Professional Post-Editing: The Alliance Between Artificial Intelligence and Human Expertise

Professional post-editing is the process of proofreading and correcting a neural machine translation by a professional human translator.

Born alongside machine translation, professional post-editing earned its patent of nobility when algorithm based machine translation (Systran, Apertium, or Google Translate until 2016) saw a significant rise in its reliability and accuracy through becoming neural machine translation. However, in spite of such a leap, machine translation remains in certain cases, far from perfect and yields results requiring proofreading and refinement.

Tangibly, the translator pre-translates the source text via one or multiple neural translation engines. Following comparison between different machine translated versions, the translator selects and compiles the best translations in order for him to proofread and refine them.

The 6 steps of professional post-editing:

Machine Pre-translation: One or multiple pre-translations can be achieved through online machine translation softwares, translation APIs, and machine translation tools integrated into localisation platforms.
Quality Assessment: The professional translator assesses the pre-translation quality based on various criteria such as accuracy, coherence, fluency, and pertinence to the context. If multiple pre-translations were done, the professional translator will then compile the best translations following quality assessment.
Problem Recognition: The professional translator then locates and identifies errors, incoherences, mistranslations and stylistic issues within the machine translation. That may include grammatical errors, inadequate word choice, terminology errors or global text comprehension issues.
Correction and refinement: The translator performs necessary corrections to enhance and refine the initial machine translation. This step implies sentence modifications, text restructuring, inadequate term substitution, fine adjustment of terminology, and addition of missing information or, removal of unnecessary segments.
Coherence Control: The translator ensures the translation remains coherent throughout the entirety of the text while respecting specified stylistic and terminological standards and guidelines. The translator also uses terminology management tools n order to guarantee the coherent use of specialized terms.
Review and Proofreading: Once the modifications are performed, the translator carefully reviews the post-edited translation to ensure its quality, precision and fluency.

The benefits of post-editing:

Time Saving: Professional post-editing cuts translation working hours to half or even a third of the standard time. Instead of entirely translating a text, the translator focuses on proofreading and refinement of an existing machine translation. This proves to be particularly beneficial with large projects or tight deadlines.
Profitability: A significant cost-effective method seeing that a large part of the translation is machine-produced. Clients can achieve significant resource savings through the use of machine translation softwares calling upon a translator solely for proofreading and correction.
Post-editing is also beneficial for the translators themselves. It allows them to work more efficiently and take advantage of AI abilities all while maintaining their crucial role in text adaptation through the conformity of linguistic and cultural nuances.

For which project?

It is used for standardized document translation, largely consisting of short sentences, implying perfect readership comprehension with no intention to trigger judgement or an emotional response.

Therefore, it is convenient for:

Software translation
Database translation
Translation of a number of technical documents, reference systems, and standard equipement user manuals.
Product Database translation.
Tourist information documents and lodging guides translation
Legal and administrative contract translation

The limitations of post-editing:

Limited quality: One of post-editing’s biggest obstacles resides within the initial machine translation quality. Machine translation softwares are not perfect and could yield low quality translations requiring a more intensive proofreading process. Furthermore, post-editing may be complex and challenging considering that the translator must juggle between text enhancement and the preservation of the original meaning.
Loss of the original writing style: Machine translation is incapable of reproducing the style and tone of the text and will try to standardize the target text.
A lack of accuracy

When to avoid it? One must not opt for professional post-editing when:

The text is sensitive or confidential
When the source text is intended to leave an impact on the readership (commitment, affection, awarness, image perception, loyalty, alteration of behavior or attitude, purchasing…). Such is the case for marketing translations.
Due to a faulty comprehension or interpretation, the text may have rather negative consequences (textes requiring utmost accuracy for example).
The text requires cultural or contextual adaptation
When you must work with “unconventional” languages.

In the aforementioned cases, opt for 100% human translation, ISO17100 translation or transcreation.

Authorization, Regulations and Confidentiality Policy applicable to Post-Editing

As a language service provider with ISO17100 certification for translation services, Atenao is authorized to offer Post-Editing services without the ISO18587 certification specific to Post-Editing.

Atenao’s Confidentiality Policy is applicable to all client texts and documents in 100% Human Translation, ISO 17100 Translation and Transcreation. The use of third party Neural Machine Translation (NMT) engines engages the Confidentiality Policy specific to the selected Neural Machine Translation engine. Atenao solely uses professional versions of Neural Machine Translation engines for their stricter regard to confidentiality. All texts and documents transmitted to Atenao for the purpose of Post-Editing are not permanently stocked by the used Neural Machine Translation engines. The said texts and documents are temporarily stored, strictly for the duration required to perform the translation. Once executed, all texts subject to Machine Translation processing are deleted. It is important to note that your texts are not used to improve the service quality of the used Neural Machine Translation engines.

Why entrust your translation project to a multilingual services agency?

Mastering neural translation engines: The Agency manages large projects over the majority of sectors and terminologies. It is accustomed to feeding data into neural translation engines and knows how to determine the most convenient one as per the project’s conditions and constraints (language combination, thematic…)
High performance tools: Expertise in both neural translation engines and post-editing tools. CAT tools (Computer Assisted Translation tools) post editing dedicated features such as translation memory, terminology management, saving and reusing translation segments to accelerate the process.
Skilful professional translation: Not all translators are alike and your multilingual services agency knows the majority of skilled translators on a worldwide scale and the most likely one to optimize your project quality.
Deadline optimization: The agency can dissect your project into as many necessary parts and professional translators as your project deadline requires.
The price: Resorting to an agency doesn’t come cheap, but thankfully, the agency is well aware of the real cost of post-editing services. Add to that the additional services and expertise each step of the way and it comes down more cost-effective to work with an agency rather than a freelance translator.