• Partner:

  • DFKI

  • TU Berlin


Video Coming Soon
The core idea of SpeechEval is to carry out automatic end-to-end tests of spoken dialog systems (SDS) for their evaluation or the testing of specific sub-components. These kind of tests take place with a user simulation on the speech level and can be realized via telephone line (ISDN). The simulation itself is based on a large human-machine corpus and aims to replicate general user behavior, i.e., how people interact with spoken dialog systems.

General research challenges:

  • Automatic tests of voice user interfaces
  • Extraction of general user behavior from a dialog corpus
  • Fast implementation of user simulations for dialog systems in any domain