Systems and methods for identifying entity mentions referencing a same real-world entity

• US 9,069,858 B1
• Filed: 03/15/2013
• Issued: 06/30/2015
• Est. Priority Date: 09/18/2012
• Status: Active Grant

• Current Assignee: Salesforce.com, Inc.
• Original Assignee: Salesforce.com, Inc.
• Inventors: Talmor, Alon
• Primary Examiner(s): AL HASHEMI, SANA A
• Application Number:
  US13/842,621
• Time in Patent Office: 837 Days
• Field of Search: 707/608, 707/687, 707/706, 707/791, 707/828, 707/821, 707/953
• US Class Current: 1/1
• CPC Class Codes: G06F 16/2425 Iterative querying; Query formulation based on the results of a preceding query
• CPC Class Codes: G06F 16/24575 using context
• CPC Class Codes: G06F 16/288 Entity relationship models
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 16/9538 Presentation of query results
• CPC Class Codes: G06F 16/954 Navigation, e.g. using categorised browsing

Abstract

The technology disclosed relates to identifying entity reflections that refer to a same real-world entity. In particular, it relates to using statistical functions to make probabilistic deductions about entity attributes, which are used to construct optimal combinations of entity attributes. These optimal combinations of entity attributes are further used to generate search queries that return more precise search results with greater recall.

Claims

1. A method for identifying entity mentions referencing a same real-world entity, the method including:

selecting one or more core entity attributes that represent a real-world entity as a first search attribute set for use in searching biographical sources, including in the selection applying one or more probability distribution functions or joint probability distribution functions to estimate resulting cohort size;

generating one or more searches for processing by a plurality of biographical sources using the first search attribute set;

electronically receiving, responsive to the first search attribute set, entity reflections that include supplemental entity attributes for the real-world entity;

combining the core and supplemental attributes in an anchor entity candidate data object with extended entity attributes that represent the real-world entity;

selecting one or more extended entity attributes as a second search attribute set for use in searching web sources, including applying one or more further probability distribution functions or joint probability distribution functions to estimate resulting cohort size,

generating one or more further web searches using the second search attribute set;

electronically receiving, responsive to the second search attribute set, more entity reflections that include meta entity attributes for the real-world entity; and

updating the anchor entity candidate to include one or more of the meta entity attributes.

2. The method of claim 1, further including:

selecting one or more meta entity attributes from the further web searches to include in a third search attribute set, including applying one or more additional probability distribution functions or joint probability distribution functions to estimate resulting cohort size;

generating one or more additional searches including the meta entity attributes in the third search attribute set;

electronically receiving, responsive to the third search attribute set, additional entity reflections that include additional meta entity attributes; and

updating the anchor entity candidate to include one or more of the additional entity attributes.

3. The method of claim 1, wherein at least some of the entity reflections represent one or more web mentions or web profiles of the real-world entity.

4. The method of claim 1, further including:

using an entity seed that includes one or more core entity attributes to search across the biographical sources.

5. The method of claim 1, further including:

using an entity profile from a primary biographical source that includes the core entity attributes selected as the first search attribute set used to search the plurality of biographical sources.

6. The method of claim 1, further including:

generating a set of predicate results from one or more web crawls using predicate analysis based on affix schemas with predetermined context words before and after entity mentions;

using one or more of the first, second or third search attribute sets to generate one or more searches for processing against the predicate results set;

electronically receiving entity reflections from searching the predicate results set that include predicated entity attributes for the real-world entity; and

updating the anchor entity candidate to include one or more of the predicated entity attributes.

7. A method for connecting entity reflections to real-world entities in an ambiguous environment, the method including:

selecting one or more core entity attributes that represent a real-world entity as a first search attribute set for use in searching biographical sources, including in the selection applying one or more probability distribution functions or joint probability distribution functions to estimate resulting cohort size;

generating one or more searches for processing by a plurality of biographical sources using the first search attribute set;

electronically receiving, responsive to the first search attribute set, entity reflections that include supplemental entity attributes for the real-world entity;

calculating attribute scores for supplemental attributes using a probability contribution function, wherein the attribute scores specify a quantitative assessment of similarity between the supplemental attributes and the core attributes;

merging supplemental attributes with attributes scores above a predefined threshold with core attributes in an anchor entity candidate data object with extended entity attributes that represent the real-world entity;

selecting one or more extended entity attributes as a second search attribute set for use in searching web sources, including applying one or more further probability distribution functions or joint probability distribution functions to estimate resulting cohort size,

generating one or more further web searches using the second search attribute set;

electronically receiving, responsive to the second search attribute set, more entity reflections that include meta entity attributes for the real-world entity;

calculating attribute scores for meta entity attributes using a probability contribution function, wherein the attribute scores specify a quantitative assessment of similarity between the meta entity attributes and the extended entity attributes; and

updating the anchor entity candidate to include one or more of the meta entity attributes with attribute scores above the predefined threshold.

8. The method of claim 7, further including:

selecting one or more meta entity attributes from the further web searches to include in a third search attribute set, including applying one or more additional probability distribution functions or joint probability distribution functions to estimate resulting cohort size;

generating one or more additional searches including the meta entity attributes in the third search attribute set;

electronically receiving, responsive to the third search attribute set, additional entity reflections that include additional meta entity attributes;

calculating attribute scores for additional meta entity attributes using a probability contribution function, wherein the attribute scores specify a quantitative assessment of similarity between the additional meta entity attributes and meta entity attributes; and

updating the anchor entity candidate to include one or more of the additional entity attributes with attribute scores above the predefined threshold.

9. The method of claim 7, further including:

applying a threshold based classification to one or more of the first, second or third search attribute sets, including;

clustering attributes with attribute scores above an upper threshold in a main cohort,

clustering attributes with attribute scores below a lower threshold in a secondary cohort and

clustering attributes with attribute scores between the upper and lower classification thresholds in a primary cohort;

merging the attributes with highest attribute scores within the primary and secondary cohorts to create anchor entity reflections for the primary and secondary cohorts;

matching the anchor entity reflections of the primary and secondary cohorts to the entity reflections within the main cohort, including in the matching applying probability contribution function to compare attribute scores;

creating an anchor cohort by consolidating one or more matched cohorts with the main cohort; and

searching further based on the anchor cohort to find matches or near matches of the real-world entity.

10. The method of claim 7, wherein the attribute scores specify overall similarity of an attribute to the real-world entity.

11. The method of claim 7, wherein at least some of the entity reflections represent one or more web mentions or web profiles of the real-world entity.

12. The method of claim 7, further including:

using an entity seed that includes one or more core entity attributes to search across the biographical sources.

13. The method of claim 7, further including:

using an entity profile from a primary biographical source that includes the core entity attributes selected as the first search attribute set used to search the plurality of biographical sources.

14. The method of claim 7, further including:

generating a set of predicate results from one or more web crawls using predicate analysis based on affix schemas with predetermined context words before and after entity mentions;

using one or more of the first, second or third search attribute sets to generate one or more searches for processing against the predicate results set;

electronically receiving entity reflections from searching the predicate results set that include predicated entity attributes for the real-world entity; and

updating the anchor entity candidate to include one or more of the predicated entity attributes.

15. A computer system for identifying entity mentions referencing a same real-world entity, the system including:

a processor and a computer readable storage medium storing computer instructions configured to cause the processor to;

select one or more core entity attributes that represent a real-world entity as a first search attribute set for use in searching biographical sources, including in the selection applying one or more probability distribution functions or joint probability distribution functions to estimate resulting cohort size;

generate one or more searches for processing by a plurality of biographical sources using the first search attribute set;

electronically receive, responsive to the first search attribute set, entity reflections that include supplemental entity attributes for the real-world entity;

combine the core and supplemental attributes in an anchor entity candidate data object with extended entity attributes that represent the real-world entity;

select one or more extended entity attributes as a second search attribute set for use in searching web sources, including applying one or more further probability distribution functions or joint probability distribution functions to estimate resulting cohort size;

generate one or more further web searches using the second search attribute set,

electronically receive, responsive to the second search attribute set, more entity reflections that include meta entity attributes for the real-world entity; and

update the anchor entity candidate to include one or more of the meta entity attributes.

16. The computer system of claim 15, further configured to cause the processor to:

select one or more meta entity attributes from the further web searches to include in a third search attribute set, including applying one or more additional probability distribution functions or joint probability distribution functions to estimate resulting cohort size;

generate one or more additional searches including the meta entity attributes in the third search attribute set;

electronically receive, responsive to the third search attribute set, additional entity reflections that include additional meta entity attributes; and

update the anchor entity candidate to include one or more of the additional entity attributes.

17. The computer system of claim 15, wherein at least some of the entity reflections represent one or more web mentions or web profiles of the real-world entity.

18. The computer system of claim 15, further configured to cause the processor to:

use an entity seed that includes one or more core entity attributes to search across the biographical sources.

19. The computer system of claim 15, further configured to cause the processor to:

use an entity profile from a primary biographical source that includes the core entity attributes selected as the first search attribute set used to search the plurality of biographical sources.

20. The computer system of claim 15, further configured to cause the processor to:

generate a set of predicate results from one or more web crawls using predicate analysis based on affix schemas with predetermined context words before and after entity mentions;

use one or more of the first, second or third search attribute sets to generate one or more searches for processing against the predicate results set;

electronically receive entity reflections from searching the predicate results set that include predicated entity attributes for the real-world entity; and

update the anchor entity candidate to include one or more of the predicated entity attributes.

21. A computer system for connecting entity reflections to real-world entities in an ambiguous environment, the system including:

a processor and a computer readable storage medium storing computer instructions configured to cause the processor to;

select one or more core entity attributes that represent a real-world entity as a first search attribute set for use in searching biographical sources, including in the selection applying one or more probability distribution functions or joint probability distribution functions to estimate resulting cohort size;

generate one or more searches for processing by a plurality of biographical sources using the first search attribute set;

electronically receive, responsive to the first search attribute set, entity reflections that include supplemental entity attributes for the real-world entity;

calculate attribute scores for supplemental attributes using a probability contribution function, wherein the attribute scores specify a quantitative assessment of similarity between the supplemental attributes and the core attributes;

merge supplemental attributes with attributes scores above a predefined threshold with core attributes in an anchor entity candidate data object with extended entity attributes that represent the real-world entity;

select one or more extended entity attributes as a second search attribute set for use in searching web sources, including applying one or more further probability distribution functions or joint probability distribution functions to estimate resulting cohort size;

generate one or more further web searches using the second search attribute set;

electronically receive, responsive to the second search attribute set, more entity reflections that include meta entity attributes for the real-world entity;

calculate attribute scores for meta entity attributes using a probability contribution function, wherein the attribute scores specify a quantitative assessment of similarity between the meta entity attributes and the extended entity attributes; and

update the anchor entity candidate to include one or more of the meta entity attributes with attribute scores above the predefined threshold.

22. The computer system of claim 21, further configured to cause the processor to:

select one or more meta entity attributes from the further web searches to include in a third search attribute set, including applying one or more additional probability distribution functions or joint probability distribution functions to estimate resulting cohort size;

generate one or more additional searches including the meta entity attributes in the third search attribute set;

electronically receive, responsive to the third search attribute set, additional entity reflections that include additional meta entity attributes;

calculate attribute scores for additional meta entity attributes using a probability contribution function, wherein the attribute scores specify a quantitative assessment of similarity between the additional meta entity attributes and meta entity attributes; and

update the anchor entity candidate to include one or more of the additional entity attributes with attribute scores above the predefined threshold.

23. The computer system of claim 21, further configured to cause the processor to:

apply a threshold based classification to one or more of the first, second or third search attribute sets, including;

clustering attributes with attribute scores above an upper threshold in a main cohort,

clustering attributes with attribute scores below a lower threshold in a secondary cohort and

clustering attributes with attribute scores between the upper and lower classification thresholds in a primary cohort;

merge the attributes with highest attribute scores within the primary and secondary cohorts to create anchor entity reflections for the primary and secondary cohorts;

match the anchor entity reflections of the primary and secondary cohorts to the entity reflections within the main cohort, including in the matching applying probability contribution function to compare attribute scores;

create an anchor cohort by consolidating one or more matched cohorts with the main cohort; and

search further based on the anchor cohort to find matches or near matches of the real-world entity.

24. The computer system of claim 21, wherein the attribute scores specify overall similarity of an attribute to the real-world entity.

25. The computer method of claim 21, further including:

generate a set of predicate results from one or more web crawls using predicate analysis based on affix schemas with predetermined context words before and after entity mentions;

use one or more of the first, second or third search attribute sets to generate one or more searches for processing against the predicate results set;

electronically receive entity reflections from searching the predicate results set that include predicated entity attributes for the real-world entity; and

update the anchor entity candidate to include one or more of the predicated entity attributes.