|the project will be transferred to the VU at 1/1/2001.|
Based on the observation that multi-user virtual worlds are becoming an increasingly important phenomenon in the Internet-based Cyberspace (see for example www.blaxxun.com and www.colonycity.com), we propose to investigate the indexing and retrieval problem for multi-user virtual worlds.
Our research is meant to extend the ACOI framework for the indexing and retrieval of multimedia objects by providing special purpose feature detectors for virtual worlds, that fit within the ACOI architecture. The ACOI architecture is built around a Monet database that stores information concerning the structure and contents of multimedia objects available on the World Wide Web. The assumption underlying the indexing and retrieval model underlying the ACOI framework is that the structure and contents of multimedia objects may be expressed by a grammar that is augmented with media-specific detectors for analysing the object. These detectors act as token generators, providing the information to be stored in the Monet database.
For the construction of feature detectors for multi-user virtual worlds, we need to develop the technology to analyse world description files and VRML encodings of the geometric structure and objects contained in a world, and an ontology or knowledge representation to store and reason about the contents of the world and its relation to other worlds. Further we need to investigate how we can assist the end user in formulating and refining queries with respect to locations within a world, informational resources contained within a world, and navigational preferences, that is the wish to migrate to different worlds.
The initial approach will be to extend the musical feature detector, see MIDI, which is based on a combination of feature grammars and descriptive logic, to the contents of virtual world. In a later stage, we will combine our virtual world feature detector with the other multimedia feature detectors developed within the ACOI framework. To validate our approach, we intend to create a small virtual world ourselves, in order to establish the effectiveness of our indexing and retrieval method.
Our research will result in actual feature detectors running within the ACOI framework, a knowledge representation or ontology for collecting information about virtual worlds, and appropriate applets to formulate queries and deliver a report of the outcome of consulting the ACOI index database.