VERBMOBIL: the evolution of a complex large speech-to-speech translation system

Abstract

The ambitious goal of the project Verbmobil 1 is the development of a portable speech-to-speech translation system dealing with face-toface dialogs. It constitutes a new generation of translation sytems in which spontaneously spoken language, speaker independence and speaker adaptability are among the main features. Verbmobil brings together researchers from the fields of speech processing, computational linguistics and artificial intelligence and goes beyond the state of the art in these areas. Besides the speech and language processing issues, the specific constraints of the project represent an extreme challenge on the part of project management, software engineering and test and evaluation of the system: -size and complexity: 150 researchers from 29 organizations at different sites on three continents are involved in the software development, -integration of heterogeneous software: in order to reuse existing software, hardware and know-how, only a few restrictions were given to the partners. In the following article we describe the Verbmobil scenario, the system architecture and the system evolution.

VERBMOBIL: the use of prosody in the linguistic components of a speech understanding system

“…In parallel to this deep analysis performed by these modules, the dialogue module conducts a shallow processing, i.e., the important dialogue acts are detected in the WHG and are roughly translated. A more detailed account of the architecture can be found in [15] and [55]. Fig.…”

From within-word model search to across-word model search in large vocabulary continuous speech recognition

“…• VERBMOBIL II: the data of this corpus consist of German spontaneous speech and were collected for the VERBMOBIL project (Bub & Schwinn, 1996). The training data consists of 61·5 h of speech data taken from the VERBMOBIL CDs 1-41.…”

Enabling effective design of multimodal interfaces for speech-to-speech translation system: An empirical study of longitudinal user behaviors over time and user strategies for coping with errors

“…It should be noted that mediated interpersonal communication systems (e.g., S2S translation systems) have been used in a very limited way in this context. Early user research with S2S translation systems has been conducted under the Verbmobil project (Bub and Schwinn, 1996) and in our previous work (Shin et al, 2006). Recent advances in S2S systems, however, allow us to conduct further detailed user modeling studies, such as that considered in this work.…”