What is Machine Translation Post-Editing (MTPE)?

Sep 26, 2022

Machine Translation Post-Editing (MTPE) is the process whereby humans edit machine translated text. Translators are commonly called a post-editor who handle this language editing process. Post-editing involves the correction of machine translation output to ensure that it meets a level of quality negotiated in advance between the client and the post-editor. The concept of post-editing is linked to pre-editing. In the process of translating a text via machine translation, the best results are gained by pre-editing the source text. Pre-editing is the process where a human edits a document before applying machine translation. Pre-editing should facilitate the process of machine translation by checking spelling and grammar, avoiding complex or ambiguous syntactic structure, and verifying term consistency. The main goal of pre-editing is by adapting the source document to improve the raw output of machine translation to reduce the post-editing workload. By applying the principles of controlled language, it increases the accuracy and quality of MTPE. Ideally after post-editing is finished the text should be verified to ensure quality, free of mistakes.

MTPE is developing as the optimal solution for various industries and business translation projects. Practically all computer assisted translation (CAT) tools now support post-editing of machine translated output. In recent years, MTPE comes with a variety of tools that make the life of post-editors easier. The first of these is translation memory (TM), which is a database of previously translated segments. If a segment stored in the translation memory (TM) appears in the source text again, the MT will automatically fill in the corresponding target segment for the post-editor to review. Another distinct feature is the term base, which is a manually entered list of bilingual terms for a specific industry or subject. These term bases are especially great for post-editors to assist them in the translation process. Also, in cases where multiple post-editors are working on a project it enables them to keep certain terms consistent throughout multiple translations. Another useful feature is a Quality Assurance (QA) tool, which helps spot any errors or inaccuracies that may have been overlooked in the post-editing process. QA tools are meant to ensure that the translation output is high-quality. Last but not least is data privacy protection. It’s best to opt for a customized MT engine—built using data provided by the proprietor company—instead of using commercial public MT engines.

What are the advantages of MTPE?

MTPE is used as a business strategy for translation with three main factors: time, cost, and quality. The delivery term for a project, the budget available, and the desired level of quality are all variables that influence a manager’s decision for their MTPE projects. MTPE combined with the post-editor’s experience and fine-tuning of data, it can offer companies these advantages:

-Reduce time spent by increasing translation efficiency resulting in quicker deliveries

-Reduce translation cost

-Increase consistency in translation comparable to purely human translation

-Large volumes handled in less time while maintaining adequate quality

-Accelerate distribution of multilingual products (time-to-market)

To ensure these advantages, MTPE needs to be continuously optimized to guarantee all the benefits mentioned above. For MTPE to work, a coherent translation process needs to be integrated professionally applying equally fundamental and interdependent factors. We recommend that you choose MTPE knowing and accepting the balance between time, cost, and quality. MTPE is advantageous if you want to translate high volumes of documents in a short period of time, but you may sacrifice on the overall quality of your translation. You must decide what to prioritize. You must choose but choose wisely.

What are the best practices for MTPE?

Here are some of the best practices for MTPE:

-Pre-editing is recommended where linguists check the source text before it is machine translated. Linguists should make sure the source text is error-free of spelling and grammar mistakes, consistent with terminology, and comprehensible throughout the document.

-Provide your MT engine with relevant term bases and glossaries. MTPE is more efficient with industry-specific terminology and reference data. This is especially the case for technical translations and other niche fields. The results of data-driven neural machine translation are much easier to edit. This helps the post-editor immensely.

-Post-editors should be educated not to under-edit or over-edit. Under-editing, or insufficient editing, runs the risk of ending up with an inadequate translation. Post-editors must make sure no information in the source text is left out in the translation. By contrast, over-editing like making preferential or stylistic edits that aren’t necessary, risks deviating from the source text. Additionally, the translation shouldn’t include information absent in the source text. It is okay if the output of MTPE feels less natural than human translation considering it delivers the original meaning properly. If the translation is correct overall and the meaning is comprehensible it’s best to leave it there.

-Utilize a Translation Management System (TMS) for better consistency. Using a TMS for post-editing provides a process of analysis that involves all relevant parties such as the pre-editor of the source text, the post-editor, and the manager. Feedback can be funneled through a TMS such as comments on the quality of machine translation, common mistakes, and frequent mistakes. This can be used to improve the accuracy of MT and make post editing projects efficient. Setting up a system for this type of review and feedback enables clients to improve their custom MT engine over time, which leads to improvements for the future.

What is ISO 18587:2017?

ISO 18587:2017provides requirements for the process of full, human post-editing of machine translation output and post-editors' competences. ISO 18587:2017 is intended to be used by LSPs, their clients, and post-editors. It is only applicable to content processed by MT systems.

In 2017, ISO 18587:2017 requirements were published and became visible proof to the certification of the post-editing standard. It demonstrates standardized quality processes, comprehensively transparent project procedures, including personal consultation, adjustment of the processes and the highest level of data protection. LSPs must apply the same high-quality standards to MTPE projects as they do to regular specialized translations. Post-editing is defined as light and full post-editing in the context of the European Commission Translation Service.

The first approach light post-editing is where MTPE only focuses on the essentials such as grammar, spelling, and translation accuracy. Light post-editing aims at making the output simply understandable. Light post-editing implies minimal intervention by the post-editor, with the aim of ensuring quality is "good enough" and "understandable"; the expectation is that the client will use it for inbound purposes only, often when the text is needed urgently, or has a short delivery term.

The second approach is full post-editing, where the editor will focus not only on the essentials but also take the time at making it stylistically appropriate and consistent throughout, additionally focusing on details like whether all expressions are localized. The expectation of full post-editing is the outcome will be a translation that is understandable, stylistically appropriate, used for assimilation and dissemination, and for inbound and outbound purposes. The quality is expected to be publishable and equivalent to that of a human translation. Full post-editing involves a greater level of intervention to achieve a degree of quality which needs to be negotiated between client and post-editor.

idioma® offers its ISO 18587:2017 certified MTPE service to many of its clients. We provide a customized neural machine translation engine to increase the quality of MT output which requires less post-editing effort. We use our translation management system (TMS) and computer-assisted translation (CAT) tool where post-editing entries, speed and linguistic quality assessment results of the post-edited texts can be compared. This tracking and measuring mechanism built in our MTPE environment increases efficiency and improves quality. To satisfy our clients requests we incorporate the best guidelines for MTPE while carefully evaluating the clients’ necessities to execute an optimal solution. We thoroughly explain the cost and benefits involved when deciding together with our clients on MTPE. With the proper tools, practices, and mindset in place, post-editing can be an alternative to traditional translation. idioma® ensures the highest standards with its MTPE translation services.

If you have any questions about MTPE or would like a translation quote, please contact us at info@idioma.com.