Making Discourse Graphs Indexable & Discoverable

Discourse Modeling

Idea

We want to create a clear path for folks to publish discourse graphs that can be published and referenced by others:  we want to make shared Discourse Graphs indexable and discoverable

Data Shapes

We want to specify and promulgate a set of schema or data shapes -- like a ShEx schema-- for Discourse Graphs, so that existing graphs can be recognized in the wild, and so that existing data can be more easily organized into a d/graph.

rationale:  if we can automate shape creation, we can create discourse graphs at scale, and support the conversion of datasets to d/graphs at scale.

Namespaces

We want to map out a a strategy for creating discourse graph namespaces: a namespace practice that includes

• a unique mapping of an identifying string to an instance of an item
• a way of capturing & indicating timestamped snapshots of the referent

We believe this implies developing community of practice about namespace resolution: resolving the situation where graphs overlap and conflict, but there is broad agreement that they address the same topic. This could be as simple as appending further descriptors when there is a potential name collision, wikipedia style. This project might involve creating or discovering a clustering rule: a way of determining/declaring the family of related d/graphs a for any focal d/graph. Ideally, we'd also like to align on a practice  for attaching metadata to a d\graph.

Implementation

We think a good start to this process will be to

• define a templateable unit for a discourse graph
• write instructions for how to generate a stub
• and set up a wiki for d/graphs to start working on the namespace and discovery issue