Automatic classification method and apparatus
AssigneeHewlett-Packard Development Company, L.P.
InventorsGeorge H. Forman, Henri J. Suermondt
Life of this patent
Estimating the term…
Abstract
Methods and apparatus for classifying an instance (i.e., a data item or a record) automatically into one or more classes selected from a set of potential classes are described. This multi-class classification approach enables instances to be classified with high accuracy, even when the number of classes is very large. In accordance with one classification scheme, a subset of two or more classes to which the instance is determined to most likely belong is selected from the set of potential classes. A classifier that is generated from a set of training records corresponding to a class set inclusive of the selected subset of classes is applied to the instance a to identify at least one class to which the instance most likely belongs.
Claims20
A method of classifying an instance into one or more classes selected from a set of potential classes, comprising:
- selecting from the set of potential classes a subset of two or more classes to which the instance is determined to most likely belong;
- and
- applying to the instance a scrutiny classifier generated from a set of training records corresponding to a class set inclusive of the selected subset of classes to identify at least one class to which the instance most likely belongs.
A system for classifying an instance into one or more classes selected from a set of potential classes, comprising:
- a ballpark classifier configured to select from the set of potential classes a subset of two or more classes to which the instance is determined to most likely belong;
- and
- a scrutiny classifier configured to identify from the selected subset of classes at least one class to which the instance most likely belongs.
A computer program residing on a computer-readable medium for causing a processor executing the computer program to classifying an instance into one or more classes selected from a set of potential classes, the computer program comprising instructions to:
- select from the set of potential classes a subset of two or more classes to which the instance is determined to most likely belong;
- and
- apply to the instance a scrutiny classifier generated from a set of training records corresponding to a class set inclusive of the selected subset of classes to identify at least one class to which the instance most likely belongs.
Dependencies are inferred from the claim text.