The best approach to using Machine Translation (MT) is to include it within your existing translation workflow. For many large Localisation Service Providers(LSPs), MT is a key productivity enabler for almost all of the production departments. This use of MT is the result of a focus on translating everything possible through automation in order to streamline and reduce the outside translation costs. What the technology savvy LSPs do is to create untranslated TMX files for their client data and then MT these TMX files, using them later as secondary translation memories during the post-edit phase of the project.
This creates two sets of TMX files to accompany a project dispatched to an external translator. One set consists of the previous Translation Memories as received from the client; the other set is that after being put through the MT system. The combined use of existing TM files, and new TM files created by the MT process means that every single word of the project is now translated.
The next step in the process is to have the project post-edited by either an internal team of translators or outsourced to be edited by an external team. Translators post edit at the word count speed of 5,000 words per day. On the other hand, the average translator can only translate, from raw copy, 2,000 per day. So the savings in time and efficiencies are clear to measure.
Integrating MT output into the process like this means little or no change to the existing translation workflow. However, the harnessing of MT technology does have a profound impact on costs and turnaround times. This simple equation results in increased profit margins on such projects. The production centres using MT invariably record the highest profitability. Machine Translation as a solution is now seen as a "no brainer" for LSPs as a way to increase turnover and efficiencies.