Developing Deployable Spoken Language Translation Systems given Limited Resources

Matthias Eck · Repository KITopen (Karlsruhe Institute of Technology) · 2008

Approaches are presented that support the deployment of spoken language translation systems. Newly developed methods allow low cost portability to new language pairs. Proposed translation model pruning techniques achieve a high translation performance even in low memory situations. The named entity and specialty vocabulary coverage, particularly on small and mobile devices, is targeted to an individual user by translation model personalization.

Read the paper · More papers on PaperTik