WP 5 Communication

Nebula Engine (Radon)
Since the company's inception in 1998, Radon Labs has developed its own technology. The core foundation of our game development is the NEBULA engine.
Nebula is an Open Source Engine which means it is absolutely free of charge. Anyone can download the source code and use it to develop their own projects. The Nebula Toolkit, a commercial plugin for Alias Wavefront Maya, integrates the Nebula Engine into one of the industry's standard tools, is also available from Radon Labs.
Core features of the engine are its expandability which allows modules (e.g. physics, pathfinding, audio, etc.) to be easily implemented to fit the need of the project and its versatility to be used for any kind of genre. Most commercial engines are limited to one game genre. Nebula can be used to develop any kind of 3D application not just games.
The Nebula Engine is in development since 1995 by Radon Labs' Chief Technical Officer, André "Floh" Weissflog, who started programming on a KC85 in the former GDR. Nebula 2, the current version, is used in many applications worldwide: games, 3d visualisation both commercial and scientific. Nebula 3 is currently under development, first parts are to be released in 2008.

Nebula Client (Radon)
Radon Labs' contribution to RASCALLI consists of the so-called "Nebula Client" (NC), a stand-alone Windows application which serves two purposes:
1) Provide an interface to the actual user with the help of a 3D animated character who offers a much more "human" interface
2) the NC serves as an interface to the various components of RASCALLI: the Platform, developed by OFAI and the MaryServer by DFKI for speech synthesis
The NC features a scene rendered in 3D real-time which shows the Agent or Rascallo (a male Agent) sitting in a chair. Via a simple Graphical User Interface the user can input questions and comments as texts which are then sent to the OFAI Platform. This request is processed and a file in Behaviour Markup Language (BML) is sent back to the Client with instructions which animations to play in which order, e.g. point gestures or a gaze. This takes care of the animation of the Agent. Another request is sent to the MaryServer which answers with a Wave file consisting of synthesized speech and timing information telling the Client how long to play Wave.
The Agents answer consists of text output accompanied by speech and animation to give the user a lively impression as if he is engaged in a conversation. The Agent may either give an answer which contains the information the user requested or refer to a weblink (URL) which contains further information. This could be any source on the WWW to contain such information like Wikipedia, IMDB or NNDB. By clicking on the link the user is taken to the respective website.
Two buttons in the user interface allow the user to give the Agent positive or negative feedback to praise and reprimand him respectively. This is used to train the Agent and make him remember which of his answers were helpful to the user. A profile of the Agent is stored on the OFAI platform server so the Agents information is stored after the user logs out.

Multi-Modal Generation Component (OFAI)
The multi-modal generation component generates the speech, gestures and facial expressions Rascalli use to communicate with the user. The output is encoded in an XML format that includes SSML (speech synthesis markup language) and BML (behaviour markup language) markup, making it interpretable by the MARY speech synthesis and the Rascalli User Interface (Nebula Client) developed by Radon.

The generation component implements a template-based approach by encoding vocabulary, phrases, gestures etc. - which can be combined with the output of the Rascalli tools and context data - in the form of Velocity templates. The use of Velocity, which is an industry-strength templating engine written in Java, allows to design and refine templates seperately from the application code.

In order to enhance the robustness of the system when confronted with utterances that are not domain-specific, we integrate chat bot software as a fallback strategy for generating replies. The chat bot component is accessed as a tool in terms of the Rascalli platform, and its output can be embedded inside applicable templates just like any other tool output. Currently, we have integrated the Program D AliceBot implementation into the ChatBot tool.

Jabber Protocol Interface (OFAI)
As a simple user interface mechanism, OFAI is developing a communication interface for the Rascalli system that allows Rascalli agents to connect to the Jabber (an XML-based open source instant messaging protocol) network. Various Jabber clients provide an easy-to use text-based interface by which the user can contact her Rascallo (the Rascallo can be added to the user's contact list just like other - human - contacts), and can be notified or contacted by her Rascallo in an unobstrusive manner.

Generalization of QUIZ Questions (DFKI)
Quiz application aims to answers the questions based on our database content, e.g. the social relations, the cause of death, the awards. The generalization of QUIZ questions is designed at DFKI in a (semi)-automatic way.  Information extraction components have learned a lot of linguistic patterns from a large amount data, referring to different relations defined in the RASCALLI project. We utilize these linguistic patterns and generate different questions and their paragraphes as our QUIZ questions. The question foci are generated on top of the gossip database design.

Back to workpackages

RASCALLI is supported by the European Commission Cognitive Systems Programme (IST-27596-2004).


RASCALLI develops a new type of personalized cognitive agents, the Rascalli, that live and learn on the Internet.

Rascalli combine Internet-based perception, action, reasoning, learning, and communication.

Rascalli come into existence by creation through the user. The users not only create their Rascalli but also train them to fulfil specific tasks, such as be experts in a quiz game or assist the user in a music portal.