OBJECT CLASSIFICATION USING TAXONOMIES
First Claim
1. A method for classifying objects from a source catalog to a target catalog, comprising:
- determining a probability that an object from a source catalog belongs to a class in a target catalog for respective classes in the target catalog using a baseline statistical classification algorithm;
assigning a source object to a target class corresponding to a desired target class probability for the source object if the target class probability meets a desired threshold;
determining a classification cost for assigning an unassigned source object to a target class for respective target classes from a set of desired target classes, comprising;
determining an assignment cost for respective desired target classes for the unassigned source object, the assignment cost comprising a function of the probability that the unassigned source object belongs to the target class;
determining a separation cost of the unassigned source object for respective desired target classes, comprising combining a function of distance between a desired target class for the unassigned source object and an assigned source object'"'"'s target class and a function of similarity between the unassigned source object and the assigned source object; and
combining the assignment cost and separation cost of the source object for respective desired target classes; and
assigning the unassigned source object to a target class corresponding to a desired classification cost for the unassigned source object.
2 Assignments
0 Petitions
Accused Products
Abstract
As provided herein objects from a source catalog, such as a provider'"'"'s catalog, can be added to a target catalog, such as an enterprise master catalog, in a scalable manner utilizing catalog taxonomies. A baseline classifier determines probabilities for source objects to target catalog classes. Source objects can be assigned to those classes with probabilities that meet a desired threshold and meet a desired rate. A classification cost for target classes can be determined for respective unassigned source objects, which can comprise determining an assignment cost and separation cost for the source objects for respective desired target classes. The separation and assignment costs can be combined to determine the classification cost, and the unassigned source objects can be assigned to those classes having a desired classification cost.
-
Citations
20 Claims
-
1. A method for classifying objects from a source catalog to a target catalog, comprising:
-
determining a probability that an object from a source catalog belongs to a class in a target catalog for respective classes in the target catalog using a baseline statistical classification algorithm; assigning a source object to a target class corresponding to a desired target class probability for the source object if the target class probability meets a desired threshold; determining a classification cost for assigning an unassigned source object to a target class for respective target classes from a set of desired target classes, comprising; determining an assignment cost for respective desired target classes for the unassigned source object, the assignment cost comprising a function of the probability that the unassigned source object belongs to the target class; determining a separation cost of the unassigned source object for respective desired target classes, comprising combining a function of distance between a desired target class for the unassigned source object and an assigned source object'"'"'s target class and a function of similarity between the unassigned source object and the assigned source object; and combining the assignment cost and separation cost of the source object for respective desired target classes; and assigning the unassigned source object to a target class corresponding to a desired classification cost for the unassigned source object. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)
-
-
16. A system for classifying objects from a source catalog to a target catalog, comprising:
-
a baseline classifier configured to determine a probability that an object from a source catalog belongs to a class in a target catalog for respective classes in the target catalog; an object assignment component configured to assign a source object to a target class corresponding to a desired probability for the source object if the desired probability meets a desired threshold; a classification cost determination component configured to determine a classification cost of an unassigned source object for respective target classes from a set of desired target classes, comprising; an assignment cost determination component configured to determine an assignment cost of respective desired target classes for the unassigned source object, the assignment cost comprising a function of the probability that the unassigned source object belongs to a desired target class; a separation cost determination component configured to determine a separation cost of the unassigned source object for the respective desired target classes, the separation cost comprising a combination of a function of similarity between the unassigned source object and an assigned source object and a function of distance between a desired target class for the unassigned source object and the assigned source object'"'"'s target class; and a combination component configured to combine the assignment cost and separation cost of respective source objects for respective desired target classes; and an unassigned object assignment component configured to assign respective unassigned source objects to corresponding target classes having a desired classification cost for the unassigned source objects. - View Dependent Claims (17, 18, 19)
-
-
20. A method for classifying products from a source catalog to a target catalog, comprising:
-
determining a probability that an object from a source catalog belongs to a class in a target catalog for respective classes in the target catalog using a baseline statistical classification algorithm; assigning a source object to a target class corresponding to a highest target class probability for the source object if the target class probability meets a desired threshold; determining a classification cost for assigning an unassigned source object to a target class for respective target classes from a set of desired target classes, comprising; determining an assignment cost for respective desired target classes for the unassigned source object, comprising; determining a probability that an object belongs to respective desired target classes using the baseline statistical classification algorithm; and calculating a function of the probability for respective desired target classes; determining a separation cost of the unassigned source object for respective desired target classes, comprising; determining a function of distance between the target class for the unassigned source object and the assigned source object'"'"'s target class comprising determining a number of node hops from the target class for the unassigned source object to the assigned source object'"'"'s target class in a hierarchical taxonomy tree of target classes; determining a function of similarity between the unassigned source object and an assigned source object comprising determining a similarity between the unassigned source object'"'"'s source class and the assigned source object'"'"'s source class from the source catalog; and combining the function of similarity and the function of distance with a normalization factor; and summing the assignment cost with a combination of a regularization parameter and the separation cost; and assigning the unassigned source object to a target class corresponding to a lowest classification cost for the unassigned source object.
-
Specification