Why Human Post-Editing Is Essential for Accurate Machine Translations

Have machine translations reached full parity with humans? Not quite—which makes a human touch essential. The accuracy and effectiveness of machine translations will depend on numerous factors, such as language combination, content domain, complexity and technical level, whether the content is prosaic or creative, how well-written the source content is, input format, and more.

Due to these infinite variations, machine translations are not able to detect every linguistic nuance and produce a correct translation in every instance. In most cases, to guarantee a highly reliable translation output when utilizing machine translation, human post-editing is indispensable. Moreover, if you are using a Custom Neural Machine Translation Engine (CMTE), feedback from a human linguist is necessary for retraining purposes to improve performance.

At Trusted Translations, we have expert linguists trained specifically to post-edit machine-generated translations. Our skilled post-editors rely on their experience with machine or automated translation output not only to provide corrections, but also to improve future output.

With the right technology tools in place, the post-editing process will improve the output quality of a current translation project, allow you to store that quality data in your translation memory (TM), and also improve the efficiency and accuracy of the machine translation engine for future projects.

Translating and Post-Editing Processes

The processes of translating and post-editing machine-generated content are quite different; each one calls for distinct skills and training. Not all translators are comfortable doing post-editing; even if they sign on for post-editing, they often require training.

Quite recently, academia has started developing curricula to address the growing demand for human post-editors. This training begins with the mental process used when reviewing bilingual content. When translating content from scratch, faced with a “target” (the would-be translation) that is a blank canvas, translators normally:

• Read the source content carefully.
• Formulate the translation internally (i.e., in their brain).
• And finally, type their version in the target language (translation) in a software program or interface.

Depending on the content and project, the translator may edit the text immediately, conduct additional research, or return to it later when the entire text has been translated.

In post-editing machine translated content, however, the process is very different, and it is generally performed with the use of a computer-assisted translation tool (CATT):

• First, the post-editor (linguist) reads a sentence, or segment, of the MT output (that is, the process begins at the target).
• The MT output is compared to the source text.
• The post-editor makes an in-the-moment judgment of the MT output’s quality, based on linguistic knowledge and explicit instructions.
• If the MT output is acceptable, they continue working to improve the text.
• If the MT output is of poor quality, they delete whatever the MT is offering and re-translate it from scratch (as in the translation process listed above).

As part of a machine-translation post-editing process, the project will often include a translation memory (TM) that contains translation segments from previous translation jobs of similar/related material, usually from clients or other similar projects. These previously translated segments contained in the TM, may be either a fuzzy (an approximation) or a full match of the current source segment. Taking a look at what the TM offers, the post-editor can either use these previous translation segments in their entirety or make the necessary adjustments.