Back to results
US 9,058,332 B1Granted
Blended ranking of dissimilar populations using an N-furcated normalization technique
AssigneeGoogle Inc.
InventorsMatthew Thomas Darby, Naga Naresh Karuturi
Life of this patent
Estimating the term…
Abstract
Systems and methods for blending dissimilar, ordered populations into a single selection for users are disclosed herein. In an aspect, content items belonging to distinct parent populations which display a large disparity in the value which is used for ranking purposes, can be displayed together in a single continuously ranked list for simple browsing and selection by users. Further, a score can be assigned to the respective media content items based at least in part on a median value of a distribution of media content items corresponding to the respective parent populations and this score can be used as a normalized, universal value with which to rank content from all dissimilar populations together.
Claims22
A system comprising:
- a memory having stored thereon computer executable components;
- and
- a processor configured to execute the following computer executable components stored in the memory:
- an analysis component that analyzes social network tags related to media content items belonging to respective parent populations based at least in part on user browsing information to determine user interest of the media content items, wherein the respective parent populations contain statistical information associated with the population of media content items, the statistical information comprising at least organization and interpretation of data within the population;
- a scoring component that assigns respective scores to the media content items based at least in part on a distribution of media content items corresponding to the respective parent populations according to the following:
- determine respective view counts for media content items of the parent populations;
- group respective view counts into buckets by powers of ten for the parent populations;
- identify respective buckets for parent populations having a maximum count value;
- determine respective geometric means of the respective buckets for parent populations having the maximum count value;
- determine respective offsets for the parent populations based upon respective differences between the geometric means;
- and
- determine the respective scores of media content items based upon the respective offsets of the parent populations;
- and
- a ranking component that ranks the media content items based at least in part on the respective scores;
- and
- an ordering component that aggregates the media content items belonging to the respective parent populations into an ordered list based at least in part on the ranking.
13 more elements
A method, comprising:
- analyzing, by a system including a processor, social network tags related to media content items belonging to respective parent populations based at least in part on user browsing information to determine user interest of the media content items, wherein the respective parent populations contain statistical information associated with the population of media content items, the statistical information comprising at least organization and interpretation of data within the population;
- assigning, by the system, respective scores to the media content items based at least in part on a distribution of media content items corresponding to the respective parent populations, the assigning comprising:
- ascertaining respective view counts for media content items of the parent populations;
- grouping respective view counts into buckets by powers of ten for the parent populations;
- identifying respective buckets for parent populations having a maximum count value;
- ascertaining respective geometric means of the respective buckets for parent populations having the maximum count value;
- ascertaining respective offsets for the parent populations based upon respective differences between the geometric means;
- and
- generating the respective scores of media content items based upon the respective offsets of the parent populations;
- and
- ranking, by the system, the media content items based at least in part on the score;
- and
- consolidating, by the system, the media content items belonging to the respective parent populations into an ordered list based at least in part on the ranking.
10 more elements
A non-transitory computer-readable medium having instructions stored thereon that, in response to execution, cause a system including a processor to perform operations comprising:
- analyzing social network tags related to media content items belonging to respective parent populations based at least in part on user browsing information to determine user interest of the media content items, wherein the respective parent populations contain statistical information associated with the population of media content items, the statistical information comprising at least organization and interpretation of data within the population;
- assigning respective scores to the media content items based at least in part on a distribution of media content items corresponding to the respective parent populations, the assigning comprising:
- determining respective view counts for media content items of the parent populations;
- grouping respective view counts into buckets by powers of ten for the parent populations;
- identifying respective buckets for parent populations having a maximum count value;
- determining respective geometric means of the respective buckets for parent populations having the maximum count value;
- determining respective offsets for the parent populations based upon respective differences between the geometric means;
- determining the respective scores of media content items based upon the respective offsets of the parent populations;
- and
- ranking the media content items based at least in part on the score;
- and
- accumulating the media content items belonging to the respective parent populations into an ordered list based at least in part on the ranking.
9 more elements
Dependencies are inferred from the claim text.