The Behavior Markup Language (BML) is meant to be a general mark-up language to express multimodal behavior of Embodied Conversational Agents (ECAs). The language is to provide framework for
In both cases, the language is to describe behavior in an abstract way, without any reference to the ECA model in question or the actual engine realizing the behavior.
- defining unimodal and multimodal
gestures (involving the face, the head, the hands, the body, ...) for automatic generation or repository storage
- defining timing, synchronization and other aspects of a sequence of behaviors to be performed.
The work to define BML is a part of the SAIBA project (http://wiki.mindmakers.org/projects:saiba:main), initiated by a group of international researchers in 2005. A dedicated international group has been working on defining the "core" of BML. Case studies have been discussed, and at research groups the first realizes and extensions with specific features have been implemented.
In order to get a view of the current status of the work, visit the BML wiki at http://wiki.mindmakers.org/projects:BML:main.