Adaptive data classification for data mining

• US 8,140,584 B2
• Filed: 12/09/2008
• Issued: 03/20/2012
• Est. Priority Date: 12/10/2007
• Status: Active Grant

• Current Assignee: Aumni Data Incorporated
• Original Assignee: Aumni Data Incorporated
• Inventors: Guha, Aloke
• Primary Examiner(s): LE, DEBBIE M
• Application Number:
  US12/331,363
• Publication Number:
  US 20090164416A1
• Time in Patent Office: 1,197 Days
• Field of Search: 707/802, 707/737, 707/999.002
• US Class Current: 707/802
• CPC Class Codes: G06F 16/31 Indexing; Data structures therefor; Storage structures

Abstract

A method and system for adaptive classification during information retrieval from unstructured data are provided. The method includes receiving input from a user defining a classification. A sample set of unstructured data based on the user defined classification defined is determined. The sample set of unstructured data is analyzed to determine a classification mapping that maps attributes of the sample set of unstructured data to class labels for the classification. The attributes of a set of data objects in a second set of unstructured data are indexed and one or more data objects in the set of data objects are mapped to the class label based on the classification mapping. Feedback based on the user'"'"'s response to an interaction with results is determined using the class label. Finally, adaptive classification mapping is performed based on analysis of feedback by adjusting the sample set of data objects.

Claims

1. A computer-implemented method comprising:

using a programmed digital computer to automatically perform steps comprising;

receiving input from a user defining a classification;

determining a sample set of unstructured data based on the classification defined by the user;

analyzing the sample set of unstructured data to determine a classification mapping that maps attributes of the sample set of unstructured data to class labels for the classification;

indexing attributes of a set of data objects in a second set of unstructured data;

mapping one or more data objects in the set of data objects to the class label based on the classification mapping that maps indexed attributes of the one or more data objects to the class label;

receiving an ad-hoc query from the one or more users;

providing relevant results from the unstructured data based on the classification mapping of the indexed attributes of the one or more data objects and the query;

determining feedback based on the user'"'"'s response to an interaction with results determined using the class label, wherein the user'"'"'s response comprises explicit feedback received from one or more users who are viewing results determined using the classification mapping, and the explicit feedback comprises an indication that a result is misclassified in the classification; and

adapting the classification mapping based on an analysis of the misclassified result received as feedback by adjusting the sample set used to determine the classification mapping to the class label.

2. The computer-implemented method of claim 1, the steps performed by the programmed digital computer further comprising:

receiving an analytics query from the one or more users or viewing a predefined analytics query results display;

providing analytic results from the one or more data objects being mapped to the class label and the analytics query;

receiving the explicit feedback comprising an indication that an analytic result is misclassified in the classification; and

using the misclassified analytics result to adapt the classification mapping.

3. The computer-implemented method of claim 1, wherein the feedback is implicit feedback that infers relevant unstructured data is misclassified.

4. The computer-implemented method of claim 3, the steps performed by the programmed digital computer further comprising:

providing results for a query;

determining a data object in the results that is considered not incorrectly classified based on which documents in the results are selected; and

using the data object attributes and the feedback to adapt the classification mapping.

5. The computer-implemented method of claim 4, the steps performed by the programmed digital computer further comprising:

providing results for a query;

determining a data object in the results that is considered incorrectly classified based on a navigation history of a user selecting the results; and

using the data object to adapt the classification mapping.

6. The computer-implemented method of claim 1, wherein adapting the classification comprises:

adding a data object to the sample set; and

adapting the classification mapping based on the addition of the data object to the sample set.

7. The computer-implemented method of claim 6, the steps performed by the programmed digital computer further comprising mapping a new set of data objects to the associated class labels based on the adapted classification mapping.

8. An apparatus comprising:

one or more processors; and

logic encoded in one or more tangible media for execution by the one or more processors and when executed operable to;

receive input from a user defining a classification;

determine a sample set of unstructured data based on the classification defined by the user;

analyze the sample set of unstructured data to determine a classification mapping that maps attributes of the sample set of unstructured data to class labels for the classification;

index attributes of a set of data objects in a second set of unstructured data;

map one or more data objects in the set of data objects to the class label based on the classification mapping that maps indexed attributes of the one or more data objects to the class label;

receive an ad-hoc query from the one or more users;

provide the results from the relevant unstructured data based on the classification mapping of the indexed attributes of the one or more data objects and the query;

determine feedback based on the user'"'"'s response to an interaction with results determined using the class label, wherein the feedback comprises explicit feedback received from one or more users who are viewing results determined using the classification mapping, and the explicit feedback comprises an indication that a result is misclassified in the classification; and

adapt the classification mapping based on an analysis of the misclassified result received as feedback by adjusting the sample set used to determine the classification mapping to the class label.

9. The apparatus of claim 8, wherein the logic when executed is further operable to:

receive an analytics query from the one or more users or viewing a predefined analytics query results display;

provide analytic results from the one or more data objects being mapped to the class label and the analytics query;

receive the explicit feedback comprising an indication that an analytic result is misclassified in the classification; and

use the misclassified analytics result to adapt the classification mapping.

10. The apparatus of claim 8, wherein the feedback is implicit feedback that infers relevant unstructured data is misclassified.

11. The apparatus of claim 10, wherein the logic when executed is further operable to:

provide results for a query;

determine a data object in the results that is considered not incorrectly classified based on which documents in the results are selected; and

use the data object attributes and the feedback to adapt the classification mapping.

12. The apparatus of claim 11, wherein the logic when executed is further operable to:

provide results for a query;

determine a data object in the results that is considered incorrectly classified based on a navigation history of a user selecting the results; and

use the data object to adapt the classification mapping.

13. The apparatus of claim 8, wherein logic operable to adapt the classification comprises logic operable to:

add a data object to the sample set; and

adapt the classification mapping based on the addition of the data object to the sample set.

14. The apparatus of claim 13, wherein the logic when executed is further operable to map a new set of data objects to the associated class labels based on the adapted classification mapping.