Imagine it's 2025 and there's a massive data network that agents access to run queries and produce recommendations for the citizenry.
BigTree is one way that network could be structured so individuals maintain control over the publicity and authenticity of their data while for a graph-based CMS based on the idea of iteratively combining structured data.
The root of BigTree are bit strings. Each string represents a binary encoding of some piece of information. Some information is atomic, meaning it is specific enough that it can't be changed without altering the significance:
A tree is then constructed by building branches from the root (/) to each fact:
There are multiple branches, however that could reach to each of these positions.
The goal with BigTree is to create tools for making composite documents which create alternate paths to concrete facts. As multiple paths to concrete information develop they can be used to infer relationships between data items.
The area where I would like to apply this system is as the backing for a conversational recommender on a smart phone. Information is presented in terms of screens which describe and collect information on a specific concept. The software navigation operates in terms of a concept graph which is mapped onto interface screens based on the user's configuration.
For example, the program, which is called "Tip," would start at /tip/start. By default, this would look like:
This is constructed from layered templates:
This blob is then linked back to the root through several paths:
As users interact with the system, a graph space begins to be filled in. Intelligent agents in the system could do statistical analyses of the graph structures and present users with interactions to fill in the graph to refine probabilities.