CUSTOMER STORY

With XTM, harnessing extra value from your language assets

The new feature of XTM and SYSTRAN

KEY FIGURES

2022

Foundation of XTM

26 billion

words translated per year

227

unique languages

+22 BLEU

points increase by combining both Neural Fuzzy + Specialization

Generic vs Customized Neural Machine Translation

Machine translation usage is becoming more widespread and more strategic for enterprises. Professionals are more and more demanding in terms of quality and productivity. Therefore, the need for customization is higher than ever. Customization must be easy to use and fast to execute in order to serve all the different use cases within a company.

Throughout the years, machine translation went through major technological shifts:

Rule based MT (RBMT): customization was mandatory for MT to be usable.
Statistical MT (SMT): led to a major leap in quality with the use of massive data but there was little focus on customization at that time.
NMT: introduced a real breakthrough in terms of quality and fluency, still the demand for customization from the professional users increased a lot. Now custom NMT is becoming a standard.

Customized translations with SYSTRAN model studio

Solution

How does fuzzy matches work ?

The XTM AI-enhanced TM feature sends fuzzy matches* to the neural machine translation engine as reference material. Those fuzzy matches come from the translation memory and provide additional clues to the MT engine. Translation memories and machine translation work hand in hand to find the best translation. The quality of Machine Translation gets optimized in real time to save post-editing efforts.

Validated translation memories are key here and a very valuable asset for companies who translate contents. With the AI-enhanced TM feature, validated translations from the translation memories help the MT engine produce a high-quality translation based on validated content stored in the TM.

*A fuzzy match consists in a similar sentence with a reasonably high percentage of match. They are not exactly the same, but quite similar.

About XTM

XTM International was formed in 2002 when they spotted a need for high-quality, enterprise-class translation technologies that would keep pace with the fast-evolving demands of globalization. Today, the ability to update and localize large volumes of content – quickly and easily – is business-critical for more organizations than ever before. Their innovative Enterprise Translation Management System with an integrated CAT tool helps achieve this. XTM Cloud TMS is based on open standards and is securely accessed via a browser.

RESULT

Improvement by combining both Neural Fuzzy + Specialization > 18 + 4 = +22 BLEU points increase expected

Ronald Egle

Content Systems Administrator Ariel Corporation

By using XTM and SYSTRAN, we’ve been able to find extra value in our language assets. We had a phenomenal setup with our content management system feeding automatically into and out of XTM. Adding SYSTRAN MT to the process made it even better.