Before Hybrid translation was introduced in the Arcturus version, the translation workflow used to be mostly performed by human translators. Existing tools such as Wezen’s Correlations Mechanism and Translation Memories were already there to ensure translation quality and reduce delivery times, and they still will. However, a new brick in this workflow will enable an even faster translation process: Machine Translation!
Machine Translation in the translation workflow of a translation task
Before human translators start working on translation tasks in the
Analysis step of the translation tasks' workflow, translation tasks that contain segments with a translation memory match under 75% will be filled automatically with machine translation before being set to “Translation” status if machine translation has been activated on the project. No more untranslated segments when human translators open translation tasks for the first time.
Inside the Translation Studio
Using the Machine Translation widget in the Translation Studio, users can type source content in the search bar and request suggestions of translations from Machine Translation providers.
The visual indicator MT indicates that the segment was automatically filled with machine translation between the
Translation phase so that human translators are aware of it and know that they should carefully review these machine-translated segments.
As machine translation intervenes after the
Analysis phase and only fills segments with a translation memory match under 75%, all the segments with a translation memory match above 75% will be filled via the Correlations Mechanism or the Translation Memory. This ensures segments are translated using previously validated and quality translations.
Instead of leaving the remaining segments empty, they are filled using machine translation so that human translators can simply post-edit them and save a lot of time!
Machine translation quality has greatly improved thanks to recent research in Natural Language Processing algorithms and Machine Learning. Translations from these modals are getting closer and closer to what human translators would provide. However, it is important to note that it won't remove the need for human translators, but it should be considered as an assisting tool to speed up the translation process.
By default, a machine translation provider can be set for the whole project, no matter the source and target languages. However, for each pair of languages, a different machine translation provider can be selected to best match the content to translate and optimize quality.
As a manager, you can navigate through the
Settings section and configure the default local configuration and language-specific local configurations. In order for machine translation to work, please contact Wezen Helpdesk to configure the MT provider's credentials first.
Afterward, in the
Machine Translation section of each local configuration, the following parameters can be set:
|Scope||Scope to which machine translation will be applied|
|Field list||List of fields to:|
List of fields
|Daily character limit||Limit of characters which have been translated using machine translation during the current day to avoid over-consuming machine translation and monitor costs induced by using machine translation providers.||Any integer greater than 0.|