Inferring user interests

• US 8,473,500 B2
• Filed: 11/07/2011
• Issued: 06/25/2013
• Est. Priority Date: 05/01/2007
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Baluja, Shumeet, Jing, Yushi, Sivakumar, Dandapani, Yagnik, Jay
• Primary Examiner(s): Chbouki, Tarek
• Application Number:
  US13/290,455
• Publication Number:
  US 20120054205A1
• Time in Patent Office: 596 Days
• Field of Search: None
• US Class Current: 707/749
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: G06Q 30/0274 Split fees

Abstract

The subject matter of this specification can be embodied in, among other things, a method that includes determining, for a portion of users of a social network, label values each comprising an inferred interest level of a user in a subject indicated by a label, associating a first user with one or more second users based on one or more relationships specified by the first user, and outputting a first label value for the first user based on one or more second label values of the one or more second users.

Claims

1. A computer-implemented method comprising:

accessing, by a computer system, at least a portion of a graph that includes i) nodes that are associated with users of a social network and ii) edges that connect nodes of the graph;

selecting, by the computer system, a node from the accessed portion of the graph, wherein the node is associated with a user of the social network;

identifying, from the accessed portion of the graph, one or more neighboring nodes that are connected to the selected node by one or more of the edges, wherein the one or more neighboring nodes are associated with one or more other users of the social network;

determining, for the selected node, a plurality of interest values that indicate levels of interest of the user in a plurality of subjects and that are determined based on a plurality of neighbor interest values that are multiplied by one or more of the weights that are associated with the one or more of the edges that connect the selected node to the one or more neighboring nodes, wherein the plurality of neighbor interest values i) are associated with the one or more neighboring nodes and ii) indicate levels of interest of the one or more other users in the plurality of subjects, wherein, for at least one interest value from the plurality of interest values, the determining causes the at least one interest value to change from a first value that indicates a lack of interest or an unknown level of interest by the user in a corresponding subject, to a second value that indicates that the user has an interest in the corresponding subject;

adjusting each of at least a portion of the plurality of interest values to account for a relative strength of the user'"'"'s level of interest in each of at least a portion of the plurality of subjects, wherein each of the at least a portion of the plurality of interest values are adjusted based, at least in part, on a combination of the plurality of interest values; and

providing the adjusted plurality of interest values in association with the selected node of the graph.

2. The method of claim 1, wherein determining the plurality of values includes, at least:

determining, for the selected node, a first interest value that indicates a first level of interest of the user in a first subject, wherein the first interest value is determined based on a first combination of one or more first neighbor interest values, wherein the first neighbor interest values are i) associated with the one or more neighboring nodes and ii) indicate levels of interest of the one or more other users in the first subject; and

determining, for the selected node, a second interest value that indicates a second level of interest of the user in a second subject, wherein the second interest value is determined based on a second combination of one or more second neighbor interest values, wherein the one or more second neighbor interest values are i) associated with the one or more neighboring nodes and ii) that indicate levels of interest of the one or more other users in the second subject;

wherein adjusting each of the at least a portion of the plurality of interest values includes, at least, adjusting the first interest value based on the combination of the plurality of interest values that includes both the first interest value and the second interest value.

3. The method of claim 1, wherein the one or more weights indicate strengths of one or more relationships between the selected node and the one or more neighboring nodes that are connected to the selected node by the one or more edges.

4. A system comprising:

a computer system comprising one or more processors;

a data structure generator of the computer system to access at least a portion of a graph that includes i) nodes that are associated with users of a social network and ii) edges that connect nodes of the graph;

an inferred label generator of the computer system to;

select a node from the accessed portion of the graph, wherein the node is associated with a user of the social network;

identify, from the accessed portion of the graph, one or more neighboring nodes that are connected to the selected node by one or more of the edges, wherein the one or more neighboring nodes are associated with one or more other users of the social network;

determine, for the selected node, a plurality of interest values that indicate levels of interest of the user in a plurality of subjects and that are determined based on a plurality of neighbor interest values that are multiplied by one or more of the weights that are associated with the one or more of the edges that connect the selected node to the one or more neighboring nodes, wherein the plurality of neighbor interest values i) are associated with the one or more neighboring nodes and ii) indicate levels of interest of the one or more other users in the plurality of subjects, wherein, for at least one interest value from the plurality of interest values, the determination of the plurality of interest values causes the at least one interest value to change from a first value that indicates a lack of interest or an unknown level of interest by the user in a corresponding subject, to a second value that indicates that the user has an interest in the corresponding subject; and

adjust each of at least a portion of the plurality of interest values to account for a relative strength of the user'"'"'s level of interest in each of at least a portion of the plurality of subjects, wherein each of the at least a portion of the plurality of interest values are adjusted based, at least in part, on a combination of the plurality of interest values; and

an interface of the computer system to provide the adjusted plurality of interest values in association with the selected node of the graph.

5. The system of claim 4, wherein determining the plurality of values includes, at least:

determining, for the selected node, a first interest value that indicates a first level of interest of the user in a first subject, wherein the first interest value is determined based on a first combination of one or more first neighbor interest values, wherein the first neighbor interest values are i) associated with the one or more neighboring nodes and ii) indicate levels of interest of the one or more other users in the first subject; and

determining, for the selected node, a second interest value that indicates a second level of interest of the user in a second subject, wherein the second interest value is determined based on a second combination of one or more second neighbor interest values, wherein the one or more second neighbor interest values are i) associated with the one or more neighboring nodes and ii) that indicate levels of interest of the one or more other users in the second subject;

wherein adjusting each of the at least a portion of the plurality of interest values includes, at least, adjusting the first interest value based on the combination of the plurality of interest values that includes both the first interest value and the second interest value.

6. The system of claim 4, wherein the one or more weights indicate strengths of one or more relationships between the selected node and the one or more neighboring nodes that are connected to the selected node by the one or more edges.

7. A computer-implemented method comprising:

accessing, by a computer system, at least a portion of a graph that includes i) nodes that are associated with users of a social network and ii) edges that connect nodes of the graph, wherein at least a portion of the edges are associated with weights that indicate strengths of corresponding relationships between nodes of the graph;

selecting, by the computer system, a node from the accessed portion of the graph, wherein the node is associated with a user of the social network;

identifying, from the accessed portion of the graph, one or more neighboring nodes that are connected to the selected node by one or more of the edges, wherein the one or more neighboring nodes are associated with one or more other users of the social network;

determining, for the selected node, a plurality of interest values that indicate levels of interest of the user in a plurality of subjects, wherein the plurality of interest values are determined based on a plurality of neighbor interest values that are multiplied by one or more of the weights that are associated with the one or more of the edges that connect the selected node to the one or more neighboring nodes, wherein the plurality of neighbor interest values are i) associated with the one or more neighboring nodes and ii) indicate levels of interest of the one or more other users in the plurality of subjects;

adjusting each of at least a portion of the plurality of interest values to account for a relative strength of the user'"'"'s level of interest in each of at least a portion of the plurality of subjects, wherein each of the at least a portion of the plurality of interest values are adjusted based, at least in part, on a combination of the plurality of interest values; and

providing the adjusted plurality of interest values in association with the selected node of the graph.

8. The method of claim 7, wherein adjusting each of the at least a portion of the plurality of interest values comprises normalizing each of the at least a portion of the plurality of interest values based on the combination of the plurality of interest values.

9. The method of claim 8, wherein normalizing each of the at least a portion of the plurality of interest values based on the combination of the plurality of interest values comprises normalizing each of the plurality of interest values based on a sum of the plurality of interest values.

10. The method of claim 8, wherein the selecting, identifying, determining, adjusting, and providing steps are performed with respect to each node in the accessed portion of the graph.

11. The method of claim 7, wherein the combination of the plurality of interest values comprises a sum of the plurality of interest values.

12. The method of claim 11, wherein each of the plurality of interest values is determined based on one of:

a current interest value, a combination of one or more neighbor interest values, or a combination of one or more neighbor interest values and a current interest value;

where the current interest value is associated with a current state of the selected node and indicates a level of interest of the user in a subject; and

where the one or more neighbor interest values are i) associated with the one or more neighboring nodes and ii) indicate levels of interest of the one or more other users in the subject.

13. The method of claim 12, wherein the plurality of interest values are determined based on the combination of the one or more neighbor interest values and the combination of the one or more neighbor interest values comprises a sum of the one or more neighbor interest values.

14. The method of claim 7, wherein at least the determining, adjusting and providing steps are performed iteratively.

15. The method of claim 14, wherein the iterated steps are stopped when each of the plurality of interest values are within a predetermined threshold of the interest values from the previous iteration.

16. The method of claim 14, wherein when performing the selecting, the identifying, the determining, and the adjusting for a second node that is connected to the selected node by one or more of the edges, a second plurality of interest values are determined for the second node using the adjusted plurality of interest values for the selected node.

17. The method of claim 7, wherein determining the plurality of values includes, at least:

determining, for the selected node, a first interest value that indicates a first level of interest of the user in a first subject, wherein the first interest value is determined based on a first combination of one or more first neighbor interest values, wherein the first neighbor interest values are i) associated with the one or more neighboring nodes and ii) indicate levels of interest of the one or more other users in the first subject; and

determining, for the selected node, a second interest value that indicates a second level of interest of the user in a second subject, wherein the second interest value is determined based on a second combination of one or more second neighbor interest values, wherein the one or more second neighbor interest values are i) associated with the one or more neighboring nodes and ii) that indicate levels of interest of the one or more other users in the second subject;

wherein adjusting each of the at least a portion of the plurality of interest values includes, at least, adjusting the first interest value based on the combination of the plurality of interest values that includes both the first interest value and the second interest value.

18. The method of claim 7, wherein the one or more weights indicate strengths of one or more relationships between the selected node and the one or more neighboring nodes that are connected to the selected node by the one or more edges.

19. A system comprising:

a computer system comprising one or more processors;

a data structure generator of the computer system to access at least a portion of a graph that includes i) nodes that are associated with users of a social network and ii) edges that connect nodes of the graph, wherein at least a portion of the edges are associated with weights that indicate strengths of corresponding relationships between nodes of the graph;

an inferred label generator of the computer system to;

select a node from the accessed portion of the graph, wherein the node is associated with a user of the social network;

identify, from the accessed portion of the graph, one or more neighboring nodes that are connected to the selected node by one or more of the edges, wherein the one or more neighboring nodes are associated with one or more other users of the social network;

determine, for the selected node, a plurality of interest values that indicate levels of interest of the user in a plurality of subjects, wherein the plurality of interest values are determined based on a plurality of neighbor interest values that are multiplied by one or more of the weights that are associated with the one or more of the edges that connect the selected node to the one or more neighboring nodes, wherein the plurality of neighbor interest values are i) associated with the one or more neighboring nodes and ii) indicate levels of interest of the one or more other users in the plurality of subjects; and

adjust each of at least a portion of the plurality of interest values to account for a relative strength of the user'"'"'s level of interest in each of at least a portion of the plurality of subjects, wherein each of the at least a portion of the plurality of interest values are adjusted based, at least in part, on a combination of the plurality of interest values; and

an interface of the computer system to provide the adjusted plurality of interest values in association with the selected node of the graph.

20. The system of claim 19, wherein adjusting each of the at least a portion of the plurality of interest values comprises normalizing each of the at least a portion of the plurality of interest values based on the combination of the plurality of interest values.

21. The system of claim 20, wherein normalizing each of the at least a portion of the plurality of interest values based on the combination of the plurality of interest values comprises normalizing each of the plurality of interest values based on a sum of the plurality of interest values.

22. The system of claim 20, wherein the selecting, identifying, determining, adjusting, and providing steps are performed with respect to each node in the accessed portion of the graph.

23. The system of claim 19, wherein the combination of the plurality of interest values comprises a sum of the plurality of interest values.

24. The system of claim 23, wherein each of the plurality of interest values is determined based on one of:

a current interest value, a combination of one or more neighbor interest values, or a combination of one or more neighbor interest values and a current interest value;

where the current interest value is associated with a current state of the selected node and indicates a level of interest of the user in a subject; and

where the one or more neighbor interest values are i) associated with the one or more neighboring nodes and ii) indicate levels of interest of the one or more other users in the subject.

25. The system of claim 24, wherein the plurality of interest values are determined based on the combination of the one or more neighbor interest values and the combination of the one or more neighbor interest values comprises a sum of the one or more neighbor interest values.

26. The system of claim 19, wherein the inferred label generator iteratively performs at least the determining, adjusting, and providing steps.

27. The system of claim 26, wherein the iterated steps are stopped when each of the plurality of interest values are within a predetermined threshold of the interest values from the previous iteration.

28. The system of claim 26, wherein, when the inferred label generator performs the selecting, the identifying, the determining, and the adjusting for a second node that is connected to the selected node by one or more of the edges, the inferred label generator uses the adjusted plurality of interest values for the selected node to determine a second plurality of interest values for the second node.

29. The system of claim 19, wherein determining the plurality of values includes, at least:

determining, for the selected node, a first interest value that indicates a first level of interest of the user in a first subject, wherein the first interest value is determined based on a first combination of one or more first neighbor interest values, wherein the first neighbor interest values are i) associated with the one or more neighboring nodes and ii) indicate levels of interest of the one or more other users in the first subject; and

determining, for the selected node, a second interest value that indicates a second level of interest of the user in a second subject, wherein the second interest value is determined based on a second combination of one or more second neighbor interest values, wherein the one or more second neighbor interest values are i) associated with the one or more neighboring nodes and ii) that indicate levels of interest of the one or more other users in the second subject,

wherein adjusting each of the at least a portion of the plurality of interest values includes, at least, adjusting the first interest value based on the combination of the plurality of interest values that includes both the first interest value and the second interest value.

30. The system of claim 19, wherein the one or more weights indicate strengths of one or more relationships between the selected node and the one or more neighboring nodes that are connected to the selected node by the one or more edges.