memoQ 9.0 - Take advantage of the latest Machine Translation technologies

  • 13 June 2019
  • by The memoQ Team
memoQ 9.0 - Take advantage of the latest Machine Translation technologies

memoQ is evolving along with the ever-changing translation industry landscape. In memoQ 9.0, we have a special focus on neural machine translation. We now support Microsoft’s custom NMT offering, introduce a new Japanese MT plugin, and offer updates to several MT plugins already in memoQ


Machine translation (MT) has already changed the way translators work. Over and over again we see the value of machine translation in speeding up the translation process. Many of the features already in memoQ can be used in workflows that include MT. In addition, a steady stream of additions and improvements means you can rely on memoQ to help you take the best advantage of the opportunities and challenges that come with MT technology.

Until recently, most machine translation systems were built on either rules-based or statistical models. Users had some success in specific, limited applications but, in general, the translation results gave rise to widely held opinions that MT wasn’t good enough to be a practical solution for translating more in less time with suitable quality. Then came Neural Machine Translation (NMT), a model based on machine learning. NMT learns how to translate. The more translated and human-verified content you supply, the smarter the model gets. Well-trained NMT in the relevant subject matter domain inevitably provides translations that become increasingly more accurate and need less post-editing. NMT has proven to be an effective solution in many more situations—all of which rely on human translators in essential roles.

The fast pace of advancements in NMT, along with the proliferation of commercial and customized options, and connectors between various tools and systems, means that even smaller companies and individuals can use MT to good advantage. Indeed, some see it as a requirement for the future.

Overall, MT creates more jobs for translators and more translation work (for example, high-volume content that was simply too expensive and time-consuming to translate in a purely human process, even with CAT tools). For some, using MT frees up time to do more human translation that is not appropriate for MT. For others, MT brings new kinds of linguistic and technical work such as post-editing and providing feedback for training MT engines.

Per Balázs Kis, Co-Founder and Chairman of the Board at memoQ, “...because the translations are well-formed, and for the most part, grammatical...this also makes it more difficult to spot missing or altered parts.” As a result, translators are needed to ensure the output is high-quality content, and this can be trickier than it seems. The more fluent the MT is, the more likely we are to trust that it is already correct, and the harder it is to find errors and perfect.

memoQ has remedies for this challenge of NMT. If you have a solid term base with key terminology (for example, in memoQ you can specify that the Hungarian "szerződés" is to be translated as "contract" and not as "agreement") memoQ helps eliminate unwanted options. Plus, memoQ QA can warn you if a term is missing or incorrectly translated, as well as running several other automated quality checks (for example, non-translatable items, formats and values in numbers and dates, consistency, and punctuation). Nonetheless, professionals are still needed to spot other types of issues (for example, misunderstood source text or improper register and style).

What’s new in memoQ 9.0 for machine translation? There’s a lot, with a special focus on NMT.

There’s a brand new plugin in memoQ 9.0, for Mirai Translation, a machine translation service from this innovative Japanese company.
 
memoQ 9.0 now supports the latest version of Microsoft’s machine translation plugin, Translator Text API v3.0. This version is based solely on neural machine translation and supports more than three dozen languages. One of the features, Custom Translator, is the feature that allows you to build custom NMT models and then use memoQ's new Microsoft MT plug-in to connect to those models.
 
You train Microsoft’s MT with your own resources, such as translation memories and term bases. Custom Translator guarantees full privacy. Your training data is shared only with those you choose, and sharing can be revoked by you at any time.
 
Do you use DeepL, Kantan, Crossland or Systran? All their plugins have been updated in memoQ 9.0 so you can enjoy the benefits of their latest releases.
 
Finally, memoQ 9.0 comes with a painstakingly updated Machine Translation SDK (software development kit). We made requirements more transparent to make it easier for third-party developers to create MT plugins for memoQ. This update also has new content and a more comprehensive, better commented dummy plugin.
Are you using MT? Save time, increase your productivity and deliver the highest quality of work with these new developments for MT in memoQ 9.0!