XTM Cloud 13.3: Terminology module, NMT, and filter templates
XTM Cloud 13.3: Terminology module, NMT, and filter templates illustration
Melissa Favre-Lorraine
AuthorMelissa Favre-Lorraine
Reading time 2 minutes

Last week, we announced which XTM Labs features were being released in XTM Cloud 13.3. Today, we’ll explore a range of new features, including key improvements to our Terminology module, Neural Machine Translation (NMT) integrations, filter templates, and more.

Enhanced Terminology module

Having a strong terminology program is key to creating and maintaining a strong brand image globally. We’re now introducing our latest AI technology, the NLP Multistemmer. This technology ensures that terms are better recognized by the Terminology module, as the NLP Multistemmer understands the root of a word and its variations. This is particularly challenging, and therefore important, for highly inflected languages like Arabic, French, or Polish.

For example, the word ‘play’ in English could become ‘plays’, ‘playing’ or ‘played’. In French, a multitude of variations exist for ‘jouer’: ‘joue’, ‘joues’, ‘jouons’, ‘jouez, ‘jouent’, ‘jouant’, ‘jouais’, ‘jouait’, ‘jouions’, ‘jouiez’, ‘jouaient’, ‘jouerai’, ‘jouera’, ‘jouerons’, and many more. In this case, it may be harder for the terminology module to recognize the root  (regardless of prefix or suffix) each time. This could lead to key terms being missed, and forbidden words not getting flagged, creating a risk for inaccurate localization.

As the NLP Multistemmer is able to understand the root of a word as well as its variations, it ensures that all the variations of a term will be highlighted. This technology addresses a very common TMS challenge, and we’re incredibly proud to be introducing this solution with XTM Cloud 13.3.

Neural Machine Translation Integrations

Google AutoML

XTM Cloud users can now leverage their Google AutoML glossaries. Using the right terminology is key, especially when leveraging MT. By being able to define product names, ambiguous words, and borrowed words, the NMT engine is able to produce a more accurate translation, requiring little or no rework. This ensures that the right terms are used and that content is automatically localized accurately and in line with your brand voice. Additionally, it reduces cost and turnaround times, as linguists won’t have to spend as much time correcting NMT results.

For instance, using Google AutoML glossaries, we could localize this blog post and ensure that our product names, XTM Cloud and XTM Labs, never get translated according to the glossary rules.

SYSTRAN AI-enhanced TM

(75%–84% match) fuzzy matches to provide full matches. The functionality has now been extended to all fuzzy match categories, providing an even more powerful solution.

AI-enhanced TM is a unique functionality combining AI, SYSTRAN machine translation, and your translation memory (TM). The MT engine is able to access previously localized content in your TM, use those as partial matches, and then fill-in the “gaps” with SYSTRAN’s NMT engine. That way, you get full matches created automatically that are also aligned with your brand and consistent with existing content. Here’s how it works:

Filter Templates

Font mapping

You can now use font mapping when creating filter templates to define how fonts used in the source file need to be replaced in the target file for specific languages. This can be very useful if you are translating a project into multiple languages that use different character sets, such as Japanese, Russian, or Arabic, for instance. Using the right fonts for each set will ensure that the characters are displayed properly when created in the target languages. This font mapping is independent of the font displayed in Workbench: the font replacement occurs when the target document is generated.

Segmentation rules

Segmentation rules can now be used for filter templates. These rules enable you to break your source text into smaller units for localization, making them easier to translate. These units are arranged by selecting specific segmentation rules, such as the end of a segment being determined, for example, by:

  • A full stop
  • A punctuation mark
  • A paragraph break

This can be useful for source files that are string-based, like in software development where several sentences, phrases, or lists are bundled together. These can end up being quite long and could be recognized as a single segment.

Plus, having longer segments that include several sentences could also impact your TM, as it could decrease your number of perfect matches. Having your segments neatly defined and segmented automatically using filter templates will make the localization process much easier and more efficient, as well as keeping yourTM as accurate as possible.

Advanced anonymization options in email notifications

XTM Cloud Administrators can now define advanced anonymization options to ensure that private information, such as company names or proper names, does not get displayed in automatic email notifications. As these get sent to project managers and LSPs, this ensures that data is protected at all times, even when it is sent to external agencies.

REST API Features

With each release of XTM Cloud, new REST API features are added. Here’s a list of new actions you can do in 13.3:

  • Obtain metrics at file bundle level via REST API.
  • Open Workbench in read-only mode via REST API.
  • Define a workflow due date through new REST API methods.
  • Download target files to also include no-content and non-analyzable files at job level.
  • Update file properties (tags and metadata).
  • Get more visibility into workflow changes as a callback is now sent whenever an active step is deleted.
  • Decide if the workflow statuses should be retained when cloning a project.
  • Get information on whether project files are joined.
  • New task and workflow step management endpoints for Linguist users.


Want to find out more about XTM Cloud 13.3?

We’re here to help. Enter your name and email address below and we will contact you soon.