Adaptive recognition of implicit structures in human-organized layouts

Abstract

Card-handling using hypertext editor can be a p owerful methodology for generation of ideas or understanding of complex problems. To support such activity, recognizing implicit structure in the arrangement of cards would be useful. But, because the structures to be r ecognized a r e by nature ambiguous and highly dependent on user-specic perception, it is dicult for conventional rule-based s p atial parsing algorithm to achieve this task. We propose techniques for building spatial parser suitable for nding such ambiguous structures based on the mechanics of human perception. Moreover, our parser is adaptively customized t o r eect a p articular user's preferences through an interactive suggestion process, supported by application of a genetic algorithm.

Finding hyper-structure in space: spatial parsing in 3D

“…We have therefore chosen a global edge selection strategy that maximizes the total sum of levels corresponding to the edges selected and the types assigned. This is contrary to the local edge selection strategy chosen in (11).…”

“…As mentioned before our 3D parsing strategy is strongly influenced by the algorithm described in (11). Another approach would be to take departure in data clustering as known from data mining applications, as for example (21).…”

“…For each such axis aligned bounding box we partition all other objects in View Space into the surrounding 26´ · · µ disjoint subspaces that result from partitioning the space by the six (infinite) planes defined by the faces of the bounding box. This is a direct extension to 3D of the partitions shown in Figure 5 and used in (11).…”

“…Contrary to (11) and (3) a list in a pure 3D environment can be oriented arbitrarily, as objects can be rotated and scaled freely. As a consequence of the above the list expert works in the coordinate system (a, b, c) defined by the three vectors:…”

Perceptual grouping

“…This problem is exacerbated on large interactive surfaces [22], where arm fatigue may play a role. To help minimize the interaction required, previous research investigated techniques to infer perceptual structures from drawings [14,29,30], to analyze selection gestures for perceptual information [6], and to provide suggestions based on one's selection [10]. In contrast to these algorithms, humans are very skilled at visually identifying object groups.…”