NAVIGATION SYSTEM WITH POINT OF INTEREST CLASSIFICATION MECHANISM AND METHOD OF OPERATION THEREOF

• US 20130166480A1
• Filed: 12/21/2011
• Published: 06/27/2013
• Est. Priority Date: 12/21/2011
• Status: Active Grant

• Current Assignee: TeleNav Incorporated
• Original Assignee: TeleNav Incorporated
• Inventors: Popescu, Liviu Teodor, Narasimha, Pramod Lakshmi, Husain, Aliasgar Mumtaz
• Granted Patent: US 9,026,480 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

A method of operation of a navigation system includes: generating a training data from a randomly sampled uncategorized point of interest; generating a trained classifier model by training a classifier model using the training data; generating a category identifier, a confidence score, or a combination thereof for an uncategorized point of interest using the trained classifier model; generating a categorized point of interest by assigning the category identifier to the uncategorized point of interest; calculating a weighted confidence score based on a weighted F-measure for the category identifier, a pair of the category identifier and the confidence score; and consolidating the categorized point of interest based on the weighted confidence score for the category identifier being meeting or exceeding a threshold for displaying on a device.

Claims

1. A method of operation of a navigation system comprising:

generating a training data from a randomly sampled uncategorized point of interest;

generating a trained classifier model by training a classifier model using the training data;

generating a category identifier, a confidence score, or a combination thereof for an uncategorized point of interest using the trained classifier model;

generating a categorized point of interest by assigning the category identifier to the uncategorized point of interest;

calculating a weighted confidence score based on a weighted F-measure for the category identifier, a pair of the category identifier and the confidence score; and

consolidating the categorized point of interest based on the weighted confidence score for the category identifier being meeting or exceeding a threshold for displaying on a device.

2. The method as claimed in claim 1 further comprising:

generating a category tree having the categorized point of interest; and

processing the category tree by eliminating a dangling node and a descendent node.

3. The method as claimed in claim 1 further comprising:

calculating the weighted F-measure for the category identifier based on a weighted precision, a weighted recall and a search weight.

4. The method as claimed in claim 1 wherein calculating the weighted confidence score includes:

calculating a total confidence score for the category identifier for the categorized point of interest based on the pair of the category identifier and confidence score; and

generating the weighted confidence score based on the total confidence score and the weighted F-measure for the category identifier for the categorized point of interest.

5. The method as claimed in claim 1 wherein generating the training data includes:

generating a random query result using a positive keyword and a negative keyword for the category identifier to query the randomly sampled uncategorized point of interest;

verifying the random query result is positive or negative for the category identifier and generating the training data by associating the random query result which is positive with the category identifier;

updating the trained classifier model by training the classifier model using the training data;

updating the category identifier, the confidence score, or a combination thereof by classifying the randomly sampled uncategorized point of interest utilizing the trained classifier model;

generating an initially classified POI based on the confidence score for the category identifier being less than a fixed threshold; and

correcting the initially classified POI and updating the training data with a correctly categorized point of interest.

6. A method of operation of a navigation system comprising:

generating a training data from a randomly sampled uncategorized point of interest;

generating a trained classifier model by training a classifier model using the training data;

generating a category identifier, a confidence score, or a combination thereof for an uncategorized point of interest using the trained classifier model;

generating a categorized point of interest by assigning the category identifier to the uncategorized point of interest;

calculating a weighted confidence score based on a weighted F-measure for the category identifier, a pair of the category identifier and the confidence score;

consolidating the categorized point of interest based on the weighted confidence score for the category identifier being meeting or exceeding a threshold for displaying on a device; and

processing mutually exclusive category identifiers for the categorized point of interest by eliminating an incorrect category identifier.

7. The method as claimed in claim 6 further comprising:

calculating the weighted F-measure for the category identifier based on a precision, a recall and a search weight.

8. The method as claimed in claim 6 wherein calculating the weighted confidence score includes:

calculating a total confidence score for the category identifier for the categorized point of interest based on the pair of the category identifier and the confidence score; and

generating the weighted confidence score based on the total confidence score and the weighted F-measure for the category identifier for the categorized point of interest.

9. The method as claimed in claim 6 wherein eliminating the incorrect category identifier includes:

eliminating the incorrect category identifier with the smaller confidence score when there are only two category identifiers assigned to the categorized point of interest; and

eliminating the incorrect category identifier with a smaller total probability confidence when there are more than two category identifiers assigned to the categorized point of interest.

10. The method as claimed in claim 6 wherein eliminating the incorrect category identifier includes:

searching a total category set for the categorized point of interest;

generating a minimum category set contains the incorrect category identifier from the total category set; and

generating a maximum category set without the incorrect category identifier by eliminating the minimum category set from the total category set for the categorized point of interest.

11. A navigation system comprising:

an active learning module for generating a training data from a randomly sampled uncategorized point of interest;

a model training module, coupled to the active learning module, for generating a trained classifier model by training a classifier model using the training data;

a classification module, coupled to the model training module, for generating a category identifier, a confidence score, or a combination thereof for an uncategorized point of interest and generating a categorized point of interest with the trained classifier model; and

a consolidation module, coupled to the classification module, for calculating a weighted confidence score based on a weighted F-measure for the category identifier, a pair of the category identifier and the confidence score and consolidating the categorized point of interest based on the weighted confidence score for the category identifier being meeting or exceeding a threshold for displaying on a device.

12. The system as claimed in claim 11 further comprising:

a category tree generation module for generating a category tree having the categorized point of interest; and

a category tree process module, coupled to the category tree generation module, for processing the category tree by eliminating a dangling node and a descendent node.

13. The system as claimed in claim 11 further comprising:

an accuracy measure generation module for calculating the weighted F-measure for the category identifier based on a weighted precision, a weighted recall and a search weight.

14. The system as claimed in claim 11 wherein the consolidation module includes:

a total confidence score module for calculating a total confidence score for the category identifier for the categorized point of interest based on the pair of category identifier and confidence score;

a multiplication module, coupled to the total confidence score module, for generating the weighted confidence score based on the total confidence score and the weighted F-measure for the category identifier for the categorized point of interest; and

a clean-up module, coupled to the multiplication module, for eliminating the category identifier with weighted confidence score less than the threshold.

15. The system as claimed in claim 11 wherein an active learning module includes:

a keyword search module for generating a random query result using a positive keyword and a negative keyword for the category identifier to query the randomly sampled uncategorized point of interest;

a category assignment module, coupled to the keyword search module, for verifying the random query result is positive or negative for the category identifier and generating the training data by associating the random query result which is positive with the category identifier;

a training data module, coupled to the category assignment module, for storing the training data generated by the category assignment module;

the model training module, coupled to the training data module, for updating the trained classifier models by training the classifier models using the training data;

a random samples classification module, coupled to the model training module, for updating the category identifier, the confidence score, or a combination thereof by classifying the randomly sampled uncategorized point of interest utilizing the trained classifier model;

an initially classified POIs selection module, coupled to the random samples classification module, for generating an initially classified POI based on the confidence score for the category identifier being less than a fixed threshold; and

an error correction module, coupled to the initially classified POIs selection module, for correcting the initially classified POI and updating the training data with the correctly classified point of interest.

16. The system as claimed in claim 11 further comprising:

a mutually exclusive categories module, coupled to the consolidation module, for processing mutually exclusive category identifiers for the categorized point of interest by eliminating an incorrect category identifier.

17. The system as claimed in claim 16 further comprising:

an accuracy measure generation module for calculating the weighted F-measure for the category identifier based on a weighted precision, a weighted recall and a search weight.

18. The system as claimed in claim 16 wherein the mutually exclusive categories module includes:

a confidence module, coupled to the consolidation module, for eliminating the incorrect category identifier with the smaller confidence score in the mutually exclusive category identifiers when there are only two category identifiers assigned to the categorized point of interest; and

a probability confidence module, coupled to the consolidation module, for eliminating the incorrect category identifier with a smaller total probability confidence in the mutually exclusive category identifiers when there are more than two category identifiers assigned to the categorized point of interest.

19. The system as claimed in claim 16 wherein the mutually exclusive categories module includes:

a total category module for searching a total category set from the mutually exclusive category identifiers for the categorized point of interest;

a minimum category module, coupled to the total category module, for generating a minimum category set contains the incorrect category identifier from the total category set; and

a maximum category module, coupled to the minimum category module, for generating a maximum category set without the incorrect category identifier by eliminating the minimum category set from the total category set for the categorized point of interest.

20. The system as claimed in claim 16 further comprising a category tree process module including:

a dangling node search module for searching a dangling node from the category tree;

a node elimination module, coupled to the dangling node search module, for eliminating the dangling node and a descendent node; and

a confidence score update module, coupled to the node elimination module, for updating the confidence score corresponding to the dangling node and the descendent node.