CREATING PERSONALIZED AND CONTINUOUS PLAYLISTS FOR A CONTENT SHARING PLATFORM BASED ON USER HISTORY

• US 20150039620A1
• Filed: 07/31/2013
• Published: 02/05/2015
• Est. Priority Date: 07/31/2013
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Ning, Huazhong
• Granted Patent: US 9,684,656 B2
• Time in Patent Office: Days
• US Class Current: 707/740
• CPC Class Codes: G06F 16/4387 by the use of playlists

Abstract

A method for creating personalized and continuous playlists for a content sharing platform based on user history is disclosed. The method includes grouping content items into groups based on whether users access the content items close together in time and whether the content items appear together in search results and are both selected by the users, the groupings representing a relatedness between pairs of the content items, determining probabilities that the users transition between the groups when viewing the content items, assigning, for a user of the users, a score to each group in a subset of the groups, wherein the scores represents preferences of the user for viewing the content items of the subset, selecting, for the user, a group in the subset based on the assigned scores and the probabilities, and providing a content item from the selected group to include in a playlist generated for the user.

Claims

1. A method comprising:

grouping, by a processing device, content items of a content sharing platform into groups based on whether users access the content items close together in time and also based on whether the content items appear together in search results and are both selected by users when they appear together in search results, each of the groups representing a relatedness between the content items in the group;

determining probabilities that users of the content sharing platform transition between the groups when viewing the content items;

assigning, for a user of the users, a score to each group in a subset of the groups, wherein the scores represent preferences of the user for viewing the content items of the subset;

selecting, for the user, a group in the subset based on the assigned scores and the probabilities; and

providing a content item from the selected group to include in a playlist generated for the user.

2. The method of claim 1, wherein the relatedness between the content items in the group is based on at least one of:

the content items being viewed by the same user in a time interval;

or

the content items being returned as query items and selected in a same search result by one or more of the users.

3. The method of claim 1, wherein each of the assigned scores is a function of a number of viewed content items of the group and is normalized with respect to the other assigned scores associated with the user and assigned to the groups of the subset of groups.

4. The method of claim 1, further comprising:

providing the selected content item for playback in the playlist of the user;

adjusting the scores associated with the user based on user feedback associated with the playback of the playlist; and

repeating the selecting the group, the providing the content item, and the providing the selected content item until an indication is received to stop playback of the playlist of the user.

5. The method of claim 1, wherein the subset of groups is selected based on a number of the content items viewed by the user in each of the groups of the subset.

6. The method of claim 1, wherein the selecting the group further comprises selecting the group of the subset that maximizes a function applied to:

one of the determined probabilities of transitioning from a previous selected group of the subset to the selected group; and

the assigned score for the selected group.

7. The method of claim 1, wherein the selecting the content item is based on at least one of previous appearance of the selected content item in the playlist, quality of the content item, or age of the content item in terms of time since upload to the content sharing platform.

8. The method of claim 4, wherein the assigning the score is repeated based on the adjusted scores associated with the user.

9. The method of claim 1, wherein the playlist of the user comprises a plurality of the content items that stream to the user without interruption.

10. A non-transitory machine-readable storage medium storing instructions which, when executed, cause a processing device to perform operations comprising:

grouping, by the processing device, content items of a content sharing platform into groups based on whether users access the content items close together in time and also based on whether the content items appear together in search results and are selected by users when they appear together in the search results, each of the groups representing a relatedness between the content items of the group;

determining probabilities that users of the content sharing platform transition between the groups when viewing the content items of the groups;

assigning a score associated with a user of the users, the score assigned to each group in a subset of the groups, wherein the scores for the subset represent preferences of the user for viewing the content items of the subset;

selecting, for the user, a group in the subset based on the assigned scores and the probabilities;

providing a content item from the selected group to include in a playlist generated for the user; and

adjusting the scores associated with the user based on user feedback associated with playback of the playlist.

11. The non-transitory machine-readable storage medium of claim 10, wherein the relatedness between the content items in the group is based on at least one of:

the content items being viewed by the same user in a time interval;

or

the content items being returned as query items and selected in a same search result by one or more of the users.

12. The non-transitory machine-readable storage medium of claim 10, wherein the score associated with the user is a function of a number of viewed content items of the group and is normalized with respect to the other scores associated with the user and assigned to the groups of the subset of groups.

13. The non-transitory machine-readable storage medium of claim 10, further comprising repeating the selecting the group and the selecting the content item, wherein the assigning the score associated with the user to the each group is repeated based on new data representing current behavior and feedback of the user received since a previous iteration of the assigning.

14. The non-transitory machine-readable storage medium of claim 10, wherein the selecting the group further comprises selecting the group of the subset of the groups that maximizes a function applied to both of a probability of the probabilities of transitioning from a previous selected group of the subset of groups to the selected group and the assigned score for the selected group.

15. The non-transitory machine-readable storage medium of claim 10, wherein the selecting the content item is based on at least one of appearance of the selected content item in the playlist previously, quality of the content item, or age of the content item in terms of upload to the content sharing platform.

16. A system comprising:

a processing device;

a memory coupled to the processing device, the memory to store a plurality of content items; and

a user modeling component executable by the processing device from the memory, the user modeling component to;

group content items of a content sharing platform into groups based on whether users access the content items close together in time and also based on whether the content items appear together in search results and are both selected by users when they appear together in search results, each of the groups representing a relatedness between the content items of the group;

calculate probabilities that users of the content sharing platform transition between the groups when viewing the content items of the groups; and

generate a user model for a user of the content sharing platform, the user model comprising a plurality of user preference scores each associated with a group of a subset of the groups, wherein the user preference score for each group of the subset represents preferences of the user for viewing the content items of the group of the subset; and

a playlist generation component executable by the processing device from the memory, the playlist generation component to;

select, for the user, a group of the subset based on the user preference scores and the probabilities; and

provide a content item from the selected group to include in a playlist generated for the user.

17. The system of claim 16, wherein the score associated with the user is a function of a number of viewed content items of the group and is normalized with respect to the other scores associated with the user and assigned to the groups of the subset of groups.

18. The system of claim 16, wherein the playlist generation component further to:

adjust the user preferences scores based on user feedback associated with playback of the playlist; and

repeat the selecting the group and the providing the content item, wherein the user preference scores associated with each group is adjusted based on new data representing current behavior and feedback of the user received since a previous iteration of the assigning.

19. The system of claim 16, wherein the playlist generation component to select the group further comprises the playlist generation component to select the group of the subset of the groups that maximizes a function applied to both of a probability of the probabilities of transitioning from a previous selected group of the subset of groups to the selected group and the user preference score for the selected group.

20. The system of claim 16, wherein the playlist generation component selects the content item based on at least one of appearance of the selected content item in the playlist previously, quality of the content item, or age of the content item in terms of upload to the content sharing platform.