Recognition system with an automated development tool

Claim 2

In a data processing system, a method of generating an object recognition system for recognizing goal objects, each comprised of one or more input entities, the method performed by the data processing system comprising: defining input entity recognition objects relative to input entities; defining goal recognition objects relative to input entities through recognition object trees, each having input entity recognition objects as leaves and a goal object at its root; and generating from plural recognition object trees a decision tree of an object recognition system for recognizing goal objects from the input entities, the decision tree being separate from the recognition object trees and comprising decision nodes testing for recognition objects and comprising goal objects, from plural recognition object trees, at leaves of the decision tree, the step of generating the decision tree comprising, beginning at the root of the decision tree, recursively selecting, for the nodes of the decision tree, recognition objects from the recognition trees beginning from the leaves of the recognition trees, and grouping recognition trees to define branches from the nodes of the decision tree based on whether the recognition trees include the selected recognition objects.

Claim 3

A method as claimed in claim 1 wherein the step of generating an object recognition system comprises generating a decision tree for a group of recognition trees, the decision tree indicating at each node a test for a recognition object from the group of recognition object trees and branching from each node based on the results of the test, the decision tree testing for recognition objects of the recognition tree leaves near its root and providing goal recognition object decisions at its leaves.

Claim 4

A method as claimed in claim 1 wherein the step of generating an object recognition system comprises generating a decision tree for each group of recognition trees having the same number of leaves.

Claim 5

A method as claimed in claim 3 wherein the step of grouping recognition trees to define branches comprises including in true branches recognition trees which do not include the recognition object of the node but which include a recognition object which is likely to be confused with the recognition object of the node.

Claim 6

A method as claimed in claim 1 wherein the step of grouping recognition trees to define branches comprises including in true branches recognition trees which do not include the recognition object of the node but which include a recognition object which is likely to be confused with the recognition object of the node.

Claim 7

A method as claimed in claim 2 wherein the step of generating an object recognition system comprises generating a decision tree for each group of recognition trees having the same number of leaves.

Claim 8

In a data processing system a method of recognizing goal objects, each comprised of one or more input entities, the method performed by the data processing system comprising: defining input entity recognition objects relative to input entities; defining goal recognition objects relative to input entities through recognition object trees, each having input entity recognition objects as leaves and a goal object at its roots; and recognizing a goal recognition object from input entities by processing a decision tree which is separate from the recognition object trees but which comprises decision nodes testing for recognition objects, from the recognition object trees, with the goal recognition objects at leaves of the decision tree, the step of processing the decision tree comprising creating a discovered item database including the input entities and recognition objects identified at the nodes of the decision tree, the discovered item database supporting subsequent tests for recognition objects in the decision tree; and wherein the decision tree indicates at each node a test for recognition object, and the step of generating an objection recognition system comprises generating a decision tree for a group of recognition trees, the decision tree indicating at each node a test for a recognition object from the group of recognition object trees and branching from each node based on the results of the test, the decision tree testing for recognition objects of the recognition tree leaves near its root and providing goal recognition object decisions at its leaves.

Claim 9

In a data processing system a method of recognizing goal objects, each comprised of one or more input entities, the method performed by the data processing system comprising: defining input entity recognition objects relative to input entities; defining goal recognition objects relative to input entities through recognition object trees, each having input entity recognition objects as leaves and a goal object at its roots; and recognizing a goal recognition object from input entities by processing a decision tree which is separate from the recognition object trees but which comprises decision nodes testing for recognition objects, from the recognition object trees, with the goal recognition objects at leaves of the decision tree, the step of processing the decision tree comprising creating a discovered item database including the input entities and recognition objects identified at the nodes of the decision tree, the discovered item database supporting subsequent tests for recognition objects in the decision tree; and wherein the decision tree indicates at each node a test for recognition object, and the step of generating an objection recognition system comprises generating a decision tree for a group of recognition trees, the decision tree indicating at each node a test for a recognition object from the group of recognition object trees and branching from each node based on the results of the test, the decision tree testing for recognition objects of the recognition tree leaves near its root and providing goal recognition object decisions at its leaves.

Claim 10

In a data processing system a method of recognizing goal objects, each comprised of one or more input entities, the method performed by the data processing system comprising: defining goal recognition objects relative to intermediate and input entity recognition objects at nodes of recognition object trees, each recognition object tree having input input entity recognition objects at its leaves and a goal recognition object at its root; defining from plural recognition object trees a decision tree of tests for recognition objects, the decision tree being separate from the recognition object trees and having goal recognition object decisions at its leaves; and from the decision tree root performing tests at decision tree nodes on input entities for recognition objects of the recognition object tree nodes; and wherein the decision tree indicates at each node a test for a recognition object, and the step of generating an object recognition system comprises generating a decision tree for a group of recognition trees, the decision tree indicating at each node a test for a recognition object from the group of recognition object trees and branching from each node based on the results of the test, the decision tree testing for recognition objects of the recognition tree leaves near its root and providing goal recognition object decisions at its leaves.

Claim 11

In a data processing system a method of recognizing goal objects, each comprised of one or more input entities, the method performed by the data processing system comprising: defining goal recognition objects relative to intermediate and input entity recognition objects at nodes of recognition object trees, each recognition object tree having input input entity recognition objects at its leaves and a goal recognition object at its root; defining from plural recognition object trees a decision tree of tests for recognition objects, the decision tree being separate from the recognition object trees and having goal recognition object decisions at its leaves; and from the decision tree root performing tests at decision tree nodes on input entities for recognition objects of the recognition object tree nodes; and wherein the decision tree indicates at each node a test for a recognition object, and the step of generating an object recognition system comprises generating a decision tree for a group of recognition trees, the decision tree indicating at each node a test for a recognition object from the group of recognition object trees and branching from each node based on the results of the test, the decision tree testing for recognition objects of the recognition tree leaves near its root and providing goal recognition object decisions at its leaves.

Claim 12

A method as claimed in claim 8 wherein the step of processing the decision tree comprises creating a discovered item database including the input entities and recognition objects identified at the nodes of the decision tree, the discovered item database supporting subsequent tests for recognition objects in the decision tree.

Claim 13

A method as claimed in claim 8 wherein a decision tree is defined for each group of recognition trees having the same number of leaves.

Claim 14

A method as claimed in claim 8 wherein a decision tree is defined for each group of goal objects comprised of a common number of input entities, the method further comprising: processing a plurality of successive input entities to recognize one or more goal objects comprised of different numbers of input entities; and where processing of successive input entities from a first input entity results in recognition of plural goal objects which share the first input entity, selecting as the recognized goal object the one of the plural goal objects having the greatest number of input entities and satisfying a maximal requirement as specified for that goal object, and repeating the process on the remaining input entities until a maximum number of input entities have been associated with goal objects.

Claim 15

In a data processing system a method of recognizing goal objects, each comprised of one or more input entities, the method performed by the data processing system comprising: defining input entity recognition objects relative to input entities; defining goal recognition objects relative to input entities through recognition object trees, each having input entity recognition objects as leaves and a goal recognition object at its root; generating from the recognition object tree, a decision tree of tests for recognition objects, the decision tree being separate from the recognition object trees and having goal recognition object decisions at its leaves; and from the decision tree root, performing tests on input entities for recognition objects at the decision tree nodes; and wherein the step of generating an object recognition system comprises generating a decision tree for a group of recognition trees, the decision tree indicating at each node a test for a recognition object from the group of recognition object trees and branching from each node based on the results of the test, the decision tree testing for recognition objects of the recognition tree leaves near its root and providing goal recognition object decisions at its leaves.

Claim 16

In a data processing system a method of recognizing goal objects, each comprised of one or more input entities, the method performed by the data processing system comprising: defining input entity recognition objects relative to input entities; defining goal recognition objects relative to input entities through recognition object trees, each having input entity recognition objects as leaves and a goal recognition object at its root; generating from the recognition object tree, a decision tree of tests for recognition objects, the decision tree being separate from the recognition object trees and having goal recognition object decisions at its leaves; and from the decision tree root, performing tests on input entities for recognition objects at the decision tree nodes; and wherein the step of generating an object recognition system comprises generating a decision tree for a group of recognition trees, the decision tree indicating at each node a test for a recognition object from the group of recognition object trees and branching from each node based on the results of the test, the decision tree testing for recognition objects of the recognition tree leaves near its root and providing goal recognition object decisions at its leaves.

Claim 17

In a data processing system a method of recognizing goal objects, each comprised of one or more input entities, the method performed by the data processing system comprising: defining input entity recognition objects relative to input entities; defining goal recognition objects relative to input entities through recognition object trees, each having input entity recognition objects as leaves and a goal recognition object at its root; generating from the recognition object trees a decision tree of tests for recognition objects, the decision tree being separate from the recognition object trees and having goal recognition object decisions at its leaves; and from the decision tree root, performing tests on input entities for recognition objects at the decision tree nodes; and wherein the step of generating a decision tree comprises, beginning at the root of the decision tree, recursively selecting for the nodes of the decision tree recognition objects from the recognition trees beginning from the leaves of the recognition trees and grouping recognition trees to define branches from the nodes of the decision tree based on whether the recognition trees include the selected recognition objects.

Claim 18

In a data processing system a method of recognizing goal objects, each comprised of one or more input entities, the method performed by the data processing system comprising: defining input entity recognition objects relative to input entities; defining goal recognition objects relative to input entities through recognition object trees, each having input entity recognition objects as leaves and a goal recognition object at its root; generating from the recognition object trees a decision tree of tests for recognition objects, the decision tree being separate from the recognition object trees and having goal recognition object decisions at its leaves; and from the decision tree root, performing tests on input entities for recognition objects at the decision tree nodes; and wherein the step of generating a decision tree comprises, beginning at the root of the decision tree, recursively selecting for the nodes of the decision tree recognition objects from the recognition trees beginning from the leaves of the recognition trees and grouping recognition trees to define branches from the nodes of the decision tree based on whether the recognition trees include the selected recognition objects.

Claim 19

A method as claimed in claim 12 wherein the step of generating an object recognition system comprises generating a decision tree for each group of recognition trees having the same number of leaves.

Claim 20

A method as claimed in claim 13 wherein the step of grouping recognition trees to define branches comprises including in true branches recognition trees which do not include the recognition object of the node but which include a recognition object which is likely to be confused with the recognition object of the node.

Claim 21

A pattern recognition system for recognizing goal objects, each comprised of one or more input entities, the system comprising in a data processing system: a recognition object database of recognition trees, a recognition tree for each goal object, each recognition tree comprising a goal recognition object node at its root, intermediate recognition object nodes and input entity recognition object nodes at its leaves, the recognition object nodes comprising predicates for defining recognition objects relative to other recognition objects in the tree and input entities; a decision tree generator for generating decision trees separate from recognition object trees, there being a decision tree for each group of recognition trees having the same number of leaves, each decision tree indicating at each node a test for a recognition object of the corresponding group of recognition object trees and branching from each node based on the results of the test, the decision trees indicating tests for recognition objects of the recognition tree leaves near its root and providing goal recognition decision at its leaves, the generator beginning at the root of a decision tree, recursively selecting for the nodes of the decision tree recognition objects from the recognition trees beginning from the leaves of the recognition trees and grouping recognition trees to define branches from the nodes of the decision tree based on whether the recognition trees include the selected recognition objects; and a recognition engine, for recognizing an object from input entities, which selects a decision tree corresponding to the number of input entities, performs test for recognition objects of the recognition object database at the decision tree nodes beginning at the root of the decision tree until a recognition object decision is made at a decision tree leaf, and creates a discovered item database including the input entities and recognition objects identified at the nodes of the decision tree, the discovered item database supporting subsequent tests for recognition objects in the decision tree.

Claim 22

A pattern recognition system for recognizing goal objects, each comprised of one or more input entities, the system comprising in a data processing system: a recognition object database of recognition trees, a recognition tree for each goal object, each recognition tree comprising a goal recognition object node at its root, intermediate recognition object nodes and input entity recognition object nodes at its leaves, the recognition object nodes comprising predicates for defining recognition objects relative to other recognition objects in the tree and input entities; a decision tree generator for generating decision trees separate from recognition object trees, there being a decision tree for each group of recognition trees having the same number of leaves, each decision tree indicating at each node a test for a recognition object of the corresponding group of recognition object trees and branching from each node based on the results of the test, the decision trees indicating tests for recognition objects of the recognition tree leaves near its root and providing goal recognition decision at its leaves, the generator beginning at the root of a decision tree, recursively selecting for the nodes of the decision tree recognition objects from the recognition trees beginning from the leaves of the recognition trees and grouping recognition trees to define branches from the nodes of the decision tree based on whether the recognition trees include the selected recognition objects; and a recognition engine, for recognizing an object from input entities, which selects a decision tree corresponding to the number of input entities, performs test for recognition objects of the recognition object database at the decision tree nodes beginning at the root of the decision tree until a recognition object decision is made at a decision tree leaf, and creates a discovered item database including the input entities and recognition objects identified at the nodes of the decision tree, the discovered item database supporting subsequent tests for recognition objects in the decision tree.

Claim 23

A method as claimed in claim 12 where the decision tree includes true branches from fence recognition objects toward goal recognition objects which do not include the fence recognition objects.