RECOMMENDER SYSTEM FOR IDENTIFYING A NEW SET OF MEDIA ITEMS RESPONSIVE TO AN INPUT SET OF MEDIA ITEMS AND KNOWLEDGE BASE METRICS

• US 20100161595A1
• Filed: 01/11/2010
• Published: 06/24/2010
• Est. Priority Date: 02/03/2005
• Status: Active Grant

• Current Assignee: Apple Inc.
• Original Assignee: Strands Incorporated (CRIBIS Holding SRL)
• Inventors: Martin, Francisco J., Shur, Jim, Torrens, Marc
• Granted Patent: US 8,312,017 B2
• Time in Patent Office: Days
• US Class Current: 707/722
• CPC Class Codes: G06F 16/435 Filtering based on additional data, e.g. user or group profiles
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0631 Item recommendations
• CPC Class Codes: H04N 21/466 Learning process for intelligent management, e.g. learning user preferences for recommending movies services using the results of monitoring in broadcast systems H04H60/61
• CPC Class Codes: H04N 21/4668 for recommending content, e.g. movies
• CPC Class Codes: H04N 21/4825 using a list of items to be played back in a given order, e.g. playlists
• CPC Class Codes: H04N 7/163 by receiver means only

Abstract

Systems and methods are disclosed for identifying a new set of media items in response to an input set (or “query set”) of media items and knowledge base metrics. The system uses a knowledge base consisting of a collection of mediasets. Various metrics among media items are considered by analyzing how the media items are grouped to form the mediasets in the knowledge base. Such association or “similarity” metrics are preferably stored in a matrix form that allows the system to efficiently identify a new set of media items that complements the input set of media items.

Claims

1. A computer-implemented method for dynamically generating an output set of media items responsive to an input set of media items comprising the steps of:

(a) receiving an input set comprising at least one input media item;

(b) accessing a knowledge base comprising a plurality of collected mediasets, each knowledge base mediaset comprising at least one media item;

(c) for an input media item, computing a corresponding value of a selected metric relative to each one of a plurality of media items in the knowledge base;

(d) responsive to the computed metric values, selecting at least one of the knowledge base media items as a candidate output media item;

(e) selecting among the candidate output media items to form the output set of media items; and

(f) delivering the output set of media items to a user.

2. A method according to claim 1 wherein the input set of media items comprises first meta data identifying songs.

3. A method according to claim 2 wherein the output set of media items comprises second meta data identifying songs.

4. A method according to claim 1 wherein:

said computing step includes computing a corresponding metric value for each input item relative to each one of a plurality of media items in the knowledge base by applying a co-concurrency metric responsive to a number of times that the respective input media item appears together with the corresponding media item in the same mediaset in the knowledge base; and

said selecting a candidate output media item includes selecting a media item for which the computed metric value exceeds a predetermined threshold metric value.

5. A method according to claim 1 wherein:

the knowledge base mediasets are ordered and said computing step includes computing a corresponding metric value for each input item relative to each one of a plurality of media items in the knowledge base by applying at least one of a pre-concurrency metric and a post-concurrency metric; and

said selecting a candidate output media item includes selecting a media item for which the computed metric value exceeds a predetermined threshold metric value.

6. A method according to claim 1 wherein:

said selecting among the candidate output media items to form the output set of media items includes applying a predetermined selection criterion responsive to the media item distribution from which the candidate output media items were selected.

7. A method according to claim 1 wherein:

the input set of media items comprises first meta data identifying songs;

the knowledge base comprises a plurality of collected mediasets of songs; and

the output set of media items comprises second meta data identifying songs.

8. A method according to claim 7 wherein said delivering the output set of media items to a user comprises displaying or transmitting the second meta data to the user.

9. A method according to claim 8 and further comprising filtering the selected candidate output media items to apply a predetermined additional external constraint to form the output set of media items.

10. A method according to claim 9 wherein said filtering step includes maximizing the sum of the selected metric values among each pair of media items included in selecting the output set.

11. A method according to claim 9 wherein said filtering step includes applying a predetermined preference or criterion selected by the user.

12. A method for dynamically generating an output set of media items responsive to an input set of media items comprising the steps of:

(a) receiving an input set comprising at least one input media item;

(b) accessing a knowledge base comprising a plurality of collected mediasets, each knowledge base mediaset comprising at least one media item;

(c) for an input media item, computing a corresponding value of a selected metric relative to each one of a plurality of media items in the knowledge base;

(d) selecting among the knowledge base media items the items having the highest computed metric values as candidate output media items; and

(e) selecting among the candidate output media items to form the output set of media items.

13. A method according to claim 12 wherein the selected metric comprises one of pre-concurrency metric, a post-concurrency metric and a co-concurrency metric.

14. A method according to claim 12 wherein:

the input set of media items comprises first meta data identifying songs;

the knowledge base comprises a plurality of collected mediasets of songs; and

the output set of media items comprises second meta data identifying songs.

15. A method according to claim 14 and further comprising filtering the selected candidate output media items to form the output set of media items.

16. A method according to claim 15 wherein said filtering step applies a filter selected by a user to limit the output set of media items to a determined type of items.

17. A method according to claim 15 wherein said filtering step applies a filter selected by a user to limit the output set of media items to items that meet defined keyword-based meta data criteria, thereby personalizing the recommendations.

18. A method of pre-processing a knowledge base of mediasets to quantify an implicit association between pairs of media items in the knowledge base, each mediaset comprising a plurality of media items, and the method comprising the steps of:

for each item in a mediaset, identifying every other item in the same mediaset, so as to define all of the pairs of media items in that mediaset;

repeating said identifying step for every other mediaset in the knowledge base, thereby defining every pair of items that appears in any mediaset in the knowledge base; and

for each identified pair of media items, adjusting a concurrency metric value responsive to each additional occurrence of the same pair of items in the knowledge base.

19. A method according to claim 18 wherein the concurrency metric value is a co-concurrency metric value, and said adjusting step comprises incrementing the co-concurrency metric value responsive to every additional occurrence of the same pair of items in the knowledge base.

20. A method according to claim 18 wherein:

the media items are ordered within the mediasets in the knowledge base;

said identifying step includes defining ordered pairs of items that appear in any mediaset in the knowledge base in the same order;

the concurrency metric value is an ordered concurrency metric value, and said adjusting step comprises incrementing the ordered concurrency metric value responsive to every additional occurrence of the same ordered pair of items in the knowledge base.