Video recommendation based on video co-occurrence statistics

Patent metadata

Field	Value
Publication number	US-11601703-B2
Application number	US-201615276605-A
Country	US
Kind code	B2
Filing date	Sep 26, 2016
Priority date	Dec 14, 2011
Publication date	Mar 7, 2023
Grant date	Mar 7, 2023

Abstract

A system and method provides video recommendations for a target video in a video sharing environment. The system selects one or more videos that are on one or more video playlists together with the target video. The video co-occurrence data of the target video associates the target video and another video on one or more same video playlists and frequency of the target video and another video on the video playlists is computed. Based on the video co-occurrence data of the target video, one or more co-occurrence videos are selected and ranked based on the video co-occurrence data of the target video. The system selects one or more videos from the co-occurrence videos as video recommendations for the target video.

First claim

We claim:

1. A computer method for generating video recommendations for a target video in a video sharing environment, comprising:
identifying, using a hardware processor, a plurality of video playlists associated with the target video, an identified video playlist of the video playlists containing the target video and one or more other videos, wherein the target video and each of the other videos are hosted by a video service;
generating, using the hardware processor, video co-occurrence data for the target video on the plurality of video playlists, wherein generating video co-occurrence data comprises:
for an identified video playlist, generating one or more video pairs, a video pair containing the target video and another video on the video playlist; and
for a generated video pair, computing a frequency of the video pair on the plurality of the video playlists;

generating, using the hardware processor, ranking scores for the other videos on the plurality of the video playlists based on the computed frequency with which the other videos are paired with the target video on the plurality of the video playlists, a ranking score of a video on a video playlist representing a measurement of likelihood of the video being viewed by a user viewing the target video;
in response to determining that a first video and a second video from the other videos have a same computed frequency with the target video, selecting, using the hardware processor, the first video to pair with the target video based on similar uploading time, wherein times at which the target video and each of the other videos were uploaded to the video service are extracted from each of the target video and each of the other videos, wherein the first video in the other videos receives a higher ranking score than the second video in the other videos based on a difference in the times at which the target video and the first video were uploaded to the video service being less than the difference in the times at which the target video and the second video were uploaded to the video service;
generating, using the hardware processor, one or more video recommendations for the target video based on the ranking scores of the other videos;
causing, using the hardware processor, the one or more video recommendations for the target video to be presented while concurrently causing the target video to be presented;
identifying, using the hardware processor, new video recommendations based on the ranking scores of the other videos in response to the user viewing the target video; and
replacing, using the hardware processor, the one or more video recommendations with the new video recommendations while continuing to present the target video by deleting at least one of the previously presented one or more video recommendations.

2. The method of claim 1 , wherein a video playlist is described by metadata associated with the video playlist, the metadata identifying one or more videos on the video playlist and an order of the videos to be played.

3. The method of claim 1 , wherein the video co-occurrence data describes a plurality of associations between the target video and other videos on the plurality of the video playlists, and the plurality of the associations between the target video and other videos on the plurality of the video playlists are determined based on at least one of:
video content of the target video and other videos;
sources of the target video and other videos; and
video quality of the target video and other videos.

4. The method of claim 1 , further comprising:
generating aggregated ranking scores for the other videos on the plurality of the video playlists according to a plurality of weighted ranking factors; and
ordering the other videos on the plurality of the video playlists based on aggregated ranking scores of the other videos and the computed frequency with which the other videos are paired with the target video on the video playlists.

5. The method of claim 1 , further comprising ordering the other videos based on a distance between the target video and the other videos, the distance between the target video and another video determined based on an order of the target video and the another video to be played.

6. The method of claim 1 , further comprising:
generating secondary video co-occurrence data of the target video, the secondary video co-occurrence data of the target video indicating a number of hops between the target video and another video on a video playlist, the number of hops determined based on an order of the target video and the another video to be played.

7. The method of claim 6 , wherein a video having a fewer hops from the target video is more closely related to the target video than another video having a larger number of hops.

8. The method of claim 1 , wherein one or more of the plurality of the video playlists were created by users of the video sharing environment.

9. A non-transitory computer-readable storage medium storing executable computer program instructions for generating video recommendations for a target video in a video sharing environment, the computer program instructions comprising instructions for:
identifying a plurality of video playlists associated with the target video, an identified video playlist of the video playlists containing the target video and one or more other videos, wherein the target video and each of the other videos are hosted by a video service;
generating video co-occurrence data for the target video on the plurality of video playlists, wherein generating video co-occurrence data comprises:
for an identified video playlist, generating one or more video pairs, a video pair containing the target video and another video on the video playlist; and
for a generated video pair, computing a frequency of the video pair on the plurality of the video playlists;

generating, using the hardware processor, ranking scores for the other videos on the plurality of the video playlists based on the computed frequency with which the other videos are paired with the target video on the plurality of the video playlists, a ranking score of a video on a video playlist representing a measurement of likelihood of the video being viewed by a user viewing the target video;
in response to determining that a first video and a second video from the other videos have a same computed frequency with the target video, selecting, using the hardware processor, the first video to pair with the target video based on similar uploading time, wherein times at which the target video and each of the other videos were uploaded to the video service are extracted from each of the target video and each of the other videos, wherein the first video in the other videos receives a higher ranking score than the second video in the other videos based on a difference in the times at which the target video and the first video were uploaded to the video service being less than the difference in the times at which the target video and the second video were uploaded to the video service;
generating one or more video recommendations for the target video based on the ranking scores of the other videos;
causing the one or more video recommendations for the target video to be presented while concurrently causing the target video to be presented;
identifying new video recommendations based on the ranking scores of the other videos in response to the user viewing the target video; and
replacing the one or more video recommendations with the new video recommendations while continuing to present the target video by deleting at least one of the previously presented one or more video recommendations.

10. The computer-readable storage medi

Assignees

• Google Llc

Inventors

• Wei Li
• Zhang Kun
• He Yu
• Cai Xinmei

Classifications

• H04N21/44226

  on social networks · CPC title

• G06N5/04

  Inference or reasoning models · CPC title

• G06N5/00

  Computing arrangements using knowledge-based models · CPC title

• G06F16/735 Primary

  Filtering based on additional data, e.g. user or group profiles · CPC title

• H04N21/26258

  for generating a list of items to be played back in a given order, e.g. playlist, or scheduling item distribution according to such list (retrieval of multimedia data based on playlists G06F16/40) · CPC title