Towards the implementation and demonstration of the Delphus 3.0 platform two groups have worked on different directions, but with continuous interaction and the results were being merged day by day:
-Athens School of Fine Arts group constructed basic conceptual elements (with video and 3D technologies) and proposed arbitrary semantic hierarchies and their mapping to a simple relational database model. It was not feasible to go deeper in semantic descriptions and technologies during the seminar. However, the proposed scripting interface with the database is open to support any semantic technology in the future.
-Paris-8 group, an expert in virtual spaces and 3D technologies, proposed state-of-the-art solutions and finally undertook the hard task of developing an open platform of semantically linked objects, based on the conceptual requirements expressed through discussions between the two groups. The implementation was such that the code is clean and the platform is easily extensible towards any direction.
The implementation of such a platform was based on the following multilayer architecture:
More specifically, the role of each layer is the following:
- Layer 1: palette of primitive elements - This layer is a library of basic elements (3D shapes, materials, textures, behaviours, etc) that can be chosen to define conceptual objects.
- Layer 2: (hierarchical) object tagging and database storage – In this layer conceptual objects are linked to form semantic hierarchies, in the sense of very simple “ontologies”. This information is mapped to a relational database.
- Layer 3: rule-based synthesis and complex structures storage – In this layer, conceptual objects from the same or different hierarchies can be interlinked to form more complex structures, which can be stored in the database as new object definitions.
- Layer 4: conceptual navigation, interaction and searching – This layer provides the final setup of this virtual universe, as well as the navigation and interaction system through the interlinked conceptual objects.
During the seminar the following functionalities were implemented:
Layer 1- almost complete with a few behaviours and materials included – more behaviours and materials can be added easily
Layer 2 – a simple hierarchical tagging model has been chosen – this layer can be extended to support more sophisticated models, based on further research which could not be accomplished during the seminar
Layer 3 – This layer is almost completed, except for the storage of complex structures which could not be completed due to lack of time.
Layer 4 – This layer is complete.
As already mentioned, the system architecture is such that it can be easily extended to include more sophisticated functionalities.
The most important achievement of this meeting is that we have managed to subdue our different ideas and viewpoints and to converge to a product that will form the basis for further collaboration.
This ambitious work has already raised many important research issues and ideas which we will investigate.
Furthermore, the Delphus 3.0 platform will be the basis for the next seminar.