User of semantic web technology based database for more than a year now, I’m involved in a research project in Digital Humanities (ReSeed) that aims to develop new digital tools, based on graph data management, in order to help deal with the complex data in the processes of the cultural heritage understanding. This understanding happens at two moments:
In our work a strong focus has to be put on provenance, quality and trust of the information in the database, altough, for the last 2, several conflicting assessment should coexist. Furthermore, the provenance of our triples has to be identified at a level below one of the source, at a smaller segmentation level. For books at the string level, for pictures or 3D point clouds or 3D meshes at the area level.
Pragmatically, in order to deal with these issues, all of our data triples are comprised in very short named graphs, identifying a unit of coherent information.
For instance, events, composed at least of a place and a date, or points of a 3D model, composed of a [x,y, z] coordinate, produce named graphs of multiple triples.
On the contrary, a single description like the name of a person, or her/his birthdate build a single triple named graph.
This way we can track the provenance, and discuss quality and trust at the level of the unit of information.
I would like to lead discussion on this tracking mechanism for managing metadata at a very small granularity level. To confront it to the usage of others and to talk about the problems associated with it, among them the need to “unwrap” our data in order to build the global data graph.