MTPE and technical professional translations

26/08/2022

We recently explained how machine translation can be used in a professional business context and how the IEC/IEEE 82079-1 documentation standard ensures that technical documentation is comprehensible and usable in all languages. Now comes the reality check: How exactly do you use machine translation for technical texts, technical documentation and communication and what do you have to do for it?

Professional, systematic and efficient translation services have always been based on technology. Machine translation is yet another component that has been added and needs to be integrated in a targeted manner. In this article, we provide an overview of the development of the tools and classify how they are connected and what to look out for in order to achieve the best possible results in the right combination.

“Machines” have long been used in professional translation and language services. This field is a far cry from the pen and paper translation of the past. It is always focused on efficiency, consistency and quality, equipped with innovative tools and cast in optimal processes.

The mid-1990s was when software-supported translation first arose. CAT tools (Computer-aided/assisted translation tools) enable documents to be opened in an editor regardless of format, irrelevant content to be hidden and texts to be translated segment by segment.
Translations can be stored in integrated translation memories (TM), and then reused: when the translator loads a new document into the tool, the translation memory accesses translations from previous projects in the background and inserts (excerpts of) texts that have already been translated, so that only new texts need to be translated. Since the introduction of TM, you no longer have to search for identical texts in your own files, which is a cumbersome task. Instead, the texts are automatically detected in the next project and the translations for them are suggested or inserted immediately. A quantum leap for consistency and reduced translation costs.

In terms of consistency, the emergence and integration of terminology databases directly into the translation environment also meant that compliance with specified technical language could be implemented and checked immediately during the translation process. After all, technical language is the be-all and end-all, especially in technical documentation. A CAT tool detects a term from the database in the source text and suggests the foreign-language equivalent that is also stored. Compliance with the specifications can then be checked again as part of automated quality assurance. Therefore, companies and translators can ensure that all defined terms are used correctly and consistently.

QA tools are other “machines” used for quality assurance to check that the translation is correct and complies with the specifications. Familiar functions, such as the spell checker, are used as well as checks related to style guidelines, forbidden spellings, terminology, numbers and, finally, the correct use of tags so that protected or formatted contents of the source text are also correct in the target text. QA tools are partly integrated into CAT tools but are also available as additional software, depending on the requirements.

CAT tools, terminology databases and QA tools are aids for translators and language service providers that support the translation process but do not take away the actual translation step. This changed with the advent of machine translation (MT), which does the actual translating of texts without the active intervention of a human. For a long time, however, machine translation results were mediocre and were more amusing than useful. However, with the integration of neural networks, machine translation – now called neural MT – has experienced a surge in innovation over the last five years and is now a very practical solution as an additional tool in the translation process. The machine translation results form a very good starting point for using Machine Translation + Post-Editing (MTPE) to have the translated content checked, corrected and standardised by trained post-editors.

CAT tools are still state of the art, as are integrated terminology databases and QA tools. In addition, machine (pre-)translations from relevant engines can now be accessed via API or plug-in interfaces. In practice, this makes for a winning combination: existing translations are taken from the translation memory and new segments are first pre-translated by an MT engine. Consistency, style and accuracy are ensured through content checking and post-editing.

If the source text is suitable and, most importantly, if the appropriate MT engine is used, machine (pre-)translation, including post-editing, takes up to 50 percent less time. Translators and post-editors are therefore able to handle a higher workload while producing consistently high quality, and clients receive their translations more quickly. Given the larger volumes, this is also reflected in reduced costs.

Artificial intelligence needs human expertise. After all, humans can read into the texts, check their usefulness, recognise ambiguities and put everything together into coherent documents. The machine also cannot process all file formats. For example, when translating technical texts for a machine’s HMI/operator panel, a certain number of characters must be used, so that they can be displayed on the screen. Machine translation cannot yet guarantee that these character limitations will be observed.

Inconsistent corporate communication is another risk. This is because MT engines that can be used for free are not familiar with the company or its specific terminology. Ergo, rapid and uncontrolled translations with one or even several services can quickly lead to a proliferation of terms for one and the same thing and, therefore, can work against systematic terminology management.

Freely accessible MT engines (only available in German) are good and free of charge, but it means that they are used carelessly and may work against the protection of sensitive data and confidential content. When employees copy and paste a text, this text is always transferred to the provider’s server and processed there. Because machines are constantly learning, these texts are also used as training data and to improve the offering. The data leaves the company’s sphere of influence and there is no guarantee of what happens to it or how secure it is.

Therefore, it is important to have an approach that is fundamentally secure. At the very least, this includes ensuring that data security is guaranteed through subscriptions and licences, and checking carefully in the General Terms and Conditions how long content is stored for and what it is used for.

Relying exclusively on machine translation for technical editing and documentation would be fatal for all the reasons mentioned above. If only to avoid danger, it is never advisable, because there are great warranty and liability risks with technical documents.

However, it makes sense to use it and it is right to use it when you consider the possibilities it brings: with help from expert post-editors who check, correct and standardise the pre-translated texts, machine translation becomes suitable for use in practice. In professional hands, the post-processing is more and more standardised and the post-editing effort is constantly decreasing.

To complete the arc mentioned at the beginning: Professional, systematic and efficient translation services have always been based on technology. Machine translation is another component that has been added that can be integrated in a targeted manner – assuming that there is consultancy and expertise about machine translation and that sensible engines and post-editing are used – in order to combine all the parts sensibly and use them profitably in the optimal translation process. Especially when there are large volumes and tight deadlines. And especially in the technical field, where machines make it possible to use the specified terminology and fundamentally important repetitions so that there is as little new translation work as possible, the costs and time involved are reduced and the highest quality is still ensured by terminology and post-editing.

Do you need professional translation of technical texts, technical documentation and communications? Are you considering using machine translation? Then talk to us about your specific requirements and goals. Our technical translation and MTPE experts will be happy to advise you on all the measures and options.