18/10/2019

Highest quality in all processes – that has always been our commitment. We have been certified to the ISO 17100 translation standard since 2013 and are now also actively represented on the DIN committee. In September 2019, the next logical step was to get certified to the DIN ISO 18587 standard for the post-editing of machine-translation output.

Transparent processes are particularly important to us when collaborating with our customers: What happens to a text when it comes to us, how are our prices calculated and who is involved in the project? Since the DIN ISO 18587 standard was published in spring 2018, we have been working according to standardised processes so that we can also fulfil our commitment to be transparent and produce high quality work when editing machine translation output, a field that is in high demand. These processes include consulting with our customers, preparing quotes, conducting a feasibility analysis, ensuring our post editors are qualified, post-editing and ensuring quality in all MTPE projects (MTPE = machine translation and post-editing), as well as performing evaluation processes and ensuring data security. We had all these processes examined by an auditor in September 2019 and are pleased to be one of the first language service providers in Germany to be DIN ISO 18587 certified.

For you as a customer this means that your MTPE project is not only traceable at every step, but is also in the best hands. With our customer-oriented service package oneMTPE, you receive an all-round service from the first consultation to the continual feedback process. Our team of experts will advise and support you when you introduce machine translation and will be on hand to provide advice and assistance during initial testing. Our internal feasibility analysis ensures that no text is categorically excluded from the MTPE process, but always looked at in detail in terms of its economic efficiency and potential savings. Because one thing is clear: machine translation (MT) and human post-editing offer high potential savings, but, depending on the type of text, language pair and subject area, they can also take more time than a human translation. In order to make the most of the time and cost advantages of machine translation and post-editing, each project or project type must therefore be considered individually.

Once a text has been machine-translated, our trained and experienced post-editors come into play. They review the machine translation output, aiming to use as much of the machine generated content as possible, and have to make decisions about the usability of a segment within seconds: can the machine translation be left unchanged, does it need to be adapted or does the segment possibly need to be completely re-translated? The post-editor’s progress depends largely on the quality of the machine translation output and the translation requirements. As a rule, in addition to linguistic editing, specifications from style guides or a terminology database are also used. Although these specifications make the editing process more extensive, they also mean that the target text is more in line with individual customer requirements. However, irrespective of the different quality objectives, post-editing processes that follow ISO 18587 must result in a translation of a quality that is comparable to a native-speaking human translation.

Clean terminology management is more important than ever for machine translation: this is because an untrained machine cannot know what you call certain products or services in your company. The generic machines in particular, i.e. the common neural machine systems like DeepL, for example, which are currently on the rise, do not allow corporate terminology to be integrated into the machine translation process. That means that adapting the target text to the terminology specifications is a task for the post-editor, who, in a best-case scenario, has access to a well-maintained terminology database in their working environment. The post-editor’s work ensures that the finished translation contains all the terms stipulated by the specifications. During the post-editing phase, a well-maintained terminology database not only ensures that corporate terminology is adhered to consistently. It also has a significant impact on how quickly the post-editor can work, which has a positive effect on the time and cost of each MTPE project.

After post-editing, oneword carries out a comprehensive quality assurance process and analyses the MTPE process: How many machine translated segments had to be corrected by the post-editor? How many of the corrections were only minor modifications and how many segments were completely rewritten or re-translated? If you would like us to, we can also share these analysis results with you and in so doing, after several projects, categorise certain text types as being more or less suited to the process. As part of oneMTPE, you will also receive feedback on every project so that we can continually optimise the process together. Here we can, for example, point out missing terminology entries, make suggestions for optimising your source texts or create common style guides for MTPE projects. Regular efficiency measurements not only reveal potential savings, but also pay off right from the start.

So, at oneword, we don’t really talk of „translation at the click of a button“. Because, in spite of artificial intelligence and machine translation, every oneMTPE project involves a lot of common sense and expertise, as well as our passion for language and our high quality standards.