[Date Prev] [Date Next] [Thread Prev] [Thread Next] Indexes: Main | Date | Thread | Author

Re: [ba-ohs-talk] new version of graph model paper posted

Title: Re: [ba-ohs-talk] new version of graph model paper pos
Joslyn, Cliff: (2000) "Hypergraph-Based Representations for Portable Knowledge Management Environments",
LANL Technical Report LAUR 00-5660, 2000.

Joslyn is thinking along v. similar lines to us, essentially arguing for weighted conceptual graphs as the fundamental underlying representation for a "DKS" (Distributed Knowledge System), that can be rendered in multiple ways (cf. Eugene's paper)

Conjecture 6: Each semantic hyperweb determines a unique conceptual graph. Conversely, each
conceptual graph determines a class of semantic hyperwebs equivalent to weights.

The conjecture basically states that a cg is equivalent to an unweighted semantic hyperweb.

The issues where we part company with cgs include:
 Semantic webs are weighted structures.
 The bipartate representation of a cg echoes the \tentacle" representation of a hypergraph.
The difference is that all the tentacles together comprise a single hyperedge, whereas in the
bipartate graph these have to be called out distinctly.
 Similarly, hypergraphs are closer to the typed-link hypertext model. Here now only concepts
are represented as vertices, and relations of different types are assigned edges of different
types (labels), resulting in a labeled or multigraph mathematical structure.

He also refers to work that connects with the graph tool reqts discussion here:

3.1 Front End
First, there is a need for a general, robust front end, not something highly specialized like the Java
Ontology Editor (JOE) [22], or Protege [42]. The following features are desired:

 A full drawing/visualization environment for both simpler node-arc diagrams and more com-
plex hypergraph form.
 A professional GUI: full undo/redo, a full audit trail (time stamping, authoring, versioning),
graphical and textual annotation, etc.
 Building on commercial platforms for graph editing and on supported protocols both at the
intersection of graph markup and KR.

Recent work at LANL collaboratively with the University of Liverpool has begun development
building from the Visio platform for Windows. Visio is a superb environment for working with
reasonably sized labeled graphs, and can be adapted to the tentacle representation of hypergraphs
and cgs.