Method and system for expertise mapping based on user activity in recommender systems

• US 20050131897A1
• Filed: 12/11/2003
• Published: 06/16/2005
• Est. Priority Date: 12/11/2003
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Grasso, Antonietta, Bergholz, Andre
• Granted Patent: US 7,240,055 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 16/9536 Search customisation based on social or collaborative filtering
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

The system and method analyzes activities of users in a recommender system to identify experts. Two types of experts may be identified: authoritative experts and hub experts. An authoritative expert is a user who creates items or documents in a given field that are acted upon by a large number of users within a community who are well connected in the given field. A hub expert is a user who is aware of important items or documents within a given field and uses, works or acts upon items or documents produced by authoritative experts. The system and method also identifies authoritative items (e.g., documents). An authoritative item or document is one which has been acted upon by a large number of users within a community who are well connected in the given field (i.e., hub experts).

Claims

1. A method for identifying an expert within a community of users in a recommender system, comprising:

identifying items in a particular field provided by users within the community;

for each identified item;

determining which users within the community have acted upon the identified item;

determining which of the users who have acted upon the identified document are aware of other relevant items in the particular field; and

defining those users as hub experts in the particular field;

determining which identified items have been acted upon by a predetermined number of hub experts; and

defining those items as authoritative items.

2. The method of claim 1, further comprising associating names of the hub experts with the authoritative items.

3. The method of claim 1, acting upon an item comprises one of reading the item, reviewing the item, commenting on the item and recommending the item.

4. The method of claim 3, wherein recommending an item comprises submitting a numerical rating.

5. The method of claim 3, wherein recommending an item comprises a submitting a numerical rating and a comment.

6. The method of claim 1, wherein the determining steps comprise:

defining a connectivity matrix M, wherein element M_ij is non-zero if and only if there is an arc from node i to node j;

wherein for a given node i, a_i is an “

authority”

value and h_i is a “

hub”

value, such that;

$a_i=\underset{j}{\overset{\text{\hspace{1em}<br><br>}}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>M_{\mathrm{ji}}<br><br>h_j<br><br>\text{\hspace{1em}<br><br>}<br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{and}<br><br>\text{\hspace{1em}<br><br>}<br><br>h_i=\underset{j}{\overset{\text{\hspace{1em}<br><br>}}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>M_{\mathrm{ij}}<br><br>a_j$

whose solutions are the principal eigenvectors of M′

M_ and MM′

, respectively; and

defining vectors a=[a₁, a₂, . . . a_n]′ and

h=[h₁, h₂, . . . h_n]′

, such that a=M′

h=M′

Ma, and h=Ma=MM′

h.

7. The method of claim 6, wherein node i represents an authoritative item and node j represents a hub expert;

and

wherein a represents an authority value for an authoritative item and wherein h represents a hub value for a hub expert.

8. The method of claim 6, wherein M_ij is 1 if and only if there is an arc from node i to node j.

9. The method of claim 7, wherein M_ij is max-|rij-Rj|, where max denotes the maximal rating possible, rij denotes the rating of user i for item j and Rj denotes the average rating of item j so far.

10. A method for identifying an expert within a community of users in a recommender system, comprising:

identifying items in a particular field created by users within the community;

for each identified item;

determining which users within the community have acted upon the identified item;

determining which of the users who have acted upon the identified document are aware of other relevant items in the particular field; and

defining those users as hub experts in the particular field;

determining which users creating identified items have had their identified items acted upon by a predetermined number of hub experts; and

defining those users as authoritative experts.

11. The method of claim 10, further comprising associating names of the hub experts with the authoritative experts.

12. The method of claim 10, acting upon an item comprises one of reading the item, reviewing the item, commenting on the item and recommending the item.

13. The method of claim 12, wherein recommending an item comprises submitting a numerical rating.

14. The method of claim 12, wherein recommending an item comprises a submitting a numerical rating and a comment.

15. The method of claim 10, wherein the determining steps comprise:

defining a connectivity matrix M, wherein element M_ij is non-zero if and only if there is an arc from node i to node j;

wherein for a given node i, a_i is an “

authority”

value and h_i is a “

hub”

value, such that;

$a_i=\underset{j}{\overset{\text{\hspace{1em}<br><br>}}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>M_{\mathrm{ji}}<br><br>h_j<br><br>\text{\hspace{1em}<br><br>}<br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{and}<br><br>\text{\hspace{1em}<br><br>}<br><br>h_i=\underset{j}{\overset{\text{\hspace{1em}<br><br>}}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>M_{\mathrm{ij}}<br><br>a_j$

whose solutions are the principal eigenvectors of M′

M_ and MM′

, respectively; and

defining vectors a=[a₁, a₂, . . . a_n]′ and

h=[h₁, h₂, . . . h_n]′

, such that a=M′

h=M′

Ma, and h=Ma=MM′

h.

16. The method of claim 15, wherein node i represents an authoritative expert and node j represents a hub expert;

and

wherein a represents an authority value for an authoritative expert and wherein h represents a hub value for a hub expert.

17. The method of claim 16, wherein an arc between users B and A is only added if the ratings of A and B for the current item are within the same range, e.g., do not differ by a predetermined value.

18. A recommender system for recommending items to users in a community of interest, comprising:

a memory storing;

items provided by users in a particular field within the community; and

for each identified item;

ratings for the item made by other users in the system and a list of any hub expert users associated with the item; and

a processor for identifying items in the particular field provided by users within the community;

for each identified item;

for determining which users within the community have acted upon the identified item;

for determining which of the users who have acted upon the identified document are aware of other relevant items in the particular field; and

for defining those users as hub experts in the particular field;

for determining which identified items have been acted upon by a predetermined number of hub experts; and

for defining those items as authoritative items.

19. A method for identifying an expert within a community of users in a recommender system, comprising:

a memory storing;

items provided by users in a particular field within the community; and

for each identified item;

ratings for the item made by other users in the system;

a list of authoritative experts and a list of any hub expert users associated with the authoritative experts; and

a processor for identifying items in a particular field created by users within the community;

for each identified item;

for determining which users within the community have acted upon the identified item;

for determining which of the users who have acted upon the identified document are aware of other relevant items in the particular field; and

for defining those users as hub experts in the particular field;

for determining which users creating identified items have had their identified items acted upon by a predetermined number of hub experts; and

for defining those users as authoritative experts.