ABSTRACT
This chapter is dedicated to experts working in real domains where discoveries of unobservable events are desired. For example, consider an intelligence analysis in which analysts of criminal group behaviors are exploring missing links among members. The head (see the unobservable person at the top of Figure 8.1) of the criminal organization may phone subleaders managing local sections (Mr. A and Mr. B in
8.1 The Challenge in the Basis: Detecting Zero-Frequency Key Events ............ 119 8.2 Data Crystallization and Its Limits ............................................................... 121
8.2.1 KeyGraph: The Basic Tool for Visualizing Event Maps ................... 121 8.2.2 Data Crystallization: Extending KeyGraph for Analysis
of Hidden Events ................................................................................ 123 8.3 Refinement by Human-Interactive Annealing .............................................. 126 8.4 Application to Designing New Products from Patent Documents ................ 127
8.4.1 Redesign of Surface Inspection System ............................................. 127 8.4.2 Preliminary Study and Tasks ............................................................. 129 8.4.3 Improvement of Tasks ........................................................................ 130
8.5 The Application of DCHS to Patent Documents .......................................... 131 8.5.1 Experiment Condition and the Process .............................................. 131 8.5.2 Experimental Results ......................................................................... 134 8.5.3 Discussion .......................................................................................... 135
8.6 Conclusions ................................................................................................... 135 References .............................................................................................................. 135
Figure 8.1) a few times. By responding to these top-level commands, each local section keeps its own internal communication via different media from that the boss used for contacting the subleaders. Then, the subleaders may meet to achieve consensus before responding to the boss. The boss may not attend these meetings. In this way, a person who is hidden from meetings’ data may be the real boss.
