RECOMMENDATION DIVERSITY

• US 20080104111A1
• Filed: 10/27/2006
• Published: 05/01/2008
• Est. Priority Date: 10/27/2006
• Status: Active Grant

• Current Assignee: R2 Solutions LLC (Acacia Research Corporation)
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: White, William, Slaney, Malcolm
• Granted Patent: US 7,860,862 B2
• Time in Patent Office: Days
• US Class Current: 707/104.1
• CPC Class Codes: G06F 16/40 of multimedia data, e.g. slideshows comprising image and additional audio data retrieval of still image data G06F16/50; retrieval of audio data G06F16/60; retrieval of video data G06F16/70
• CPC Class Codes: G06F 16/634 Query by example, e.g. query by humming
• CPC Class Codes: G06F 16/683 using metadata automatically derived from the content
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: Y10S 707/916 Audio
• CPC Class Codes: Y10S 707/957 Multidimensional

Abstract

Recommendation systems and methods are disclosed that objectively determine similarities between products and quantify diversity between products for use in generating recommendations. The product interests, such as musical interests, of a user are measured based on objective characteristics of the product. Then the interests are modeled by a distribution. The resulting distribution is then used as a measure of the diversity of the user'"'"'s tastes. Based on the diversity and the characteristics of other products, recommendations are then made to the user. The systems and methods may also utilize subjective information as a secondary filter to add or remove products for which such data is known.

Claims

1. A method for recommending songs to a listener comprising:

receiving a list associated with the listener, the list identifying a plurality of first songs;

retrieving music data for each of the first songs;

analyzing the music data to create a multidimensional diversity measure of the list; and

selecting at least one second song for recommendation to the listener based on the multidimensional diversity measure.

2. The method of claim 1 wherein analyzing further comprises:

evaluating a plurality of characteristics of the music data for each of the first songs;

determining a multidimensional distribution of the first songs based on the characteristics of the music data for each of the first songs; and

modeling the multidimensional distribution to obtain the multidimensional diversity measure of the list.

3. The method of claim 2 wherein modeling further comprises:

mapping each of the plurality of first songs to a corresponding first location in a multidimensional space to determine a scatter, each dimension of the multidimensional space associated with at least one of the plurality of characteristics; and

fitting the scatter to a multidimensional figure in the multidimensional space.

4. The method of claim 3 further comprising:

selecting at least one second song, each second song when evaluated and mapped is contained within the multidimensional figure.

6. The method of claim 3 further comprising:

retrieving song data for a plurality of third songs including a location in the multidimensional space for each third song; and

selecting the at least one second song from the plurality of third songs.

8. The method of claim 1 wherein receiving further comprises:

receiving a plurality of lists associated with the listener, at least one list identifying first songs selected by the listener and at least one other list identifying first songs selected by someone other than the listener.

9. The method of claim 1 further comprising:

evaluating the plurality of characteristics for each of a plurality of test songs, each test song associated with a subjective classification;

determining a multidimensional distribution of the test songs based on the characteristics; and

selecting the characteristics to be used as dimensions of the multidimensional space based on the subjective classifications of the test songs and the multidimensional distribution of the test songs.

5. (canceled)

7. (canceled)

10. A method of recommending songs based on a diversity of songs in a playlist comprising:

receiving the playlist associated with the listener, the playlist identifying a plurality of first songs selected by the listener;

evaluating a plurality of characteristics for each of the plurality of first songs;

mapping each of the plurality of first songs to a corresponding first location in a multidimensional space to determine a scatter, each dimension of the multidimensional space associated with a characteristic;

fitting the scatter to a multidimensional figure in the multidimensional space, the multidimensional figure defining a volume indicative of the diversity of the playlist; and

selecting at least one second song based on the diversity of the playlist.

11. The method of claim 10 further comprising:

evaluating a plurality of characteristics for each of a plurality of third songs;

mapping each of the plurality of third songs to a corresponding third location in the multidimensional space;

identifying as second songs the third songs contained within the multidimensional figure; and

recommending at least one second song to the listener.

12. The method of claim 10 further comprising:

receiving a diversity selection; and

changing, based on the diversity selection, the volume of the multidimensional figure, thereby changing which third songs are identified as second songs.

16. The method of claim 11 further comprising:

retrieving song data for the plurality of third songs including a third location in the multidimensional space for each third song.

17. The method of claim 12 wherein changing further comprises:

identifying a center of the multidimensional figure in the multidimensional space; and

changing, based on the diversity selection, the volume of the multidimensional figure about the center.

13-15. -15. (canceled)

18. A system for recommending products to a consumer comprising:

an analysis module adapted to evaluate a plurality of objective characteristics for each of one or more seed products associated with the consumer;

a mapping module adapted to model the plurality of objective characteristics of each of the one or more seed products to a corresponding location in a multidimensional space to create a scatter; and

a selection module adapted to select one or more second products from a set of third products based on the scatter.

19. The system of claim 18 further comprising:

a fitting module adapted to fit a multidimensional figure to the scatter, the multidimensional figure defining a volume in the multidimensional space, the volume indicative of a diversity of seed products in the list.

20. The system of claim 19 wherein the selection module is further adapted to identify as a second products any of the set of third products that, when evaluated and mapped to the multidimensional space, are within the multidimensional figure.

21. The system of claim 20 wherein the fitting module is further adapted to receive a diversity selection and change the volume of the multidimensional figure based on the diversity selection, thereby changing which of the set of third products that, when evaluated and mapped to the multidimensional space, are within the multidimensional figure.

22. The system of claim 18 wherein the analysis module is further adapted to receive the list of one or more seed product associated with the consumer.

23. The system of claim 18 further comprising:

a product datastore containing data associated with each of the third products.

24. The system of claim 23 wherein the data includes a location in multidimensional space corresponding to each of the third products.

25. A method of recommending products based on a diversity of products in a list comprising:

receiving the list associated with a consumer, the list identifying a plurality of first products selected by the consumer;

evaluating a plurality of characteristics for each of the plurality of first products;

mapping each of the plurality of first products to a corresponding first location in a multidimensional space to determine a scatter, each dimension of the multidimensional space associated with a characteristic;

fitting the scatter to a multidimensional figure in the multidimensional space, the multidimensional figure defining a volume indicative of the diversity of the list; and

selecting at least one second product based on the diversity of the list.

26. The method of claim 25 further comprising:

evaluating a plurality of characteristics for each of a plurality of third products;

mapping each of the plurality of third products to a corresponding third location in the multidimensional space;

identifying as second products any third products contained within the multidimensional figure; and

recommending at least one second product to the consumer.

27. The method of claim 25 further comprising:

receiving a diversity selection; and

changing, based on the diversity selection, the volume of the multidimensional figure, thereby changing which third products are identified as second products.

30. The method of claim 26 wherein fitting further comprises:

fitting the scatter to a multidimensional ellipsoid.

31. The method of claim 26 further comprising:

retrieving product data for the plurality of third products including a third location in the multidimensional space for each third product.

32. The method of claim 27 wherein changing further comprises:

identifying a center of the multidimensional figure in the multidimensional space; and

changing, based on the diversity selection, the volume of the multidimensional figure about the center.

28-29. -29. (canceled)