We describe the Dr Inventor creativity support tool that aims to support and even enhance the creativity of active research scientists, by discovering un-noticed analogical similarities between publications. The tool combines text processing, lexical analysis and computational cognitive modeling to find comparisons with the greatest potential for a creative impact on the system users. A multi-year corpus of publications is used to drive the creativity of the system, with a central graph matching algorithm being adapted to identify the best analogy between any pair of papers. Dr Inventor has been developed for use by computer graphics researchers, with a particular focus on publications from the SIGGRAPH conference series and it uses this context in three main ways. Firstly, the pragmatic context of creativity support requires the identification of comparisons that are unlike pre-existing information. Secondly, the suggested inferences are assessed for quality within the context of a corpus of graphics publications. Finally, expert users from this discipline were asked to identify the qualities of greatest concern to them, which then guided the subsequent evaluation task.