We’re here in Seattle where over a dozen companies are showcasing new technologies and services at the TAUS User Conference. One of the most interesting companies from the first batch is Memsource.
Based in Prague, 2010, offers a service that combines machine translation, translation memory and professional post-editing. Memsource Cloud is a pure SaaS offering that combines machine translation with post-edits done by human translators (they also offer an on premise version, for customers who need to host the service within their data center). What’s unique about Memsource is that, in addition to combining translation memory (a record of previously done translations) and machine translations, is their approach to measuring the amount of post-editing done to machine translations, as a way of gauging, and then predicting machine translation quality, something that is hard to do via a purely automated approach.
The system measures MT quality at the segment (sentence or phrase) level, and tracks the amount of post-editing done by professionals. This analysis tracks things such as the number of keystrokes, the “edit distance” (amount of change made to the text), and time spent on each item. In addition to enabling customers to assess MT quality more objectively, it also enables them to reward post-editors based on their effort, with an overall 5-10% cost reduction in projects while improving overall quality.