10/04/2024

It’s Quality Time again. In a new episode of our expert interviews, Eva-Maria Tillmann, Head of Quality Management at oneword, spoke to Ursula Reuther, Chief Linguistic Consultant at Congree Language Technologies. An insightful discussion about how practical experience impacts standardisation and the useful impact of standards on writing and translation practice, how conformity with clear rules is a decisive factor for creating comprehensible texts that are suitable for translation, and why AI should be involved, but must not be used in isolation.

Congree Language Technologies and oneword have several areas where their work overlaps. As a software manufacturer for authoring support, Congree provides some of the leading technologies for formulating consistent texts, including defined style rules and standardised terminology – all areas that oneword focuses on too. Both companies are also DIN members. Ursula Reuther represents Congree. Since 1986, the linguist and translatologist has been working in development and project management in the fields of language processing and language technology. Eva-Maria Tillmann, Head of Quality Management, is an active member of the committees, representing oneword.

Eva-Maria Tillmann (oneword GmbH) and Ursula Reuther (Congree Language Technologies GmbH)

Eva-Maria Tillmann (EMT): It’s great that we have been able to find some time to chat, Ursula. After all, we have known each other and had a lot of respect for each other for four years now through the joint working group on translation-oriented writing at DIN, which has brought about DIN 8579. I’m looking forward to the conversation personally, but I think it will also be interesting for many of our clients to see how you as a tool manufacturer engage with the topic. How did you come to work on this standard in the first place?

Ursula Reuther (UR): Congree is actually part of the reason the standard exists. The idea for it emerged at the tekom spring conference in 2019 in a discussion with a major translation service provider. The consensus was that it would be great to have a standard for translation-oriented writing. There are various standards and metrics for evaluating translations, but, so far, there had been nothing to evaluate the source text.
However, we had no experience of how to go about developing a standard, so we looked for fellow campaigners: people who already had experience developing standards and who submitted our proposal to DIN, which also liked the idea. That’s how this working committee came about and I became a member of it.

EMT: Yes, I remember. At DIN, it was initiated by Professor Christoph Rösener, who is also project manager for the international standardisation project with me. He suggested the standardisation project at the time and I joined as an expert because our committee was asked whether we would like to work on it together with the technical editors. That was an exciting time. We worked very intensively on the text; each of us had our own chapter. I was involved in the topic of formatting. What about you?

UR: My part involved more linguistic aspects, such as terminology, grammatical and stylistic aspects, and potential spellings.

EMT: What were you able to contribute to the standard in particular thanks to your professional experience or from your background at Congree and in software development?

UR: I’ve been engaging with the issue of comprehensible writing for more than 20 years, even before Congree existed. At that time, I was still at the IAI, the Institute for Applied Information Research at Saarland University, which then merged with Congree. There we were already carrying out research projects on this topic, the results of which have also been incorporated into our products. And comprehensible writing runs through the whole story. I dealt with both the theoretical aspects and the concrete implementation of software rules to test comprehensible writing.

And, when it comes to the standard, comprehensible writing has many cross-overs with translation-oriented writing. I was therefore able to incorporate some of the rules we had already developed, and also a lot of customer feedback. The rules we have implemented have often been developed by directly interacting with customers or based on their feedback. Certain rules ensure both comprehensibility and fit-for-purpose translation. And when there were a particularly large number of queries from translators, customers usually wanted to know what they could do better in the text to make the processes run more smoothly. I was able to incorporate all of these things.

EMT: We feel the same way, of course. We also see that problems in the translation process often relate to the source text, spelling, inconsistencies or formatting. We also give our clients regular feedback on this, of course. Or we have to charge for additional work when we optimise a text, and, in the interest of our clients, we want to avoid that. This is precisely why we and our translators are very interested in ensuring that the source-language content is optimised and suitable for translation.

You say that you have always done this in principle and that Congree is a software that has been able to check source texts for a long time. What was your particular interest in having a standard for this?

UR: Most of the rules were actually already included in the product. What we have added through the standard, however, is that we have now bundled together those exact rules that we described in the standard or that we recommend following. We have added our own pre-set, i.e. a predefined rule set that contains exactly the rules that are relevant for translation-oriented writing. And this has been very well received by customers.

EMT: So customers can activate an additional pre-set, depending on whether the text is to be translated?

UR: Exactly. Either they compile their own rules or there are various predefined rule sets. For example, there is a set of rules for the tekom guideline and a best practice configuration for beginners, and now there is also one for the standard on translation-oriented writing.

EMT: So you’ve already been able to contribute many of your assets to the standard. Did the standard also add a few points that you hadn’t included before?

UR: Yes, it did. It was a win-win situation for us.

EMT: Can you explain in principle what authoring support at Congree or the conformity check looks like when customers have activated this pre-set for translation-oriented writing?

UR: The procedure is the same regardless of whether you use this package of rules or another one. The software integrates with various editors, depending on what the customer is using, from Word to editorial systems with complex document structures. The language check is integrated and users are then shown rule violations when writing. The checks can be carried out as your write, or you can first write a chapter or a paragraph and then check it. The language check then checks for spelling and grammar as well as compliance with corporate terminology. And then there is the stylistic check. If there are ambiguities or complex sentence structures that are difficult to understand, these are displayed for the editors. They are made aware of which sentences are problematic for translation and can take appropriate action.

EMT: That’s exciting, because it reminds me a lot of translation technology and the checks that exist for target-language texts. Is there any content from the standard that you think might be difficult to map or check at the moment?

UR: Yes. For example, the last chapter of the standard contains a matrix that documents which rules can actually be checked automatically by the authoring support. Some columns are still marked ‘No’. These are, for example, cultural specifics that cannot yet be covered by linguistic methods alone.

Here, too, AI is of course increasingly catching up with us, and we have now integrated AI into our software and use it to give users rephrasing suggestions where linguistic methods alone do not work.

EMT: In other words, if a rule is not followed, editors who are notified also receive a suitable suggestion from the AI on how to correct it based on a specific rule that was obviously not adhered to?

UR: Exactly. And you can then simply click to accept this suggestion.

EMT: So, it’s basically like when we use AI: some things can be done without humans, but not entirely.

UR: That’s right. Nothing is automatically replaced. Ultimately, the human is really the one who decides whether the improvement is actioned on or not. That was the overriding principle right from the start.

The linguistic check can also provide suggestions for less complex reformulations, which can then be accepted. However, if, for example, you have to change a sentence from the passive to the active voice and have to move the parts of the sentence around a bit, it’s a bit easier with AI.

EMT: We are now also working together on the upcoming ISO 18968, the international standard for translation-oriented writing. We are both on the task force to provide examples. There are many examples in the German standard, but we obviously have to revise them substantially, not only because the source language in the English standard is completely different. We have already adapted a lot, optimised a lot and still have a lot of work ahead of us, because the standard that was written entirely for the German language – DIN 8579 – is being internationalised and will be made applicable to as many languages as possible in future.

What do you think, are we on the right track? Do you have any other ideas from your practical experience that we can incorporate?

UR: The biggest challenge is definitely shaping the whole thing in a way that works for all languages because, as you say, the German standard really is related to the German language. And we are now realising from the examples that a very large number of language-specific elements do not apply to all languages. But I think we are making progress and are definitely on the right track, and I am very much looking forward to the initial feedback on our task force’s work from the many member states.

EMT: I see it the same way. And I’m looking forward to what lies ahead, because it’s really enjoyable to work on the standard with this great international group of experts.

But, for now, thank you very much for the great conversation and see you soon, Ursula.