OpenBEAT is a BS project for Center for Analysis and Design of Intelligent Agents (CADIA) at Reykjavik University, spring 2009.

The project is an open-source, modular and extensible framework for creating and outputting a set of behaviors, with integration to an Icelandic NLP source and open-source visual realizers of the behaviors. The project is implemented by Árni Hermann Reynisson (arnihr@gmail.com), Eiríkur Ari Pétursson (eirikurp06@ru.is) and Guðleifur Kristjánsson (gudleifur05@ru.is).

BEAT implements behavior generation based on English input text. BEAT does this using a pipeline structure, relying on rules derived from extensive research into human conversational behavior. Each module in the pipeline structure has a certain purpose. For instance, there is an module processing the input text in regard to grammar rules and another module assigning behavior to the output of the input text processing. The output is a fully constructed Behavioral Markup Language(BML). BEAT is extensible, so that new rules can be added in a fast manner and is designed to plug into larger systems that may also assign their own behavioral structure.

An animator can provide a text represent what an animated character is supposed to say. The text is first processed by a language module that describes the contents, using a range of linguistic markers. It uses several different techniques from the field of Natural Language Processing or NLP. From these markers a description of supporting behavior is added, based on results from human behavior studies. These results are stored in generators as rules. The text with behavior descriptions is then turned into speech and animation. The following chapters describe the difference, although based on the same principle, of the pipelines of BEAT and OpenBEAT: