Interest expansion using a taxonomy

• US 9,020,962 B2
• Filed: 10/11/2012
• Issued: 04/28/2015
• Est. Priority Date: 10/11/2012
• Status: Active Grant

• Current Assignee: Walmart Apollo, LLC (WalMart Inc.)
• Original Assignee: Wal Mart Stores Inc (WalMart Inc.)
• Inventors: Lamba, Digvijay Singh, Chai, Xiaoyong
• Primary Examiner(s): Perveen, Rehana
• Assistant Examiner(s): BUI, THUY T
• Application Number:
  US13/650,077
• Publication Number:
  US 20140108439A1
• Time in Patent Office: 929 Days
• Field of Search: 707/758
• US Class Current: 707/756
• CPC Class Codes: G06F 16/337 Profile generation, learning or modification
• CPC Class Codes: G06F 16/367 Ontology
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

Disclosed are methods for inferring interests of a user based on declared interests of the user. Text for which a user has expressed interest, e.g. “liked” is evaluated to identify at least one principal concept. A principal article for the principal concept is located in a taxonomy and the link structure of the taxonomy analyzed to identify candidate articles related to the principal article. The candidate articles are scored according to a plurality of metrics and these scored are weighted and combined for a final score. Candidate articles are selected for the score and recommendations are generated and recommendations generated based on the concepts of the selected candidate articles.

Claims

1. A method for inferring interests, the method comprising:

receiving, by a computer system, one or more indicators of approval of from a user, each of the one or more indicators having a text corresponding thereto;

evaluating, by the computer system, the one or more texts of the one or more indicators and identifying for the one or more texts one or more principal concepts corresponding thereto;

for each principal concept of the one or more principal concepts;

locating a principal article in a taxonomy corresponding to the principal concept;

identifying candidate articles according to linking of the candidate articles relative to the principal article;

evaluating similarity of the candidate articles to the principal article; and

selecting at least one of the candidate articles according to the similarity;

generating, by the computer system, recommendations for the user in accordance with concepts associated with the at least one of the candidate articles for each of the principal concepts; and

transmitting, by the computer system, for display to the user, the recommendations;

wherein evaluating the similarity of the candidate articles to the principal article further comprises evaluating a lineage similarity of the candidate articles to the principal article; and

wherein the lineage similarity of the candidate articles to the principal article is calculated according to a weighted Jaccard distance between ancestors of the principal article and ancestors of the candidate articles, each ancestor of the principal article and each ancestor of the candidate articles being weighted according to a number of descendants of the each ancestor in the taxonomy including the ancestors of the principal article and the candidate articles.

2. The method of claim 1, wherein identifying candidate articles according to linking of the candidate articles relative to the principal article further comprises identifying as candidate articles those articles in the taxonomy that include links to the principal article and articles in the taxonomy to which the principal article includes links.

3. The method of claim 2, wherein identifying candidate articles according to linking of the candidate articles relative to the principal article further comprises identifying as candidate articles those articles having a same category parent in the taxonomy as the principal article.

4. The method of claim 2, wherein identifying candidate articles according to linking of the candidate articles relative to the principal article further comprises identifying as candidate articles those articles having a same immediate category parent in the taxonomy as the principal article.

5. The method of claim 1, wherein evaluating similarity of the candidate articles to the principal article further comprises evaluating similarity of titles of the candidate articles to a title of the principal article.

6. The method of claim 1, wherein evaluating similarity of the titles of the candidate articles to the title of the principal article further comprises evaluating an edit distance between the titles of the candidate articles and the title of the principal article.

7. The method of claim 1, wherein evaluating similarity of the candidate articles to the principal article further comprises evaluating similarity of out-links of the principal article and the candidate articles.

8. The method of claim 7, wherein evaluating similarity of out-links of the principal article and the candidate articles further comprises evaluating similarity of both link text and link destination for the out-links of the principal article and the candidate articles.

9. The method of claim 7, wherein evaluating similarity of out-links of the principal article and the candidate articles further comprises evaluating similarity of placement of the out-links of the principal article and the candidate articles.

10. The method of claim 7, wherein evaluating similarity of out-links of the principal article and the candidate articles further comprises evaluating a Jaccard distance between the out-links of the principal article and the candidate articles.

11. The method of claim 1, wherein evaluating similarity of the candidate articles to the principal article further comprises evaluating a score for each candidate article relative to the principal article, the score for each candidate article computed as a weighted combination of:

an edit distance between titles of the candidate article and a title of the principal article;

a Jaccard distance between out-links of the principal article and the out-links of the candidate article;

a popularity of the candidate article;

and

a placement similarity between out-links of the principal article and the out-links of the candidate article.

12. The method of claim 1, wherein weighting each ancestor comprises weighting the each ancestor by 1/log2(N), where N is the number of descendants of the each ancestor.

13. A system for inferring interests, the system comprising one or more processors and one or more memory devices operably coupled to the one or more processors, the one or more memory devices storing executable and operational data effective to cause the one or more processors to:

receive one or more indicators of approval of from a user, each of the one or more indicators having a text corresponding thereto;

evaluate the one or more texts of the one or more indicators and identifying for the one or more texts one or more principal concepts corresponding thereto;

for each principal concept of the one or more principal concepts;

locate a principal article in a taxonomy corresponding to the principal concept;

identify candidate articles according to linking of the candidate articles relative to the principal article;

evaluate similarity of the candidate articles to the principal article; and

select at least one of the candidate articles according to the similarity;

generate recommendations for the user in accordance with concepts associated with the selected at least one of the candidate articles for each of the one or more principal concepts; and

transmit for display to the user, the recommendations;

wherein the executable and operational data are further effective to cause the one or more processors to evaluate similarity of the candidate articles to the principal article by evaluating a lineage similarity of the candidate articles to the principal article; and

wherein the executable and operational data are further effective to cause the one or more processors to calculate the lineage similarity of the candidate articles to the principal article according to a weighted Jaccard distance between ancestors of the principal article and ancestors of the candidate articles, each ancestor of the principal article and each ancestor of the candidate articles being weighted according to a number of descendants of the each ancestor in the taxonomy including the ancestors of the principal article and the candidate articles.

14. The system of claim 13, wherein the executable and operational data are further effective to cause the one or more processors to identify candidate articles according to linking of the candidate articles relative to the principal article by identifying as candidate articles those articles in the taxonomy that include links to the principal article and articles in the taxonomy to which the principal article includes links.

15. The system of claim 14, wherein the executable and operational data are further effective to cause the one or more processors to identify candidate articles according to linking of the candidate articles relative to the principal article by identifying as candidate articles those articles having a same category parent in the taxonomy as the principal article.

16. The system of claim 14, wherein the executable and operational data are further effective to cause the one or more processors to identify candidate articles according to linking of the candidate articles relative to the principal article by identifying as candidate articles those articles having a same immediate category parent in the taxonomy as the principal article.

17. The system of claim 13, wherein the executable and operational data are further effective to cause the one or more processors to evaluate similarity of the candidate articles to the principal article by evaluating similarity of titles of the candidate articles to a title of the principal article.

18. The system of claim 13, wherein the executable and operational data are further effective to cause the one or more processors to evaluate similarity of the titles of the candidate articles to the title of the principal article by evaluating an edit distance between the titles of the candidate articles and the title of the principal article.

19. The system of claim 13, wherein the executable and operational data are further effective to cause the one or more processors to evaluate similarity of the candidate articles to the principal article by evaluating similarity of out-links of the principal article and the candidate articles.

20. The system of claim 19, wherein the executable and operational data are further effective to cause the one or more processors to evaluate similarity of out-links of the principal article and the candidate articles by evaluating similarity of both link text and link destination for the out-links of the principal article and the candidate articles.

21. The system of claim 19, wherein the executable and operational data are further effective to cause the one or more processors to evaluate similarity of out-links of the principal article and the candidate articles by evaluating similarity of placement of the out-links of the principal article and the candidate articles.

22. The system of claim 19, wherein the executable and operational data are further effective to cause the one or more processors to evaluate similarity of out-links of the principal article and the candidate articles by evaluating a Jaccard distance between the out-links of the principal article and the candidate articles.

23. The system of claim 13, wherein the executable and operational data are further effective to cause the one or more processors to evaluate similarity of the candidate articles to the principal article by evaluating a score for each candidate article relative to the principal article, the score for each candidate article computed as a weighted combination of:

an edit distance between titles of the candidate article and a title of the principal article;

a popularity of the candidate article;

a Jaccard distance between out-links of the principal article and the out-links of the candidate article; and

a placement similarity between out-links of the principal article and the out-links of the candidate article.