Note: Your custom schedule will not be saved unless you create a new account or login to an existing account.
| Paper: | SS-1.6 |
| Session: | Multilinguality in Speech Processing |
| Time: | Tuesday, May 18, 14:25 - 14:42 |
| Presentation: |
Special Session Lecture |
| Topic: |
Special Sessions: Multilinguality in Speech Processing |
| Title: |
TOWARDS LANGUAGE PORTABILITY IN STATISTICAL SPEECH TRANSLATION |
| Authors: |
Alex Waibel; Carnegie Mellon University | | |
| | Tanja Schultz; Carnegie Mellon University | | |
| | Stephan Vogel; Carnegie Mellon University | | |
| | Christian Fügen; Karlsruhe University | | |
| | Matthias Honal; Karlsruhe University | | |
| | Muntsin Kolss; Karlsruhe University | | |
| | Jürgen Reichert; Karlsruhe University | | |
| | Sebastian Stüker; Karlsruhe University | | |
| Abstract: |
In this paper we presented an approach towards the tighter coupling of statically based speech translation that uses multiple layers of reduction and transformation by cascading several stochastic source-channel models. This approach more radically relies on learning techniques to overcome today’s limits of language and domain portable conversational speech translation systems. The disfluency cleaner for English achieved a recall of 77.2% and a precision of 90.2%. The same algorithms and models were effortless adapted to Mandarin Chinese giving 49.4% recall and 76.8% precision. The results on translation suggest that MT systems can be successfully constructed for any language pair by cascading multiple MT systems via English. Moreover, end-to-end performance can be improved, if the interlingua language is enriched with additional linguistic information that can be derived automatically and monolingually in a data-driven fashion. |
| |
| Back | |