Translated.net is here in Seattle demoing a new plugin that extends popular CAT (computer aided translation) tools so that translators can work effectively as teams, and use a shared pool of translation memory and machine translation services.

The plugin, which works with popular CAT tools like SDL/Trados and MemoQ, enables translators to collaborate globally, as well as within their organization. The plugin makes machine translation work in the same way as translation memory, where possible translations are suggested to translators in their editing environment. There it is easy for translators to select the best available option. The tool pulls from many sources in the following order:

First it looks in the local translation memory for the current project (if there is a match, it will generally better than other options). Then it looks in a shared translation memory across projects within an organization. Then it looks to a global translation memory, shared across all users (with over 1 billion words captured in six months). If it has no luck with this search, it then goes to a domain specific machine translation engine, and if all else fails, goes to a generic translation engine.

The MyMemory interface aggregates machine and human edited translations from many sources, and encourages users to edit and curate them as a community.

As translators select and post-edit recommended machine translations, the system learns which engines are generating the best output for different tasks and language pairs. The amount of effort put into post-editing translations is a good mechanical way to measure quality (more time and effort spent fixing a translation = lower quality).