Scope and deployment

For a long time, many have hoped (and feared) that machine translation (MT) would one day replace the time-consuming and expensive “human” translation activity. The disappointment (or the relief) was great when at the launch of every new MT technology, the output quality was simply not sufficient to do that. However, by shifting the focus and positioning MT in a more clearly defined context it is beginning to bring about real changes in how we look at translation.

Today’s main MT technologies are: rule-based machine translation (RBMT), statistical machine translation (SMT) and the newest trend towards combining the two to create ‘hybrid’ systems.

The main deployment models are the dissemination and assimilation application. MT in dissemination applications is used to support a more efficient publishing process. MT in this context is just another productivity tool integrated with TM in a traditional translation workflow environment. An important aspect of dissemination application of MT is post-editing. Post-editing does not seem too different from linguistic review, so why is it that so many translators are horrified by the thought alone? Why do companies struggle to define a suitable pricing model for post-editing?

MT in assimilation environments is used to provide real-time translations (without post-editing) for the purpose of “gisting” or information retrieval for internal staff or external customers. It is exactly in this context that the strength of MT is illustrated: not perfection but real-time usability.

Both the assimilation and dissemination model can be illustrated by successful business cases.