What aspects of artificial companions can I explore in the Rascalli demonstrators?

Via the 3D Client you can talk to your Rascalla/o. Talking here actually means that you can type in questions that will be answered by your Rascalla/o making use of different processing strategies. In addition, you may press praise and scolding buttons to tell your Rascalla/o that you were satisfied with or disappointed by the given answer. While you may only type in your text and press the praise or scolding buttons, the Rascalla/o answers to your input making use of synthesized speech, facial expressions and gestures, and presents information in a combination of spoken dialogue and media embedded in the 3D interface. Through the Music Explorer (MEX, and Visual Browser (VB) you can inform your Rascalla/o about your interests in music and musical artists. Based on the music you listen to and the artists you look up in MEX and VB, your Rascalla/o will acquire a profile of your preferences and interests which changes over time, depending on the user activities in MEX and VB and whether you select the more human-like forgetting mode or stick to remembering everything as it is typical for computers. You can monitor the interest profiles created within your Rascalla/o via the Web Interface Through the Web Interface, you may also inform your Rascalla/o about Web sites, typically RSS-Feeds, which are of interest for you. Knowing your preferences about musical artists from the interest profiles, the Rascalla/o will monitor the RSS-Feed and alert you when information about your favourite artists is available. Moreover your Rascalla/o is aware of other Rascalli, if they are made public by their users, and may exchange information with them. If your Rascalla/o is of type Adaptive Music Companion (AMC), it is able to learn from your positive and negative feedback, and accordingly adapt its reactions to your questions. The AMC is also equipped with a short-term memory to remember which actions it has already performed to answer your current question. Thus it is able to try its full action repertoire, and if you are constantly unsatisfied with your Rascalla/o's answers, it will tell you that it cannot do better at the moment, and that you should better ask another question. The AMC also memorizes the history of the current dialogue, and therefore realizes when you ask the same question over and over again. At some point, it will tell you to stop behaving like this. All this enables the AMC to show, at some point, awareness of what is going on in the interactions and accordingly react to it.

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

 
 
Internal links: