Natural language parsers to normalize addresses for geocoding

• US 9,390,084 B2
• Filed: 10/03/2014
• Issued: 07/12/2016
• Est. Priority Date: 09/28/2007
• Status: Active Grant

• Current Assignee: Verizon Patent and Licensing Incorporated (Verizon Communications Inc.)
• Original Assignee: Telogis Incorporated (Verizon Communications Inc.)
• Inventors: Mitchell, David John, Morris, Arthur Newth, Mason, Ralph James
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): COUGHLAN, PETER D
• Application Number:
  US14/506,459
• Publication Number:
  US 20150127324A1
• Time in Patent Office: 648 Days
• Field of Search: 706/62
• US Class Current: 1/1
• CPC Class Codes: G06F 40/205 Parsing
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 40/295 Named entity recognition

Abstract

The present invention provides a technique for building natural language parsers by implementing a country and/or jurisdiction specific set of training data that is automatically converted during a build phase to a respective predictive model, i.e., an automated country specific natural language parser. The predictive model can be used without the training data to quantify any input address. This model may be included as part of a larger Geographic Information System (GIS) data-set or as a stand alone quantifier. The build phase may also be run on demand and the resultant predictive model kept in temporary storage for immediate use.

Claims

1. A method for normalizing an input address, comprising:

under control of a computer system comprising computer hardware;

identifying an input address indicative of a physical address;

selecting a predictive model corresponding to the input address from a plurality of predictive models, each of the plurality of predictive models being an automated country-specific natural language parser uniquely defined for a corresponding one of a plurality of countries and jurisdictions, the selected predictive model comprising a graph having address field nodes and edges connecting the address field nodes, each of at least some of the address field nodes comprising an address field and a corresponding set of one or more address field classifications each assigned a first probability value, each of at least some of the edges assigned a second probability value; and

executing the selected predictive model to determine an address field classification for the input address from one of the address fields in the graph based at least on the first and second probability values of the address field nodes and the edges that correspond to the address field classification.

2. The method of claim 1, wherein said executing comprises executing the selected predictive module to determine a plurality of address field classifications for the input address from the address field nodes in the graph, the plurality of address field classifications including the address field classification.

3. The method of claim 1, further comprising generating the selected predictive model from a training file, the training file comprising a plurality of tokens.

4. The method of claim 3, wherein the training file defines one or more address expression formats.

5. The method of claim 3, wherein the training file is associated with a particular country or jurisdiction.

6. The method of claim 1, wherein the selected predictive model comprises a table of probabilities associated with the address fields.

7. A system for normalizing address input, the system comprising:

computer hardware programmed to;

identify an input address indicative of a physical address;

select a predictive model corresponding to the input address from a plurality of predictive models, each of the plurality of predictive models representing an automated country-specific natural language parser uniquely defined for a corresponding one of a plurality of jurisdictions, the selected predictive model comprising (1) address field data including address fields and address field classifications corresponding to the address fields and (2) probability values assigned to the address field data; and

execute the selected predictive model to match a selected one of the address field classification with the input address based at least on the probability values.

8. The system of claim 7, wherein the computer hardware is further configured to generate the selected predictive model from a training file, the training file comprising a plurality of tokens training field values.

9. The system of claim 8, wherein the training file defines one or more address expression formats.

10. The system of claim 8, wherein the training file is associated with a particular jurisdiction.

11. The system of claim 8, wherein the computer hardware is further configured to generate the selected predictive model by identifying a most likely address field of the address fields for each of at least some of the plurality of tokens.

12. The system of claim 8, wherein the computer hardware is further configured to generate the selected predictive model by determining a number of times that the plurality of tokens occur in a given address field of the address fields.

13. The system of claim 7, wherein the selected predictive model comprises a table of probabilities associated with the address fields.