Machine translation can be useful, though the approach many machine translation companies isn’t really translator-friendly. Many MT (machine translation) vendors see MT as a data-driven process rather than a workflow-driven process, and do not accommodate any feedback cycle. They believe in customer-specific training, then machine translation, then post-editing, but there is no link back from post-editing to training, thus the quality does not improve. If you sympathize with these companies, imagine yourself having to correct the same mistake over and over. We believe that systems that simply don’t consider user feedback miss out on many opportunities for improvement.
Kilgray has a wealth of machine translation experience. Prior to establishing the company, the founders, Balázs, Gábor and István worked for MorphoLogic, a company well-known for its English-Hungarian machine translation software MetaMorpho. Indeed, Gábor was the architect behind the parser of this rule-based system. We see a convergence between machine translation with human assistance and computer-assisted human translation. However, it’s all about the workflow.In an ideal situation, machine translation begins with training, then the trained system is used in production and there is a feedback cycle converting user suggestions into improvements to the rules, no matter whether these rules are statistical or syntactic.
We see the industry moving into this direction and we’d place our bet on one single MT company, but while that happens, we are open to integrate any machine translation product into memoQ. We have a set of documented APIs that allow the integration of an external text-parsing tool into memoQ. memoQ gives out the source segment to a third-party tool and retrieves the output of this tool. It can be commercial machine translation or it can be anything else that you come up with. It’s a plugin architecture and you can even have multiple MT vendors addressed at the same time. memoQ integrates this in the translation grid and the pre-translation process.
If you want a system that integrates with machine translation, contact us to check whether we already have an integration with the vendor of your choice. Or come up with own text manipulation engines and integrate it into memoQ yourself. The options are open. Machine translation is only available in memoQ’s enterprise edition.