Below are the most common problems clients report in tickets concerning the Machine translation (MT) engines.

When submitting tickets, remember that XTM Cloud is only an intermediary between you and a particular MT service provider. We are not responsible for the quality of the returned text and can only influence the overall process in a relatively small way.

MT engine fails during synchronization

Open

Solution: Perform the following initial checks in your MT connection details (administrative privileges required!) before submitting a ticket:

• Check that no additional white spaces are present.

• Check if your credentials are correct. If you have another set of credentials, check them as well.

• Check if your MT engine's API panel permits connections to the outside world (especially Google MT).

• If nothing happens when you click on Synchronize, or there is an error, open the Dev Tools in your browser and check for error messages (see the following article for help: How to use browser's a developer tools (Dev Tools)?).

MT engine fails to process some of the segments

Solution: Perform the following initial checks before submitting a ticket:

• Check that you have not hit any of the limits in your MT API. Both the character and time limits are important.

• Check if the MT engine you are using can match the remaining segments with XTM Workbench mode (we recommend that you select the After analysis and in XTM Workbench option in your MT engine connection details).

Open

• Check if the affected segments have special characters or similar, which can influence the MT output.

• The MT engine may be configured to ignore certain segments based on project settings:

  • Review the project's general information page to check if the MT engine is set to ignore specific types of TM matches.

  • Adjust settings as necessary to allow for broader translation coverage.

MT engine fails to process all segments

Solution: Perform the following initial checks before submitting a ticket:

• Check if the MT engine is activated at project level in Project editor → General info → Machine translation.

Open

• Check the project's Metrics. If there are only ICE, Leveraged and other kinds of matches then your MT is probably configured to ignore such segments and not provide a translation for them. Check your project's General info page (see the screenshot from the previous bullet point).

• Check if the MT engine has been synchronized correctly and the credentials you are using are correct (see the first issue from this section: MT engine fails during synchronization).

• Check whether any limits have been exceeded.

• Ask an MT service provider if a particular language pair is supported by their MT engine.

Note that, if an MT service provider just recently added a particular language pair to their supported languages list, it is likely that XTM Cloud has not started supporting it yet and has to upgrade the required MT functionality. Should that be the case, do not hesitate to contact the XTM International Support team.

MT engine works only in XTM Workbench mode

Solution: Perform the following initial checks before submitting a ticket:

• Check your global MT settings and verify if the After analysis or After analysis and in XTM Workbench mode is selected.

Open

• Check if other and smaller files are translated during the analysis.

Google MT has restricted access

Solution: Currently, XTM Cloud can only work with Google MT if there are no restrictions set in the Google API panel. As soon as you set a whitelist or other restriction methods, this MT engine will work with XTM Cloud.

Google MT Error: User Rate Limit Exceeded

An error message indicating User Rate Limit Exceeded appears when attempting to obtain translations.

Solution: This error typically occurs when the API usage surpasses the allowed quota. For this reason, monitor your API usage regularly and consider upgrading your plan or optimizing your requests to remain within the limits.

Subcontractor cannot use MT engine in projects

Solution: Currently, it is only possible for a subcontractor (LSP) to enable MT in a subcontracted project when the contractor sets the appropriate setting on their side.

While the UI on the LSP side might show that a particular MT is available for switching on in the project General info section, this can be a false positive. For this reason, before proceeding, the LSP should ask their Contractor if the MT engine is indeed switched on.

DeepL MT is not working

Solution: If solutions from the issues above are not applicable, there is another possible approach.

DeepL comes in two versions: V1 (for CAT tools) and V2 (for developers). XTM supports only V1. In your DeepL panel, check if the version of the MT engine there is also correct.

MT blockage

Sometimes, an MT engine might not being able to implement its mechanisms in XTM Workbench. You will then see this system message:

Open

As seen in the example above, the issue occurred with Google MT.

Solution: To resolve this, a Google API configuration must be implemented, to prevent unauthorized use and data theft. You need to configure your API key on the Google and Amazon side and there is an option for that:

Open

You would need to use the "None" or "HTTP referrers (web sites)" options, adding the https://www.xtm-cloud.com.

URL and other XTM Cloud instances required for the client’s account.

MT matches populate target randomly

Sometimes, certain segments are populated with an MT match straight away while others are not, despite this MT system being specifically configured to only produce a suggested translation and not insert a translation automatically.

Solution: One thing to remember is that running simultaneous projects contributes to ICE or Leveraged matches appearing in the segment. However, the same match does not exist at the time the MT engine is checking for matching segments when you only work on a single project.

MT is a collaborative mechanism: running multiple projects at the same time leads to the generation and display of new matches in the segment.

The "M" status instead of "MT" in the XTM Workbench segment

Sometimes, you might find it hard to understand the M status instead of MT in the XTM Workbench segment when you used machine translation, and, in fact, this is the only match in this segment.

Open

Solution: This is expected behavior if the match has not yet been inserted in the target segment and confirmed. The M status means that the segment in question has matches, but none of them have been used in the target segment. The MT status will only be visible if an MT match is used and confirmed in the target segment.

If you would like to have your MT matches inserted into target segments by default, ensure that [Populate target with match: Always, fixing inline differences] is enabled during project creation:

Open

Client API key is not working for Microsoft MT

Open

Solution: The client's API key might be associated with a particular region, and they must specify their own region. To do so, first contact the XTM International Support team and ask them to implement an option for users to enter a custom 'issueTokenUrl'.

After the Support team changes the configuration in the back-end, the MT connection details form should contain these two additional fields: Translation endpoint URL and Authorization endpoint URL.

In the case of Microsoft MT, a URL of this kind, relating to a specific region, must be specified in the Authorization endpoint URL field.

Keep in mind that the URL practically stays the same, and what is actually changed is the part indicating a particular region. For instance, in the URL: https://westeurope.api.cognitive.microsoft.com/sts/v1.0/issueToken the part westeurope is a variable value.

Open

Lack of MT matches – incorrect credentials for CustomMT

Sometimes, you might come across a problem where your XTM Cloud project is missing MT matches although you CustomMT is switched on and enabled in a project in the XTM Cloud UI.

Open

Solution: The root cause of the problem usually lies on the configuration side. Such issues might arise from an incorrect URL configuration for the MT provider, as indicated by the following log entry:

In this particular case, there are two occurrences of /xtm/custom-mt in the URL, which is causing the API to return a 404 error.

To resolve this issue, take the following steps:

1. In Cloud, navigate to Configuration → Settings → Translation → Machine Translation (select a relevant CustomMT).

Open

2. Update the Job Seeker MT base URL to: https://connectors.inten.to/xtm.

3. Set the Translation endpoint to: /custom-mt/xtm/custom-mt/ai/text/translate?smart_routing=intento_gen_ai_routing_indeed_job_seeker.

4. After making these adjustments, enable the MT once again and reanalyze the affected project. You should then see the MT matches appearing correctly.

Missing MT wordcount in the project’s Metrics after translation

Sometimes you might encounter what at first glance appears to be an issue, namely missing MT wordcount in the project’s Metrics, in the XTM Cloud UI, in a project which was pre-translated with machine engine.

Solution: The issue stems from a combination of two settings:

1. Having the same text in both the source and target segment after machine translation.

2. Having matches auto-populated in XTM Workbench.

MT matches are considered the lowest in match priority, just above Fuzzy Repetitions (for reference, see this article: Translation memory (TM) matching and its expected behavior | Priority of TM matching). Therefore, when an MT segment has the same text in both the source and target, and a new match (not present at the time the project was created) is found, it can replace the MT match. If this new TM match is configured to auto-populate segments, the MT match is replaced as the "main match," and the project’s metrics are updated accordingly.

Here is a simple scenario illustrating this issue:

1. Create two projects (Project 1 & Project 2) with MT. It can be the same file.

2. Ensure that at least one segment has the same source and target segment after MT translation.

3. Check the Metrics for Project 1.

4. Access XTM Workbench for Project 2 in edit mode.

5. Translate one of the “source = target” MT segments and finish the project, so the translation is sent to TM.

6. Access XTM Workbench for Project 1 in edit mode. The MT match that had the same source and target segment and was translated or confirmed in Project 2 will change its match type from MT to ICE (if the file was the same).

7. Check the Metrics for Project 1: the MT match will have been replaced by an ICE match by then.

MT engines are not applying for English → English language combinations

You might report an issue that the MT engine you might have implemented in the preprocessing step does not return any MT matches for the English → English language combinations.

Solution: Such mono-linguistic combinations are not supported. The following log appears on the back-end side:

If you want an English text to be populated in the target, you can do so in the XTM Workbench settings (Settings → General → Segments → Populate target with source text).

Open