overview .  download .  systems .  research .  links
   what is RavenClaw . what is Olympus . recent developments .

RavenClaw is a dialog management framework for task-oriented spoken dialog systems. It was developed at CMU, as a successor to the AGENDA dialog management architecture previously used in the CMU Communicator project.

The RavenClaw dialog management framework enforces a clear separation between the domain-specific and the domain-independent aspects of the dialog control logic. System developers focus exclusively on specifying the domain-specific aspects. In parallel, the RavenClaw Dialog Engine automatically ensures a rich set of conversational behaviors, such as error-handling, timing and turn-taking, in addition to support for universal dialog mechanisms such as repeat, help, suspend, where are we, what can I say, start over, etc.

RavenClaw is a framework for dialog management. In order to build a full spoken language interface around a RavenClaw dialog manager, other components such as a speech recognition, language understanding, language generation, speech synthesis, etc. are required. To date, RavenClaw has been used within the context of the Olympus dialog system architecture, which provides all these components, as well as the means for connecting them. Although most systems we have built use RavenClaw coupled with the Olympus components, it is possible to connect RavenClaw-based dialog managers to other system architectures (for instance, in the IPA system, we have successfully connected it with the Nuance recognizer and Gemini language understanding and generation in an OAA architecture)

RavenClaw/Olympus-based Systems

To date, a number of systems spanning multiple domains and interaction types have been developed using the RavenClaw dialog management framework. Some of these systems (e.g. RoomLine, the Let's Go! Public Bus Information System) have been successfully deployed and are accessible to the larger public. More information about these and other RavenClaw-based systems, is available here.


Both RavenClaw and the larger Olympus architecture are available for download. We have made available a number of full spoken dialog systems that were constructed using the RavenClaw/Olympus infrastructure. If you decide to download a system, and would like the latest documentation, please email dbohus@cs.cmu.edu.

Development Team

Developers: Dan Bohus Antoine Raux
Contributions from: Thomas Harris
Jahanzeb Sherwani
Alex Rudnicky
Andrew Hoskins
David Huggins-Daines
Scott Judy
Yitao Sun