Winston (1975) described a Blocks World Learning program. This program operated in a simple blocks domain. The goal is to construct representation of the definition of concepts in the blocks domain.
Example : Concepts such a “house”.
- Start with input, a line drawing of a blocks world structure. It learned Concepts House, Tent, Arch as : brick (rectangular block) with a wedge (triangular block) suitably placed on top of it, tent – as 2 wedges touching side by side, or an arch – as 2 non-touching bricks supporting a third wedge or brick.
- The program for Each concept is learned through near miss. A near miss is an object that is not an instance of the concept but a very similar to such instances.
- The program uses procedures to analyze the drawing and construct a semantic net representation.
- An example of such an structural for the house is shown below:
- Node A represents entire structure, which is composed of two parts : node B, a Wedge Brick has-part A has-part B isa Supported – by C isa Wedge, and node C, a Brick.
- Links in network include supported-by, has-part, and isa.
Winston’s program followed 3 basic steps in concept formulation:
- Select one known instance of the concept. Call this the concept definition.
- Examine definitions of other known instance of the concept. Generalize the definition to include them.
- Examine descriptions of near misses. Restrict the definition to exclude these.
■ Both steps 2 and 3 of this procedure rely heavily on comparison process by which similarities and differences between structures can be detected.
■ Winston’s program can be similarly applied to learn other concepts such as “ARCH”.
Method for Determining Object Properties and Relations
(a) edges belonging to an object
1.- exterior edges belong to object if bounded by object
2.- at object-object boundaries, T verticies are used to assign edges to objects
3.- bottom collinear line segments are connected
(b) shape of faces – triangle, quadrilateral, hexagon, octagon, etc.
(c) size of an object – tiny (.5% of visual area), small, medium (1.5-5% of visual area), large, huge (35-100% of visual area)
(d) orientation – standing, lying (applicable to bricks only)
(a) left-of, right-of: A is to the left-of B is centre of area of A is left-of centre of area of B and the rightmost part of A is left-of the rightmost part of B.
(b) above, below (similar to (a))
(c) supports, supported-by: first find base line of objects: if the lower end and an interior line of an object lies at an exterior vertex, the exterior edges radiating from that vertex are candidate base lines. Example
Eliminate both candidates if one candidate is more vertical than the interior line concerned. There are various refinements, e.g., extension of base lines through matched T’s. Now if any object B is unbounded by a base line of an object A, B is a possible support of A. Examples
Eliminate “vertical slides” (using vertical edges in supposed supporting surface) and eliminate mismatched-height cases.
(d) in front-of, behind: an object bounded by a line belonging to another object and not a base line is behind that other object; there are many refinements.
(e) marries – common face common edge. Examples
X, K, and T verticies common to 2 or more objects are used to determine the “marries” property.