizing a previous translation is ultimately also criticizing the client that hired them, as you are showing them that they did not choose correctly. Also, they may have had a particular translation committed to memory for years, may have used it in their packaging, materials, website and all sorts of communications. Telling them that there was an error in all that content is kind of hard to swallow: you’re basically telling them they’ve been sending out a flawed product that they thought perfect and finished. If the error you point out is what is often termed “critical”, that is to say, a significant spelling or grammatical error, and you find more than one in several strings, that will send them into an even bigger panic. They have to consider that the resources they deposited their trust in all this time were wrong, especially if there was more than one stage of editing involved. A translation project is ultimately based on weak, implicit trust: put it into question, and the whole edifice comes crumbling down.
Where do the errors in the TM come from in the first place? We must not forget that the vast majority of the time a translator is working in solitu, without direct contact with the client, sometimes without having ever looked at the source material the source segments were taken from, such as a website or software tool. The translator is also often part of a larger team of translators, spread out throughout a country or several, translators with varying degrees of training, expertise in the subject matter and proficiency in the target language. Yet the translator alone is responsible for coining new terms or coming up with catchy marketing phrases, and it must be admitted that not all translators are prepared for this. If under time constraints or working in simultaneous project (a reality of the market), the translator might not have enough time to do research into appropriate terminology.
This is why terminology and stylistic guidelines cannot be left to chance. When embarking in a large, ongoing project that involves standardization, branding and coding committed to a TM, measures need to be taken to ensure terminological consistence. If a project is going to be worked on by several translators with different writing and research styles, then one same editor, preferably with expertise in the field and proven written command of the target language, should ideally review the entire content. The editor would then issue stylistic guidelines and help put together a glossary.
Translators need to be aware that they have a responsibility towards the final content. If something is not clear from context, then it should be asked. The translator needs to acknowledge his or her part of responsibility, do research accordingly and learn to identify terminology issues.
It is not all on the translator’s side, though. LSP‘s and clients in general need to adopt a more flexible attitude towards their text. They must never forget that they are dealing with language, and that context reigns king over all other concerns. There are many reasons why a previously accepted segment needs to be changed, ranging from an error from the previous team to a change in the context. Keeping an open mind about the fallibility of translation memories will ultimately help improve the content for the benefit of all involved.
Entering an ongoing project that has been alive for a long time may be a blessing and a curse. On the one hand, you have all those TM matches to use as reference for terminology and phrasing. This is especially helpful if you are new to the material or have limited knowledge about the field. But you might find yourself severely limited because, even after pointing out errors and explaining why they need to be fixed, your client might decide to not do anything about it. Then it boils down to whether you can live with that or not.