Analysis: Machine Translation Usage in Memsource Cloud

Memsource was a proud sponsor of the annual Conference of the European Association for Machine Translation (EAMT) on May 28-31, 2017 in Prague. Thanks to the mix of participants from both academia and translation industry, the conference sparked several interesting discussions, including the use of machine translation (MT) in professional translation.

As a sponsor and participant, Memsource presented a poster at the event which analyzed the use of MT in Memsource Cloud. Here’s what we found.

New Home Page Analytics Dashboards to Track Localization

A new set of dashboards featuring localization analytics for jobs, costs, savings, and more is now available on the home page in Memsource Cloud.

The new home page analytics dashboards are available for users of Team, Ultimate, Biz Start, Biz Team, and Unlimited editions. They allow users to have an overview of their jobs and processes and to receive in-depth analysis of over 400 localization statistics as they log into their Cloud account.

Data: Translation memory saves 36% of your translation budget – here’s the proof

Savings from translation memory for top-100 users in Memsource, sample size 516 million words.TM_saving (png visualisation)

For our second Translation Big Data study, we’ve investigated how owning translation memory increases productivity and saves money.

The table above applies a predefined net rate scheme to a sample of 500+ million words. It clearly shows that Memsource’s most active users increase their productivity by an average of 36% by using translation memory. This means that if you had an average cost of 10 euro cents per word, for this volume you could save €18.6 million. Not bad!

Data: Machine and Professional Human Translations Identical in 5 – 20% cases

MT leverage

Click on the chart to enlarge.

Our latest data shows to what extent generic machine translation (MT) engines can help professional translators. Already 5 – 20% of the suggestions from MT are good enough to simply use them as final translation without any changes. Up to 40% of the suggestions are okay after edits, and for 80% of segments MT can provide data to autocomplete.