Hyperknowledge allows to build a federated ecosystem of knowledge work tools. It is based on an event sourcing model of generalized concept graphs, which can model structured data or documents. High-level events are interpreted into micro-events by distributed components, allowing for ad-hoc event model extensibility; components can enrich data reactively; inter-process subscription allows reactive transclusion.
We will present work in progress on the hyperknowledge protocol.
We are developing infrastructure for a federated document and concept graph model, HyperKnowledge.
We want to establish a Web of Knowledge. This allows to see how any idea is connected with other ideas in debates, through evidence, citations, influence, consequences, etc.
The long-term goal is to create social documents, with advanced ViewSpec capabilities. Text implicitly references topics. These implicit references should be replaced by explicit pointers to data structures that represent those topics. This would allow the document to come alive in the following way:
1. When the topic referred to is a data structure, we can manipulate the data with advanced, data-specific ViewSpecs, beyond what can be achieved with ViewSpecs on uninterpreted text.
2. When the topic is embedded in a concept graph, we can show the neighbourhood of related concepts elsewhere in the document or in any document that uses that concept. Advanced queries on combinations of concepts can go much further than textual search in giving a social neighbourhood to the document. By itself, visual representations of the larger concept graph would add a layer of interpretation to the document.
3. When the topic is a transclusion of another document (or fragment thereof), we can have document composition and reuse. We can also control the updates to the transcluded fragments, or clamp them to a timestamped snapshot.
4. The federation protocol will provide topic subscription, which will allow bidirectional links between heterogeneous tools.
But federated topic-aware documents depend on a federated concept graph, open enough so that different information appliances can enrich it independently, yet with subdomains which are curated by a community. At this point, we have built an event-based grammar of concept graph operations, which allows controlled topic transclusion, local topic variants (akin to git forks at a very local level), with on-demand updates from the transcluded topic (akin to git merges.) We will show how the event model allows any information appliance to introduce new operations. We will show an early prototype of the federated graph library, while integration with document-based applications is work in progress.
We will also show work on an earlier tool, Idealoom, which illustrates how documents and concept graphs can illuminate one another.
Relevance to Doug Engelbart’s Work:
Engelbart’s original demonstration was connecting multiple documents in a coherent system. The web has allowed cross-links between very heterogenous systems, but we lost back-links, provenance and history. We reintroduce those by grounding transclusion in an event model; we are building a concept graph model, which can be embedded in documents, to allow complex semantic connections between them; and we are focusing on distributed events to allow ecosystem of analysis tools to communicate.
Connection to other demos
Our tool expects to consume document operations as events. It is possible to consume non-event-based documents or graphs with transducers. But our first integration will emit events from one other front end tool, either Webleau, Trailmarks, or our own IdeaLoom, to demonstrate interoperability concretely.
Marc-Antoine Parent, Gyuri Lajos, Robert Cunningham