Conceptual Analysis of Software Structure

Speaker: Richard Cole

When: 10:00, Thursday, 20th November, 2003

Venue: 78-420

Abstract:  Software systems are often highly structured, consisting of
  artifacts: types, methods, variables, and packages; and
  relationships between these artifacts. Domain and meta models, and
  software design documentation provide additional artifacts such as
  roles, associations, use cases, and paragraphs of text.

  This talk describes a tool employing formal concept analysis and a
  knowledge base (data from software analysers + rules) to
  understand software structure. The tool is demonstrated via a case
  study: we consider the structure of the Java container class
  libraries and attempt to identify a design rational behind the
  resulting class hierarchy.

  The motivating assumption underlying this work is that software
  designs evolve over time and that either due (i) to poorly
  understood or poorly enunciated design goals, or (ii) due to
  physical or algorithmic constraints, irregularities in the
  form of broken symmetries arise in software systems. Even if the
  resulting systems are not going to be repaired, I would argue that
  an understanding of system regularities and the deviations from that
  regular structure makes the system structure easier to internalize
  (i.e. remember) and so activate in human thinking.
 

Hospitality: Roger Duke

Contact: Prof Paul Bailes (SSERG seminar co-ordinator) (p.bailes@epsa.uq.edu.au)

SSE seminar web page: http://www.itee.uq.edu.au/~sse/Seminars.html