Hierarchical multisource playlist generation

• US 10,242,098 B2
• Filed: 05/31/2016
• Issued: 03/26/2019
• Est. Priority Date: 05/31/2016
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Koenigstein, Noam, Nice, Nir, Elazar, Shay Ben, Berezin, Yehiel, Barkan, Oren, Zaccai, Tal, Shlevich, Shimon, Simhon, Nimrod Ben, Nogues, Paul, Lavee, Gal
• Primary Examiner(s): Chbouki, Tarek
• Application Number:
  US15/169,305
• Publication Number:
  US 20170344635A1
• Time in Patent Office: 1,029 Days
• Field of Search: None
• CPC Class Codes: G06F 16/4387 by the use of playlists
• CPC Class Codes: G06F 16/639 using playlists
• CPC Class Codes: G06F 16/68 Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually

Abstract

A playlist generator that utilizes multiple data sources to rank each track within a set of candidate tracks to enable selection of candidate tracks according to the ranking. Candidate tracks are each scored according to one or more features, such as acoustic similarity and/or similar usage patterns of the candidate track or artist of the candidate track to a current or previously played track or artist. Each feature is weighted according to historical listening patterns surrounding a user-selected playlist seed artist. The weighting may also be further corrected according to historical listening patterns of the particular user. When historical usage data related to a particular seed artist is limited, more generalized historical usage data related to a higher level in a genre hierarchy may be used.

Claims

1. A method implemented by a computing system for generating a playlist, the method comprising:

the computing system receiving a playlist seed;

the computing system generating a feature vector for each of a plurality of candidate tracks, each feature vector having one or more features each corresponding to respective feature dimensions;

the computing system, for each feature vector, adjusting the weight of at least one feature of the feature vector to form a dimensionally weighted feature vector, the adjustment being performed according to training data corresponding to the playlist seed, wherein the training data corresponding to the playlist seed is configured according to a genre hierarchy having a plurality of hierarchy levels, the training data being based on historical usage data of a lowest hierarchy level within which the playlist seeds resides and from which there is sufficient historical usage data;

the computing system ranking the plurality of candidate tracks according to the dimensionally weighted feature vectors of the plurality of candidate tracks to form a candidate track ranking; and

the computing system selecting one or more tracks according to the candidate track ranking to generate a playlist.

2. The method of claim 1, wherein adjusting the weight of at least one feature of each feature vector further includes adjusting according to training data further associated with individual user activity.

3. The method of claim 2, wherein the training data associated with individual user activity represents a set of corrective weighting factors for tailoring the relative weights of the features of the feature vectors according to historical listening patterns of a corresponding user.

4. The method of claim 2, wherein the computer-executable instructions are also structured to cause the computing system to track user activity with respect to the generated playlist to generate additional training data for updating one or more of the training data associated with the playlist seed or the training data associated with individual user activity.

5. The method of claim 1, wherein each feature vector includes a plurality of features.

6. The method of claim 5, wherein the plurality of features of a corresponding candidate track include one or more features indicating acoustic similarity between one or more of the candidate track, an artist of the candidate track, or an album of the candidate track to one or more of a previously played track, an artist of a previously played track, an album of a previously played track, or the playlist seed.

7. The method of claim 5, wherein the plurality of features of a corresponding candidate track include one or more features indicating crowd-sourced usage similarity between one or more of the candidate track, an artist of the candidate track, or an album of the candidate track to one or more of a previously played track, an artist of a previously played track, an album of a previously played track, or the playlist seed.

8. The method of claim 5, wherein the plurality of features of a corresponding candidate track include one or more features indicating popularity of the candidate track.

9. The method of claim 1, wherein the training data associated with the playlist seed represents a set of corrective weighting factors for tailoring the relative weights of the features of the feature vectors according to historical listening patterns of a plurality of users with respect to the playlist seed.

10. The method of claim 1, wherein the playlist seed is an artist.

11. The method of claim 1, wherein the training data associated with the playlist seed is configured according to the genre hierarchy such that as additional usage data is received at one or more hierarchy levels lower than a current hierarchy level source, a progressively greater proportion of the training data is based on the additional usage data added to the relatively lower hierarchy level.

12. The method of claim 1, wherein selecting one or more tracks according to the candidate track ranking includes selecting one or more tracks according to one or more selectivity parameters, the one or more selectivity parameters controlling a strictness factor by which track selection probabilities align with the candidate track ranking.

13. A computing system configured for generating a playlist, the computing system comprising:

one or more processors;

one or more computer-readable media having thereon computer-executable instructions that are structured such that, when executed by the one or more processors, cause the computing system to perform at least the following in response to receiving a playlist seed;

generate a feature vector for each of a plurality of candidate tracks, each feature vector having one or more features each corresponding to respective feature dimensions;

for each feature vector, adjust the weight of at least one feature of the feature vector to form a dimensionally weighted feature vector, the adjustment being performed according to training data corresponding to the playlist seed, the training data being based on historical usage data of a genre hierarchy according to a lowest hierarchy level within which the playlist seed resides and from which there is sufficient historical usage data;

rank the plurality of candidate tracks according to the dimensionally weighted feature vectors of the plurality of candidate tracks to form a candidate track ranking; and

select one or more tracks according to the candidate track ranking to generate a playlist.

14. The computing system of claim 13, wherein the training data is configured according to the genre hierarchy such that as additional usage data is received at one or more hierarchy levels lower than a current hierarchy level source, a progressively greater proportion of the training data is based on the additional usage data added to the relatively lower hierarchy level.

15. The computing system of claim 13, wherein the adjustments to form the dimensionally weighted feature vectors are also performed according to training data associated with individual user activity.

16. The computing system of claim 15, wherein selecting one or more tracks according to the candidate track ranking includes selecting one or more tracks according to one or more selectivity parameters, the one or more selectivity parameters controlling a strictness factor by which track selection probabilities align with the candidate track ranking.

17. A computer program product comprising one or more hardware storage devices having stored computer-executable instructions which are executable by one or more processors of a computing system for causing the computing system to perform the following in response to the computing system receiving a playlist seed:

generate a feature vector for each of a plurality of candidate tracks, each feature vector having one or more features each corresponding to respective feature dimensions;

for each feature vector, adjust the weight of at least one feature of the feature vector to form a dimensionally weighted feature vector, the adjustment being performed according to training data corresponding to the playlist seed, the training data being based on historical usage data of a genre hierarchy according to a lowest hierarchy level within which the playlist seed resides and from which there is sufficient historical usage data;

rank the plurality of candidate tracks according to the dimensionally weighted feature vectors of the plurality of candidate tracks to form a candidate track ranking; and

select one or more tracks according to the candidate track ranking to generate a playlist.

18. The computer program product of claim 17, wherein at least one of the one or more selectivity parameters is tunable according to an exploratory/conservative scale.

19. The computer program product of claim 17, wherein the adjustments to form the dimensionally weighted feature vectors are also performed according to detected training data associated with individual user activity.