Microscope!
Three research clusters:
Two projects of interest to this group:
Suggested improvement: networks over time
Complexity! Lack of models for theoretical primitives.
Chunking. Preliminary observations:
How do the two relate?
What is a scene?
Burke, Kenneth. A Grammar of Motives. University of California Press, 1969.
Rasheed, Z., and M. Shah. “Scene Detection in Hollywood Movies and TV Shows.” In 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings, 2:II – 343–48 vol.2, 2003.
Complications:
Proposed algorithm:
Separate actors from background through grammatical categories (object, subject, transitive verb, A blanks B)
Cluster scene particulates semantically to detect potential boundaries
Multiple passes to detect quality / movement (by analogy with video).
Ground truth, a stumbling block as always. To overcome: expert tagging, parallel copora. Any ideas?
We need better primitives to build more complex models!