How can I make sure Machine Translation processes are working properly? Can I get an alert if issues occur?

Question

Sometimes the text my users get from MT Hive is not entirely translated. The resulting MT translation contains paragraphs in the original language. Why does this happen, how can I solve this so any document is entirely translated?

Answer

Usually, when the system detects an error with the used MT profile, an event is logged, and administrator users are made aware of the incident. The information in these errors is directly taken over from the error the Machine Translator provider generates. Some screens show the related information to the user in form of a label or an alert icon, and in most cases, the incident is also reported to administrator users via email notification.

Event types reported

The events for which an email notification is triggered are listed below:

Notify MT issues: Translation failures: when the machine translation system failed to translate fully or partially
Notify MT issues: Unsupported languages: when the chosen MT configuration does not support the language combination

Find more information about these notification options on the related page (Settings > Emails and notifications)

Remember

To simplify the processing of errors and the distribution of email notifications, information related to the same event is centralized.

This means in the following cases a single notification is issued:

for each file during the mark online operation, containing all issues detected in that file.
for each order during the submission process, containing all issues detected in all files attached to the order upon submission.
for a batch MT operation ran on the Jobs screen of a project, containing the issues detected for each document processed. Several MT errors can be reported within the same notification.

Each error log contains all relevant information for administrators, including:

The summary part has the total number of issues in this document and a link to the project
The issues grouped by source/target languages
The type of issues detected (see types described above), with these additional details: Document name, document ID, Document URL, and total issues for each event type.

This information is used to work out the problem with the Machine Translation provider while keeping the Translation Team aligned about how to handle the incidents.