Operator: recognize

Operator Overview:

Recognize operators, patterned after Mark Wiesmeyer's thesis work (An Operator-Based Model of Human Covert Visual Attention) but extended in certain ways, process an attended input to integrate separate features of the same visual object and to determine whether the object matches any of a number of recognition targets. Recognition is one of a number of comprehension operators, which Allen describes in the following way in Unified Theories of Cognition: "Comprehend... involves both recognition and labeling or re-representing, as required." (more)

Operator Proposal:

In this domain, recognition proposals are learned. The search-for-step and scan-step problem-spaces can both result in recognition operator proposals. These preferences are either returned to the top by the production generic*ao*subspace-ACTI, which appears in generic.soar6, or are produced by explicit operators such as check-criterion.

Operator Application:

The recognize operator, when applied to visual objects, results in the addition to the top-state of a new ^visual-objects value. Each visual object consists of an amalgamated list of all of the features of that object, plus possibly some additional information.

Recognize operators applied to visual objects may also be used to test for specific recognition targets, and will then add the appropriate ^criterion-satisfied value to the task.

Operator Reconsider:

The operator is reconsidered when all of the co-located features are on a new visual-object.