One of the most discussed and attended events for translators during the FIT XIX World Congress were Rosana Wolochwianski’s presentation on Machine Translation Post Editing. Since I believe that Machine Translation is here to stay, whether we like it or not, I wanted to hear more about Post Editing. I found her presentation very enlightening and particularly liked the interviews with current post editors. Here is a summary of what she said.
Translation is changing in the information age. Electronic technologies now allow multilingual and semiautomatic text generation with translation memory systems, content management systems, terminology management systems and machine translation systems. The development has led to cheaper, more accessible, faster translations that are easy to update and highly consistent. This means that translators can produce more in less time.
There are two different uses of machine translated text; translation aimed at assimilation by the user and translation aimed at dissemination (spreading) by the user. For many companies, translations aimed at assimilation (for their own consumption) does not have to have such high quality, but translation aimed for publishing should still be reviewed and corrected by professionals. Professional translators usually always strive for optimum quality. These are called “traditional” translators in this presentation. However, current end users might not always strive for optimum quality.
So what is involved in machine translation post editing?
Rosana divided post editing work into three stages:
– The pre-editing phase
– The MT-tool selection phase
– The MT post editing phase
The pre-editing phase is a necessary stage for successful use of machine translation. This is when you remove typos and spelling mistakes, unnecessary hard returns or format issues from the text. Non-translatable items are also tagged.
To select the right MT-tool, you need to first perform a good analysis of the project details. Is there a reliable and large TM, does the target language have strong syntax requirements, or would a hybrid tool be appropriate, such as a tool trained for a specific client?
There are different types of MT post editing. Complete post editing is when high quality is required; the only aim is to improve speed. Then there is minimal post editing, which oftentimes generates resistance. Finally there is rapid post editing, when you only remove blatant and significant errors and is only for understandability. The two latter types are the ones that “traditional” translators have a hard time stomaching, since we only strive for the best linguistic rendering. ‘’
Pros and Cons of Machine Translation Post Editing vs. Translation
|+ Time gain+ Nothing is skipped or repeated+ No typos or spelling mistakes introduced
+ No “blank page effect”
+ Large volumes of work available
|– Recurring errors- Typos or spelling mistakes in original are not recognized- Non-translatable elements get translated
– Dull or non-user friendly interfaces
– Constant exposure to flawed language
– Lower quality environment/less creativity
– Unrealistic expectations
– Pressure to lower rates
What is needed from a post-editor?
– Long-term commitment to improve both the editor’s skills and the machines skills.
– Innovative responses to MT-errors
– Creative problem-solving
– Good keyboarding skills
– Certain degree of tolerance
– Ability to draw clear boundaries between improvements and corrections (just as a regular editor)
What does a post-editor need?
– Be able to volunteer for the task and not be forced into the role
– Be given time to learn
– Be heard when they refer a problem
– Receive proper training
– Be paid according to the time and effort applied
– Alternate with other tasks, in order to keep sharp
The machine translation post editor is a growing profession. In order for the post editor to thrive it is important that the post-editor is seen as a valuable part of the machine translation process, more as a co-developer, who can provide important information and innovation and improve performance by providing constant feedback, suggesting improvements and developing new solutions.