Corin Gurr

Throughout the history of engineering, diagrams have been used to model and reason about systems. In engineering computer systems, the relative novelty of concepts to be modeled has given rise to a plethora of diagrammatic languages, often based upon simple “graphs” consisting of nodes with edges linking them. Graphs have the advantage of being simple and thus easy to read, yet are rather inexpressive and so are typically significantly extended and embellished. Such extensions often risk swamping the simplicity of the underlying graphs with overloaded symbology and a confusion of textual annotations (Figure 1). Addressing this issue requires a theory of diagrammatic languages that explains how meaning can be attached to the components of a language both naturally (by exploiting intrinsic graphical properties) and intuitively (taking consideration of human cognition). I have constructed such a theory by analogy to theories of natural languages as studied in computational linguistics. This approach, dubbed “Computational Diagrammatics” by a colleague, separates and clarifies issues of diagram morphology, syntax, semantics, pragmatics and so forth (Figure 2); facilitating the design of diagrammatic languages that maximize expressiveness without sacrificing readability.

Links