"This talk will summarize two large-scale efforts at Northwestern: (1) CogSketch, an open-domain sketch understanding system. Our goal with CogSketch is to make it a useful research instrument for cognitive scientists (including AI researchers) and as a platform for sketch-based educational software. (2) Companion cognitive systems, a cognitive architecture which makes analogical processing central. Our goal with Companions is to create the first software social organisms, a step towards human-level artificial intelligence. I'll focus on aspects of these efforts that might be particularly interesting from the perspective of potential collaborations: knowledge capture games to explore the semantics of spatial language, modeling 3D reasoning of engineers, learning by reading, and apprenticeship learning in a strategy game."
For me, the talk was highly interesting as it opened up a perspective how to analyse sketches of maps in a conceptual way. Other pictures are available here
In a personal conversation one day later, Ken took the time and discussed with me how I could use CogSketch in my research. This approach allows to compare sketches and helps to segment them into the basic conceptually entities. The algorithms and knowledge base behind CogSketch allow to compute a similarity value on the ground of spatial properties, for example, the directional relations of entities, their neighbourhood, the containment etc. CogSketch is really amazing and I will definitely give a try!
Thank you Ken for your time and your advice and please continue with your collaborators to provide such great tools like CogSketch!
