System and method for mapping text phrases to geographical locations

• US 9,747,278 B2
• Filed: 02/23/2012
• Issued: 08/29/2017
• Est. Priority Date: 02/23/2012
• Status: Active Grant

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Palo Alto Research Center, Inc. (Xerox Holdings Corp.)
• Inventors: Bier, Eric A., Wu, Anna
• Primary Examiner(s): Thomas, Ashish K
• Assistant Examiner(s): Ohba, Mellissa M
• Application Number:
  US13/403,893
• Publication Number:
  US 20130226928A1
• Time in Patent Office: 2,014 Days
• CPC Class Codes: G06F 16/29 Geographical information databases
• CPC Class Codes: G06F 40/295 Named entity recognition

Abstract

A system and method for mapping text phrases to geographical locations is provided. Entities, each comprising one of a location, person, and place, are identified in one or more documents. Possible candidate locations associated with each entity are determined. An initial score is assigned to each location. The initial scores are adjusted and the candidate location with the highest adjusted score is selected for each entity. The selected candidate location is applied to all occurrences of the entity in the documents.

Claims

1. A method for mapping text phrases to geographical locations, comprising:

identifying in a document, an ambiguous entity that represents two or more representative entity types, wherein a first of the entity types further represents two or more candidate locations;

assigning an initial score to a second of the entity types and to each of the candidate locations;

determining the initial scores for the candidate locations, comprising;

determining for each candidate location an administrative level comprising one of a region, country, state, county, and city to which that candidate location belongs, wherein each administrative level is associated with a predetermined score;

assigning to the each of the candidate locations, the predetermined value associated with the determined administrative level for that candidate location;

when at least one of the candidate locations comprises a city, determining a value for that candidate location based on a comparative size of the candidate location by identifying an average size city in a same country as that candidate location and comparing a size of that candidate location to the average size city in the same country; and

combining the values for the administrative level and the comparative size as the initial score for each of the candidate locations comprising a city and designating the predetermined value associated with the determined administrative level for the candidate locations other than a city;

adjusting the initial scores based on one or more subroutines;

selecting one of the second representative entity type, a first of the candidate locations, and a second of the candidate locations with the highest adjusted initial score as an identified label for the ambiguous entity; and

applying the identified label to all occurrences of the ambiguous entity in the document.

2. A method according to claim 1, further comprising:

determining the initial score for the second entity type that comprises a person representative entity type, comprising;

identifying a popularity of a name for the person representative entity type; and

assigning a predetermined score to the person representative entity type based on the identified popularity.

3. A method according to claim 1, further comprising:

determining the adjusted scores based on at least one of heuristics, context, vicinity, and focus regions.

4. A method according to claim 1, further comprising:

generating an index, comprising;

identifying occurrences of the entity in the documents; and

compiling a list of the entity occurrences;

utilizing the index to apply the selected location to all occurrences of the entity in the documents.

5. A method according to claim 1, wherein the representative entity types for the ambiguous entity comprise at least one of a location, person name, and time.

6. A method for mapping text phrases to geographical locations, comprising:

identifying in a document, an ambiguous entity that represents two or more representative entity types comprising a location, person, or time;

assigning an initial score to each representative entity type and adjusting the initial score based on one or more subroutines;

determining the initial score for the location, comprising;

determining an administrative level comprising one of a region, country, state, county, and city to which the location belongs, wherein each administrative level is associated with a predetermined value and assigning the predetermined value based on the determined administrative level for that location;

when the location comprises a city, determining a value based on a comparative size of that location by identifying an average size city in a same country as the location and comparing a size of the location to the average size city in the same country; and

combining the values for the administrative level and the comparative size as the initial score for the location;

selecting the representative entity type with the highest score for the entity; and

applying the representative entity type to all occurrences of the entity in the document.

7. A method according to claim 6, further comprising:

generating entity lists, comprising;

predefining entities comprising at least one of identifying predefined words and determining regular expression patterns; and

generating a list of names for each entity; and

matching the ambiguous entity to a name in at least two of the entity lists to identify the representative entity types.

8. A method according to claim 6, further comprising:

determining the initial score for the person representative entity type, comprising;

identifying a popularity of the person representative entity type; and

assigning a predetermined score to the person representative entity type based on the identified popularity.

9. A method according to claim 6, further comprising:

determining the adjusted scores based on at least one of heuristics, context, vicinity, and focus regions.

10. A method according to claim 6, further comprising:

generating an index, comprising;

identifying occurrences of the entity in the documents; and

compiling a list of the entity occurrences;

utilizing the index to apply the selected location to all occurrences of the entity in the documents.

11. A method for mapping text phrases to geographical locations, comprising:

identifying in a document, an ambiguous entity that represents two or more locations as candidate locations;

determining an initial score for the locations associated with the candidate locations, comprising;

determining for each location an administrative level comprising one of a region, country, state, county, and city to which that location belongs, wherein each administrative level is associated with a predetermined score;

assigning to each of the locations, the predetermined value associated with the determined administrative level for that location;

when at least one of the locations comprises a city, determining a value for that location based on a comparative size of the location by identifying an average size city in a same country as that location and comparing a size of that location to the average size city in the same country; and

combining the values for the administrative level and the comparative size as the initial score for each of the locations comprising a city and designating the predetermined value associated with the determined administrative level for the locations other than a city;

adjusting the initial scores based on one or more subroutines;

selecting the location of the candidate with the highest adjusted score for the ambiguous entity; and

applying the selected location to all occurrences of the entity in the document.

12. A method according to claim 11, wherein the representative entity types for the ambiguous entity comprise at least one of a location, person name, and time.

13. A method according to claim 11, further comprising:

generating entity lists, comprising;

predefining entities; and

generating a list of names for each entity; and

matching the ambiguous entity to a name in at least two of the entity lists to identify the representative entity types.

14. A method according to claim 11, further comprising:

determining at least one focus region for one such document comprising;

determining a frequency of occurrence of each candidate location belonging to a particular region;

assigning a score to each region based on the frequencies of occurrence and weight for the location candidates; and

selecting the region with the highest score as the focus region.

15. A method according to claim 11, further comprising:

increasing the score of the locations that belong to the focus region.

16. A method according to claim 15, wherein the focus region is determined in accordance with the following equation:

17. A method according to claim 11, further comprising:

determining the adjusted scores based on at least one of heuristics, context, vicinity, and focus regions.

18. A method according to claim 11, further comprising:

dividing a map of the earth into cells each associated with latitude and longitude coordinates;

mapping one of the locations into the cells of the map;

assigning scores to the cells based on a number of other locations that are located in that cell; and

assigning the location to the cell with the highest score.

19. A method according to claim 11, wherein the location is mapped into the cells via a hash function in which a key is based on a range of latitudes and longitudes within at least one of the cells.

20. A method according to claim 11, further comprising:

generating an index, comprising;

identifying occurrences of the entity in the documents; and

compiling a list of the entity occurrences;

utilizing the index to apply the selected location to all occurrences of the entity in the documents.