US20070244880A1

US20070244880A1 - Mediaset generation system

Info

Publication number: US20070244880A1
Application number: US11/513,495
Authority: US
Inventors: Francisco Martin; Marc Torrens; Rick Hangartner; Guillermo Caudevilla-Laliena; David Bartolome; Craig Rowley
Original assignee: MYSTRANDS Inc
Current assignee: Apple Inc
Priority date: 2006-02-03
Filing date: 2006-08-31
Publication date: 2007-10-18
Also published as: US8583671B2; US20090210415A1

Abstract

Disclosed are various embodiments of systems and methods for generating composite mediasets from mediasets, each comprising media items, associated with a plurality of users. In some embodiments, individual and/or group recommendations are provided for creating a group playlist by aggregating user taste data for a plurality of users in a group. In other embodiments, systems and methods are provided which allow for sharing and playing of a group playlist by users in a group, each of which has a media playback device. In still other embodiments, systems and methods are provided for dynamically building a composite playlist that may be updated as users enter and/or leave the group.

Description

RELATED APPLICATIONS

This application is a continuation-in-part of application Ser. No. 11/346,818 filed Feb. 3, 2006 and titled “Recommender System for Identifying a New Set of Media Items Responsive to an Input Set of Media Items and Knowledge Base Metrics.” This application claims the benefit under 35 U.S.C. § 119(e) of U.S. Provisional Patent Application No. 60/772,502 filed Feb. 10, 2006, and titled “System and Method for Building and Sharing a Composite Playlist from Collective Group Tastes on Multiple Media Playback Devices.” This application also claims the benefit under 35 U.S.C. § 119(e) of U.S. Provisional Patent Application No. 60/774,072 filed Feb. 15, 2006, and titled “Mediaset Recommendations for a Group of Users.” This application also claims the benefit under 35 U.S.C. § 119(e) of U.S. Provisional Patent Application No. 60/796,724 filed May 1, 2006, and titled “Dynamically Building Composite Playlist for Merging Collective User Tastes.” Each of the foregoing applications is incorporated herein by specific reference.

BRIEF DESCRIPTION OF THE DRAWINGS

Understanding that drawings depict only certain preferred embodiments of the invention and are therefore not to be considered limiting of its scope, the preferred embodiments will be described and explained with additional specificity and detail through the use of the accompanying drawings in which:
FIG. 1 is a diagram of one embodiment of a system for synthesizing a user's taste by analyzing the user's playlist(s) and playcounts.
FIG. 2 is a diagram of one embodiment of a system for using taste analyses from a plurality of users to generate a plurality of recommended mediasets.
FIG. 3 is a diagram of one embodiment of a system for using recommended mediasets from a plurality of users to generate a recommended mediaset for a group of users.
FIG. 4 is a system diagram of a client-server embodiment of the invention.
FIG. 5 is a system diagram of a peer-to-peer embodiment of the invention.
FIG. 6 depicts the main functional components of a Device Manager according to one embodiment.
FIG. 7 depicts a Member Status screen of a Session Manager, which presents the relevant details of the data structures in the Session Manager subcomponent of the Device Manager of one embodiment.
FIG. 8 is a top-level flow diagram of the operation of the Session Manager subcomponent of the Device Manager of one embodiment.
FIG. 9 is a more detailed flow diagram of the “Prune Session Members” step in the Session Manager flow diagram of FIG. 8.
FIG. 10 is a more detailed flow diagram of the “Update Session Members” step in the Session Manager flow diagram of FIG. 8.
FIG. 11 is a more detailed flow diagram of the “Service Proxy Requests” step in the Session Manager flow diagram of FIG. 8.
FIG. 12 depicts a Playlist Queue screen of a Playlist Manager, which presents the relevant details of the data structures in the Playlist Manager subcomponent of the Device Manager of one embodiment.
FIG. 13 is a flow diagram showing the operation of the Playlist Manager subcomponent of the Device Manager of one embodiment.
FIG. 14 is a flow diagram showing the operation of the Playlist Play subcomponent of the Playlist Manager of a Media Player Client embodiment.
FIG. 15 depicts the relevant data structures in one embodiment of the Media Player Client supporting the Playlist Builder component of the system.
FIG. 16 presents a conceptual diagram of the playlist building function of one embodiment.
FIG. 17 is a flow diagram of the operation of one embodiment of the Playlist Builder component of the system.
FIG. 18 is a diagram of a process for generating a list of media items that define a user taste.
FIG. 19 is a diagram of a process for computing aggregate playlist goals and a list of desired categories.
FIG. 20 is a diagram of a process for computing relevant media items for each user from total available media items.
FIG. 21 is a diagram of a process for categorizing relevant media items according to desired attributes.
FIG. 22 is a diagram of a process for computing the status of the current group playlist's achievement of compositional goals.
FIG. 23 is a diagram of a process for choosing a set of media items to add to a group playlist based on assessing needs according to compositional goal achievement assessment.
FIG. 24 represents a step of a user addition process for detecting new users.
FIG. 25 represents another step of a user addition process in which new users are added to the system and source device information is stored.
FIG. 26 is a diagram of a process for removing users from the system.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

In the following description, certain specific details of programming, software modules, user selections, network transactions, database queries, database structures, etc., are provided for a thorough understanding of the specific preferred embodiments of the invention. However, those skilled in the art will recognize that embodiments can be practiced without one or more of the specific details, or with other methods, components, materials, etc.
In some cases, well-known structures, materials, or operations are not shown or described in detail in order to avoid obscuring aspects of the preferred embodiments. Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in a variety of alternative embodiments. In some embodiments, the methodologies and systems described herein may be carried out using one or more digital processors, such as the types of microprocessors that are commonly found in PC's, laptops, PDA's and all manner of other desktop or portable electronic appliances.
Disclosed are embodiments of systems, methods, and apparatus for generating mediasets comprising a plurality of media data items. As used herein, the term “media data item” is intended to encompass any media item or representation of a media item. A “media item” is intended to encompass any type of media file which can be represented in a digital media format, such as a song, movie, picture, e-book, game, etc. Thus, it is intended that the term “media data item” encompass, for example, playable media item files (e.g., an MP3 file), as well as metadata that identifies a playable media file (e.g., metadata that identifies an MP3 file). It should therefore be apparent that in any embodiment providing a process, step, or system using “media items,” that process, step, or system may instead use a representation of a media item (such as metadata), and vice versa.
In one embodiment, a system to provide recommendations of mediasets for a given group of users is provided. Embodiments of such a system may comprise a mechanism to store playlists and/or playcounts of each member in a community of users. Playlists and playcounts may be used to define the taste of each user and may therefore be used in performing taste analyses for each of the respective users.
In accordance with the general principles set forth above, embodiments of the invention may be used to address the problem of recommending a mediaset or group playlist to a group of users in a community or network. In some embodiments, a mediaset recommender may be provided where the input is a set of media items, and the output is a mediaset of weighted media items. Two illustrative methods for providing group recommendations of mediasets include: 1) building a common profile that expresses the taste of the group of users as a whole and applying that profile to the recommender; and 2) considering individual recommendations of each member taste, and aggregating the results.
Media players are typically capable of reproducing all types of media items and collecting playcounts and playlists. Playcounts are the number of times a media item has been played in the media player. Playlists are groupings of media items that users create to organize their libraries of media items. A system (e.g., a server) may be used to collect playcounts and playlists of media items of a community of users.
Playcounts and playlists of a user may be used to synthesize her or his taste or perform a taste analysis. In that sense, user's taste may be considered a collection of the most relevant taste data considering that user's playcounts and playlists.
The task of certain embodiments of this system is to recommend a mediaset for a group of users. A mediaset recommendation for a group of users may be the result of an aggregation process of the different mediasets that are recommended to each user of the group. Thus, some embodiments of this system may include a component that recommends a mediaset from another mediaset (by performing an analysis of the taste of each user, for example). The aggregation process may apply, for example, a voting schema and/or an optimization schema.
Mediaset recommendations for a group of users may be useful in a variety of scenarios. An example is a party where a group of people want to enjoy music together. Instead of playing the music that may be recommended to a particular individual in the group, it may be desirable to play the music that would be recommended to the group as a whole.
As such, in certain embodiments, the system's task is to find a mediaset or playlist to be recommended to a group of users. In embodiments wherein the system is collecting playcounts and/or playlists from the users, such a recommender system may be composed of three main steps:

1) Synthesizing user tastes;
2) Producing recommended mediasets for each user taste; and
3) Aggregating the set of recommended mediasets into a single mediaset to be recommended to the whole group of users.

It should be understood that numerous variations on the content, scope, and order of these steps are contemplated. For example, the step of producing recommended mediasets for each user taste may be optional.
FIG. 1 depicts how the taste 108 of a user 100 may be built from the playcounts 104 and playlists 102 of that user 100 through a process 106 called Synthesize Taste. It should be understood that process 106 is one example of a taste analysis. Process 106 produces a taste 108, which may be encoded as a mediaset. As part of process 106, media items with higher playcounts and/or those that appear more in playlists may be selected. The selected media items may include those that are played more often by the user and/or those that are used in more different contexts and situations (because they appear in many different playlists). In some embodiments, playcounts may also be associated with play dates. In such cases, the process may weigh the playcounts with the play dates and therefore give more relevance to recent playcounts than to older playcounts.
Process 106 may comprise a ranking process of the media items of a user where items with higher playcounts, more recent plays, and/or more playlist appearances get a higher ranking. The process may select the top m ranked media items as the encoding of the user's taste 108. Note that this process may produce different results over time for the same user. This may be a desirable feature in embodiments in which the goal is to encode the taste of a user as it evolves over time.
In some embodiments, the system may provide individual mediaset recommendations. For example, with reference to FIG. 2, given a group of n users 200, 208, and 216, the system may compute the individual taste (202, 210, and 218, respectively) for each user with, for example, the process described above. The system may provide a recommended mediaset for each user using a recommender engine (204, 212, and 220, respectively) that processes user tastes ( mediasets 202, 210, and 218) to produce recommended mediasets 206, 214, and 222. It should be understood that a separate recommended engine may be used for each user ( recommender engines 204, 212, and 220), as shown in FIG. 2 or, alternatively, a single recommender engine may be used for the entire group. Thus, in total the depicted system may compute n mediasets of m media items.
In some embodiments, the system may also provide for aggregating individual recommended media sets into a group recommended mediaset or group playlist. For example, with reference to FIG. 3, once the system has computed the n mediasets of m media items (one for each user 300, 304, 308 of the group), an aggregation step may be used to produce the final mediaset of p (p≦n*m) media items 314 to be recommended to the group of users. As shown in FIG. 3, the system may provide an aggregator 312, which takes as input a recommended mediaset (302, 306, and 310, respectively) from each of the users in the group and provides therefrom a group recommended mediaset 314.
The aggregation step may be performed, for example, by following different approaches that serve different goals. As previously mentioned, in two preferred embodiments, the system may follow a) a voting schema; or b) an optimization schema. A voting schema may serve the goal of finding a mediaset that the majority of users would be happy with, without considering the degree of dislike by the rest of the members of the group. On the other hand, an optimization schema may produce a mediaset that minimizes the dislike (or maximizes the like) of all the members of the group. In order to apply an optimization schema, the media items in the n recommended mediasets 302, 306, 310 may each be linked with an associated weight. The weight of a media item in a mediaset for a user may be used to indicate the relevance of that media item for the user.
With a voting schema, the aggregation process may take the p media items that appear the most in the n mediasets. If there are items that appear the same number of times in the n mediasets (tie-break), then those items may be picked randomly.
Some recommender engines may produce mediasets having weighted media items. In such embodiments, when a tie-break situation happens, instead of picking the items randomly, the process may pick the media items with highest weights. For example, considering the following mediasets:

msl={s3,s7,s8,s10}
ms2={s2,s3,s4,s10}
ms3={s3,s4,s7}
then, following the voting schema described above, the media items (s#) in the mediasets (ms#) would be ranked as follows:
s1=0, s2=1, s3=3, s4=2, s5=0, s6=0, s7=2, s8=1, s9=0, s10=2

Thus, the media items for the recommended group mediaset would be selected in the following order: s3, s4, s7, s10, s2, s8. Media items s7 and s10 are in a tie-break situation, so they could be ordered in accordance with their respective weights within their mediasets, if any. If the media items do not have associated weights, then the order of s7 and s10 may be randomized. The same would apply for items s2 and s8.
With an aggregation schema, the aggregation process may take the p media items that optimize some utility function considering all the users of the group, that is, considering all n mediasets 302, 306, 310. In order to apply the optimization schema, the media items in the n mediasets 302, 306, 310 may have an associated weight, for example, in the range from 0 to 1, where 0 means that the item is not relevant at all and 1 means that the item is the most relevant. For a given mediaset j, a media item i may therefore have a weight w(j,i). If a media item i is not in a mediaset j, then it may be considered to have a weight 0. The following example illustrates the weights associated with media items for a collection of mediasets:

ms1={s3,s7,s8,s10}
ms2={s2,s3,s4,s7, s10}
ms3={s3,s4,s7}
w(1)=[0, 0, 0.1, 0, 0, 0, 0.3, 0.2, 0, 0.9]
w(2)=[0, 0.2, 0.5, 0.4, 0, 0, 0.3, 0, 0, 0.9]
w(3)=[0, 0, 0.1, 0.8, 0, 0, 0.3, 0, 0, 0.5]

A number of different utility functions may be chosen in order to aggregate the media items. For example, a utility function may be selected to maximize the sum for all p selected items of the highest weight in any of the n mediasets. If it is desired to select p=3 items in accordance with this utility function, s10, s4, and s3 would be selected. The sum of the highest weights for these items is 0.9+0.8+0.5=2.2, which is the maximum we can get with the above example.
Alternatively, a utility function may be selected to maximize the sum for all p selected items of the lowest weight in any of the n mediasets. If it is desired to select p=3 items in accordance with this utility function, s10, s7, and s3 would be selected. The sum of the lowest weights for these items is 0.5+0.3+0.1=0.9, which is the maximum we can get with the above example.
As still another alternative, a utility function may be selected to maximize the sum for all p selected items of the mean weight of all of the n mediasets. If it is desired to select p=3 items in accordance with this utility function, s10, s4, and s7 would be selected. The sum of the average of weights for these items is 2.3/3+1.2/3+0.9/3=1.16, which is the maximum we can get with the above example. Of course, other utility functions may be employed, as will be apparent to one having ordinary skill in the art.
A recommender engine may be provided in some embodiments. In embodiments that do not include a recommender engine, the mediasets that encode the user tastes may be directly aggregated to form the mediaset that would be recommended to the whole group of users.
A system using the recommender engine may propose mediasets to discover new music, whereas a system that does not provide a recommender engine may propose mediasets with media items already known by at least one of the users in the group.
As should be apparent, the aforementioned systems and methods may produce mediaset recommendations for a group of n users so as to enable proposing a mediaset that can be enjoyed simultaneously by a group of users. The system may analyze user tastes from playlists and/or playcounts, so as to allow the user tastes to be represented as mediasets. These n user tastes can be then the input of a recommender engine that may suggest another n mediasets. An aggregation process that takes these mediasets and produces a single group mediaset may also be provided. Such a process can be done using, for example, a voting schema or an optimization schema. Similar systems may operate without a recommender engine. In such embodiments, the aggregation process may operate with the mediasets that represent the tastes of the n users, and the result may comprise a mediaset that can be recommended to the whole group of users.
Additional embodiments are disclosed and described with reference to FIGS. 4-14. In some of these embodiments, composite playlists may be built of media items in individual libraries on multiple media devices in a common location and/or through a communication network. The media devices in the group may include portable and/or non-portable devices. A means for playing the composite playlist may step through the playlist in sequence and cause the media player with the indicated media item to stream it to one or more of the other players where possible. Such embodiments may operate under, for example, a client-server architecture or a P2P distributed architecture. These embodiments may therefore be used to coordinate a set of media players to transmit media items on the specified group playlist in streaming fashion to each of the individual media players, or to a specified subset of the media players, in the sequential order they are listed on the group playlist. This may be accomplished without transferring standalone copies of the media items between the various media players.
Those skilled in the art will recognize that systems incorporating the features of one or more of the above-described embodiments may be realized as a collection of media devices whose design embodies the disclosed functional behavior, as a collection of layered protocols in the 7-level ISO Open Systems Interconnection Reference Model, or as an application task in media devices communicating using standard networking protocols.
Some preferred implementations will consist of three major components. The first component is a plurality of Session Managers which collectively coordinate the information needed to designate and manage the status of each media player with regard to a plurality of other media player devices engaged in a period of collaborative activity referred to as a “session”. In a preferred implementation, one Session Manager may be associated with each media player device, although this need not be the case in all implementations. The Session Managers may include means for verifying the eligibility of each device to join the session. This may be accomplished by virtue of the eligible users being subscribed to a service, which may provide legal access to the media items to be collectively enjoyed.
The second component of the aforementioned implementation is a Playlist Builder, which may reside, for example, on one of the media players, on a server, on a network access controller included in the system, or on a third-party server accessible to the media player devices through a communication network. The Playlist Builder may use information, such as taste data, available about the users in the group, and the collective set of media items available to the media players, to build a group playlist compatible with the collective tastes of the group.
The third component of the aforementioned implementation is a plurality of Playlist Managers, one associated with each media player device, which collectively communicate with the Playlist Builder to provide the information needed to build the group playlist and to play the media items on the group playlist. The Playlist Manager associated with each media player device may include functionality for communicating the availability of media items on the media player device, either in a local library or from a media streaming service accessible by a media player, for example. The Playlist Manager may also include a Playlist Play subcomponent, which may work in coordination with the counterpart subcomponents on the other media players in the session to step through each item on the group playlist, and may be configured to cause the media player associated with a media item to stream it to one or more of the other players as each media item is encountered.
With reference again to the drawings, further aspects of certain embodiments will now be described in greater detail. Two such alternative embodiments are shown in FIGS. 4 and 5. FIG. 4 is a system diagram of a client-server embodiment 400 and FIG. 5 is a system diagram of a peer-to-peer embodiment 500. Both embodiments of the invention may have the same essential components. However, the client-server embodiment 400 of FIG. 4 will be described in greater detail herein than the peer-to-peer embodiment. Nevertheless, It should be understood that the disclosure provided herein may apply equally to the peer-to-peer system with the appropriate system organizational differences, as depicted in the figures.
The key components of a client-server embodiment 400 are the server functional unit 402 and one or more Media Player Device clients 404. The server 402 may include three basic functional components: 1) A Session Access Controller 408, which may be used to grant permission for an individual media player to join the collaborative interaction between devices; 2) A Playlist Builder 410, may be used to construct the list and sequence of media items to be played in the group playlist during the session; and 3) A Device Manager 406 for each Media Player Device 404 in the session, which may be used to control the session-related functions of the device.
The peer-to-peer embodiment 500, as shown in FIG. 5, also includes a plurality of media player devices 504. Each of the media player devices 504 may include a device manager 506. A network access controller 502 may also be provided, which includes a session access controller 508 and a playlist builder 510 for receiving media item data from the plurality of media devices and for building a composite playlist of media data items from the media item data.
The device manager for each media player device may include a session manager and a playlist manager. For example, FIG. 6 depicts a device manager 602 having a session manager 604 and a playlist manager 606. Session manager 602 may be configured to manage the status of the media device with which it is associated.
FIG. 7 depicts a Member Status screen 702 that is associated with one embodiment of a session manager. As shown in FIG. 7, the session manager may track a series of categories of information for each of the media players in the session, such as ID, proxy, status, and time period information.
Playlist manager 606 may be configured to communicate with a playlist builder to send the media item data of the media device with which it is associated to the playlist builder and to play the media data items on the composite playlist on the media device with which it is associated.
As shown in FIG. 4 for the server-client configuration, and FIG. 5 for the peer-to-peer configuration, each media player device 404 may have an associated instance of the device manager 602 of FIG. 6. In addition, the server 402 and network access controller 502 may include a single instance of the session access controller 408, 508 and the playlist builder 410, 510.
The session access controller 408, 508 may interact, as described in greater detail later, with the session manager 604 in the device manager 602 of each media device 404 in a system to define the set of media devices comprising a session, and to enable communications between them. Similarly, the playlist builder 410, 510 may operate with the playlist manager 606 in the device manager 602 to define and perform a sequence of media items in a session.
FIG. 8 presents a top-level flow diagram of one embodiment of a session manager 604 in a device manager 602. The session manager of an individual media player device may be used to negotiate the introduction, operation, and departure of the device in a media playing session (henceforth “session”). This may include maintaining knowledge of all other media players devices in the session.
The session manager 604 may assume the presence of a single session access controller 408 on a server system 402 in the client-server configuration, or on a privileged peer-to-peer network access system 502 in the peer-to-peer configuration. In peer-to-peer embodiments, the network access system could be implemented on the system found in some peer-to-peer networks, which hosts network-level functions, while in other embodiments it could be implemented on one peer system in the network.
The session access controller's primary function in some embodiments is to serve a session sID code in response to a “request-to-enter” session message from devices seeking admission into a session as an indication permission has been granted to the device. In some embodiments, that permission may be granted if the media player mID code from the device supplied with the “request-to-enter” message is recognized as an mID eligible for admission to a session. In other embodiments, the session access controller may instead supply a dynamic mID back to the requesting media player device to serve as a unique identifier for the device in the context of the session along with the returned session sID code. In some embodiments, the session access controller may largely be functional in the networking protocol for the underlying communications network linking the media player devices, and the mID and sID may be codes derived from parameters in the network protocol that identify devices and communication sessions or transactions.
The session manager may include two major subfunctions 802 and 806, as shown in FIG. 8. Subfunction 802 mediates the admission of the media player into a session while subfunction 806 implements the process by which the media player maintains cognizance of the other media players in the session via the communication network linking.
The starting step for the session manager flow diagram of FIG. 8 assumes the media player is not participating in a session. As shown in the subfunction of 802, the session manager broadcasts a “request-to-enter” message with an optional “mID” code. If the media device successfully communicates with a session access controller that grants the media player access to a session, the media player will receive an “enter-session-sID” message indicating that the media player has been granted permission to enter the session identified by the sID code. In embodiments having only one session, the sID code would be superfluous and may be omitted or a default value returned.
If an “enter-session-sID” response message is not received, the session manager waits a random amount of time before transmitting another “request-to-enter” message. In some embodiments, the session manager may just wait a random amount of time after transmitting a “request-to-enter” message before transmitting another “request-to-enter” message if an “enter-session-sID” response is not received. In other embodiments, the session manager may wait a fixed amount of time after sending the “request-to-enter” message and then, if no “enter-session-sID” response is received, wait a random amount of time before transmitting the next “request-to-enter” message. In yet other embodiments, the session manager may wait until some external event occurs after sending the “request-to-enter” message, rather than receiving an “enter-session-sID”, and then wait a random amount of time before transmitting the next “request-to-enter” message.
Upon receipt of an “enter-session-sID” response from the session access controller, the session manager broadcasts the “in-session-mId-mId-status-sID” message, as indicated at 804. By broadcasting this message, the session manager indicates its presence in the session to the session managers in all of the other devices in the session. The parameters of this message (mID, mID, status, sID), are the mID of this media player device, the mID of another media player device in the session that knows about this device by proxy (set here to the mID of this media player device because no proxy is involved), the status of the playlist manager component 606 of the device manager 602, and the session sID.
After broadcasting the “in-session-mId-mId-status-sID” message, the session manager initiates the subfunction 806, which maintains knowledge of the other media players in the session identified by a particular sID. Subfunction 806 is a polling loop that maintains the information in a session manager state data structure, such as that shown in FIG. 7. The Device Manager 602 in each media player maintains a copy of the session manager state autonomously for each session, as identified by a unique sID, in which the media player is a participant. This state information may include several elements for each of the other media players in the session known to the subject media player: One such element is the mID of the media player. Another is the mID (referred to as the pmID) of the proxy media player in the session known to the subject media player, which knows the media player identified by mID is in the session, even if that media player is not known directly to the subject media player. A third example is the status of the media player. Still another example is the amount of time since this media player was last affirmatively known to be in the current session by the subject media player.
The polling loop in Subfunction 806 may be executed a number of times “n” as determined by the implementation. This number is relatively arbitrary, and typically is selected to achieve a desired “liveliness” criteria for the session manager and the session maintenance protocol. “Liveliness” here refers to how often the subject media player indicates its presence in the session to the other media players in the session. As the flow diagram indicates at 804, the session manager may broadcast an “in-session-mid-mid-status-sid” message to the other media players to indicate the subject media player is still active in the session and make its status known to those media players if the session manager state data table 702 includes entries for other media players. If the session manager state data table 702 is empty, the session manager may instead revert to searching for a session to join by initiating the session join subfunction 802.
Subfunction 806 may maintain knowledge of the other media players in the session by repetitively executing the three processes 810, 812, and 814 shown in FIG, 8, each of which are shown in greater detail in FIGS. 9-11, respectively. The “Prune Session Members” process 810 scans the session manager state data structure 702 and removes entries corresponding to media players from which the subject media player has not received an affirmative indication that they are still in the session, and therefore are inferred to have left the session. The “Update Session Members” process 812 processes messages from other media players that are therefore inferred to have joined or remained in the session. Finally, the “Service Proxy Requests” process 814 responds to requests from other media players if the session manager state data structure 702 includes an entry for a media player known directly to the subject media player that has been only known to another media player by proxy via the subject media player.
Describing each of the processes 810, 812, and 814 in turn, the “Prune Session Members” process 810 (FIG. 9) may perform one of the following three operations on the entry for each media player i in the session manager data structure 702.
1) The entry is removed if the time[i] that the media player was last affirmatively known to the subject media player exceeds an implementation specified timeout value. A “proxy-session-mID-sID” query is then broadcast to request if the media player corresponding to the removed entry is still known to another media player in the session.
2) A “proxy-session-mID-sID” query is broadcast to request if the media player is still known to another media player in the session, and the media player entry status has been marked as “cued” (meaning that the media player was scheduled to perform a track by the Playlist Manager process described below), and if one of the two following conditions exist: a) If the time[i] that the media player was last affirmatively known to the subject media player exceeds an implementation specified cued timeout value, meaning that the subject media player did not receive an indication that the media player corresponding to the entry transmitted a state change out of the cued state; b) If the time[i] that the media player was last affirmatively known to the subject media player does not exceed an implementation specified cued timeout value, but the media player corresponding to the entry is only known to the subject media by proxy via another media player in the session. In this case, the subject media player would only learn of a state change by the media player corresponding to the entry if that state change is broadcast by the proxy media player in response to the “proxy-session-mID-sID” query.
3) Nothing is done to the entry for the media player in the event neither of the above conditions apply.
The “Update Session Members” process 812 (FIG. 10) receives the “in-session-mID-pmID-status-sID” messages broadcast by the session manager and the “Service Proxy Requests” process 814 and updates the session manager state data structure 702. The process assumes that the protocol of the network over which the media players communicate buffers all “in-session-mID-pmID-status-sID” messages until they can be processed. Each received message in which the message sID matches the ID of the session is processed in one of the five following ways:
1) An entry corresponding to the media player with the mID of the message is added to the Session Manager state data structure if no entry with mID as the ID already exists.
2) If the mID and the pmlD of the message match, implying that this message was transmitted by a media player that is now directly known to the subject media player, the entry in the data table for the media player with mID is updated with the status from the message. The time since the media player with mID was last affirmatively known to the subject media player may be reset to 0 seconds.
3) The entry in the data table for the media player with mID may be updated with the proxy media player pmID and the status from the received “in-session-mID-pmID-status-sID” message if several conditions apply simultaneously: a) the received message is a proxy message (miID, pmID differ); b) the media player referenced by the message is currently known to the subject media by proxy (mID, pmID for the entry in the Session Manager state data structure differ) or the time since the media player with mID was last affirmatively known to the subject media player has exceeded an implementation timeout value; and c) the status of the referenced media player is unknown or the status in the message is idle. The time since the media player with mID was last affirmatively known to the subject media player may be reset to 0 seconds.
4) The entry in the data table for the media player with mID may be updated with just the proxy media player pmID from the received “in-session-mID-pmID-status-sID” message if several conditions apply simultaneously: a) the received message is a proxy message (miID, pmID differ); b) the media player referenced by the message is currently only known to the subject media by proxy (mID, pmID for the entry in the Session Manager state data structure differ) or the time since the media player with mID was last affirmatively known to the subject media player has exceeded an implementation timeout value; and c) the status of the referenced media player is known and the status in the message is not idle. The time since the media player with mID was last affirmatively known to the subject media player may be reset to 0 seconds.
5) Nothing is done if the received “in-session-mID-pmID-status-sID” is a proxy message (miID, pmID differ), the media player referenced by the message is currently known to the subject media player (mID, pmID for the entry in the session manager state data structure match), and the time since the referenced media play was last affirmatively known to the subject media player does not exceed an implementation-defined timeout value.
The “Service Proxy Request” process 814 (FIG. 11) provides information about any media players the subject media player affirmatively knows to the other media players in the session in response to the “proxy-session-mID-sID” request messages, which may have been broadcast by the “Prune Session Members” process 810 of any media player in the session identified by sID. As with the “Update Session Members” process, this process assumes that the protocol of the network over which the media players communicate buffers all “proxy-session-mID-sID” messages until they can be processed. Each received “proxy-session-mID-sID” message in which the message sID matches the ID of the session may be processed in one of the four following ways:
1) An “in session-mID-pmID-status-sID” message is broadcast, where pmID=mID and the status is that of the subject media player, if the subject media player corresponds to the mID of the request.
2) An “in session-mID-pmID-status-sID” message is broadcast, where pmID=mID, and the mID and status parameters are those in the session manager state data structure 702 for the media player referenced by the query, if the media player mID of the query is known affirmatively to the subject media player (mID of the query and pmID in the state manager state data structure match).
3) An “in session-mID-pmID-status-sID” message is broadcast, where the pmID and status parameters are those in the session manager state data structure 702 for the media player referenced by the query, if the media player mID of the query is only known to the subject media player by proxy (mID of the query and pmID in the State Manager state data structure differ).
4) Nothing is broadcast if the “proxy-session-mID-sID” message does not reference the subject media player or a media player in the Session Manager state data structure 402 known to the subject media player.
The Playlist Manager of an individual media player device may consist of two major subfunctions: First, the “Playlist Queue Updater” process 1300, as depicted in FIG. 13, which negotiates the sharing of playlist information between the media playing devices in the session. And, second, the “Playlist Play Sequencer” process 1400, as depicted in FIG. 14, which controls performance of the media items by the subject media player.
The Playlist Manager 1200 (FIG. 12) assumes the presence of a single Playlist Builder on a server system in the client-server configuration, or on a privileged peer-to-peer network access system in the peer-to-peer configuration. In peer-to-peer embodiments, the playlist builder could be implemented on the system found in some peer-to-peer networks, which hosts network-level functions, while in other embodiments it could be implemented on one peer system in the network. The “Playlist Queue Updater” may provide the data from the media player to the Playlist Builder that is used in the playlist building process. The Playlist Builder will be described in detail subsequently.
The Playlist Manager 1200 also assumes that the media player makes available several data items relevant to the playlist building process depicted conceptually in FIG. 15. In some embodiments, the media player device 1502 may store media items in a local library 1506, and provide a mechanism for querying the local directory 1508 to determine whether a specific media item is available for performance by the media player. in other embodiments, the media player 1502 may download or stream media items from a remote service over a communication network on demand, and provide a mechanism for querying a directory 1504 of media items available from the remote service for performance by the media player. FIG. 15 also depicts several other data items maintained by the media player relevant to the playlist building process, including indication 1510 of a media item that should be preemptively added to the playlist, and lists of items 1512, 1514, and 1516 of items that user does not want added, added preferentially, and added if at all possible, respectively, to the playlist.
FIG. 13 presents a top-level flow diagram of the “Playlist Queue Updater” process 1300. As the figure illustrates, this may be a repetitive process performed ad infinitum. The “Playlist Queue Updater” first determines if the user of the subject media player has requested that a particular media item be added to the playlist being built, and sends a “force-iid-pid-length-sid” message to the Playlist Builder if so. This message specifies the media item by iID, the pID of the media player (this is the same value as mID used by the Session Queue Updater described previously), the length in time units of the media item, and the sID for the session.
After any user request that a specific media item be added to the playlist has been processed, the “Playlist Queue Updater” then checks if a “have-iID-sID” request message has been received from the Playlist Builder inquiring whether the subject media player can provide a specific media item for the playlist. The process assumes that the protocol of the network over which the media players communicate buffers all “have-iiD-sID” request messages until they can be processed. Each received message in which the message sID matches the ID of the session may be processed in one of three following ways:
1) A “have-iID-pID-length-sid” message is broadcast if the subject media player pID has access to the requested media item iiD with length “length” from a local library of media items.
2) A “have-iID-pID-length-sid” message is broadcast if the subject media player pID does not have access to the requested media item iiD with length “length” from a local library of media items, but does have access to the requested media item from a remote service.
3) No response is broadcast if the media player does not have access to the requested media item.
In some embodiments, the “Playlist Queue Updater” may take into account the user's preferences with regard to media items, as indicated by the lists 1512, 1514, and 1516 in determining whether to supply a “have-iID-pID-length-sid” message in response to a “have-iID-sID” request message. For example, the “Playlist Queue Updater” may not respond to the “have-iID-sID” message even if the requested item is in the catalog 1504 or 1506 of the device if it is also on the “no play” list 1512. Similarly, the “Playlist Queue Updater” may respond optionally according to some statistical or other criteria if the item is on the “preferred” list 1514. And the “Playlist Queue Updater” may always respond if the item is on the “must play” list 1514.
The last step in an iteration of the “Playlist Queue Updater” flow diagram processes at least one “queue-iID-pID-length-sID” message from the Playlist Builder, if any have been received. The process assumes that the protocol of the network over which the media players communicate buffers all “queue-iID-pID-length-sID” messages until they can be processed. In some embodiments, the “Playlist Queue Updater” may process only a single “queue-iID-pID-length-sID” message per iteration by adding an entry to the Playlist Queue data structure in the Playlist Manager, consisting of the iID, length, and pID items from the message. In other embodiments, it may process multiple or all pending “queue-iID-pID-length-sID” messages.
In one embodiment, performance of the playlist is, in effect, directed by the Playlist Builder. As described later, the Playlist Builder may broadcast a “queue-iId-pID-length-sid” message to all the media players in the session requesting that media item be added to the Playlist Queue data structure 1202 in the Playlist Manager 1200. The Playlist Builder sends this message at the actual time the media item should be performed and the media player accepts that message as a command to perform the specified media item. In a variant of this embodiment, the Playlist Builder may send this message just sufficiently before the time the media item should be performed to allow the media item to perform any processing required to initiate the performance by the time the performance is to start.
In another embodiment, the “Playlist Play Sequencer” process 1400 of the Playlist Manager 1200 shown in FIG. 14 locally coordinates the performance of the items on the playlist that are actually performed by the subject media player with the performance of the other media items on the playlist by the appropriate media players. All media players in the session may have a copy of the Playlist Queue 1202. This copy may be imperfect: It could be missing items at the head of the queue that were put on the playlist before the subject media player joined the session, or in the middle of the queue due to communication failures. The Playlist Play Sequencer may cause the subject media player to synchronize its performances of individual media items with the performance of media items by other media players in the session in the presence of these possible gaps in the playlist.
The “Playlist Play Sequencer” 1400 may be an iterative process which achieves the synchronized performance in the presence of gaps by processing the item at the head of the Playlist Queue 1202 in, for example, one of the three following ways:
1) If the pID of the media item at the head of the queue is not the pID of the subject media player, corresponding to the left branch of the flow diagram, the “Playlist Play Sequencer” essentially just idles, monitoring the status of the media player with mID=pID in the Session Members state data structure 702 until it is inferred that the media item has been performed. The subject media player infers the media item has been performed when either a transition from played to idle is observed, or the value of the local playtime timer exceeds the performance length of the media item.
2) If the pID of the media item at the head of the queue is the pID of the subject media player, and the Session Members state data structure 702 does not include an entry for another media player with the status value playing, corresponding to the middle branch of the flow diagram, the “Playlist Play Sequencer” plays the media item. The status of the subject media player is set to playing while the item is being performed, and then set back to idle after the performance is finished to signal the performance to the other media players in the session.
3) If the pID of the media item at the head of the queue is the pID of the subject media player, but the Session Members state data structure 702 includes an entry for another media player with the status value playing, corresponding to the right branch of the flow diagram, the performance of the media item is postponed. The “Playlist Play Sequencer” repeatedly traverses this branch of the flow diagram until no other media player has the status value cued, and then sets the status value for the subject media player to cued. On the next iteration the “Playlist Play Sequencer” takes the middle branch of the flow diagram and performs the media item as described above.
As previously described, the Playlist Manager of the Device Manager in each media player device may assume the existence of an autonomous Playlist Builder on the server system in the client-server configuration, or on a privileged peer-to-peer network access system in the peer-to-peer configuration. In peer-to-peer embodiments, the playlist builder could be implemented on the system found in some peer-to-peer networks which hostsnetwork-level functions, while in other embodiments it could be implemented on one peer system in the network.
FIG. 16 depicts how the Playlist Builder 1606 draws on a knowledge base 1604 of information about media items to compose a playlist of media items in the playlist queue 1608 from the pool of media items 1602 available to the media playing devices in the session.
FIG. 17 is a flow diagram of the playlist builder 1606 process. The process assumes the availability of a method for generating the IDs of media items that are candidates for inclusion in the playlist. Candidates may be proposed based on many criteria, including information in the knowledge base 1604, and knowledge of the preferences of the users whose media players are participating in the session.
The “Playlist Builder” may be an iterative process that adds a single media item to the playlist per iteration. Each iteration may include three steps:
1) A candidate media item with ciID is generated based on information in the knowledge base 1604 and/or other criteria. A “who-has-ciID-sID” query is broadcast to all of the media players in the session to determine if any of them has access to the proposed media item.
2) If the Playlist Queue Manager of any of the media players in the session has previously broadcast an unprocessed “force-iID-pID-length-sID” message, the requested media item iID is added to the playlist 2108. The Playlist Builder may broadcast a “queue-iID-pID-length-sID” message to all of the media players in the session, directing that they each add the requested media item to their local playlist that their Playlist Queue Manager is maintaining.
3) After an implementation-determined delay, a determination is made if the Playlist Queue Manager of any of the media players in the session has broadcast an unprocessed “have-iID-pID-length-sID” response message, indicating that a media player in the session has access to the requested media item iID. As one or more of the media players may have access to the requested item, one of those media players is selected either at random, or according to some other criteria, as the media player that will perform that media item and the item is added to the playlist. The Playlist Builder may then broadcast a “queue-iID-pID-length-sID” message to all of the media players in the session directing that they each add the requested media item to their local playlist maintained by their Playlist Queue Manager 1100.
The process assumes that the protocol of the network over which the media players communicate buffers all “force-iID-pID-length-sID” request messages and “have-iID-pID-length-sID” until they can be processed. In any particular embodiment, one or more of each type of message may be processed per iteration. In addition, as the flow diagram indicates, steps 2) and 3) may be repeated a number of times limited by a timeout value to increase the responsiveness of the communications between the Playlist Builder and the media player devices.
The Playlist Builder iterations may be repeated ad infinitum. The playlist is a non-terminating sequence of media items to be performed so long as there is at least one media player in the session. Furthermore, some embodiments may support building playlists consisting solely of media items suggested by users of the media devices, and communicated to the Playlist Builder with the “force-iID-pID-length-sID” request message, by providing an option for setting a option flag to “false” so that the “auto build?” tests in the flow diagram fail.
Still other implementations are disclosed and described with reference to FIGS. 18-26. In these embodiments, playlists of selected media items may be dynamically built from a collection of media items based on the preferences of a temporally-defined group of users with media player devices. Compositional goals and the collection of media items available to meet those goals may be dynamically computed using, for example, the composite taste data for the group as a whole as individual users enter and leave the group.
Some embodiments therefore relate to methods for dynamically creating a playlist of media items responsive to the collective tastes of a temporally-defined group of individuals. Some embodiments also provide for dynamically diversifying the group playlist so that it does not in whole, or in part, unduly reflect the taste of a single member of the group, or a particular subgroup of users within the whole group.
Additional embodiments may provide for a system and method for dynamically building a playlist of media items by using the collective taste preferences of the members of a group to determine compositional goals of the playlist, and then building a group playlist that achieves those compositional goals. The system may derive the compositional goals by analyzing the taste preferences of the current members of the group. Media items available to achieve those goals are typically a subset of the media items that are identified in response to analyzing taste data and may be selected from a collection of media items available to the system. The collection of available media items may be the aggregate of the sub-collections of media items provided by the users or, alternatively, may be a pre-existing set of media items stored, for example, in a central database.
In a preferred implementation of the system, three primary processes are provided. The first process keeps track of users as they enter and/or leave the group by starting or ending communications with the system using, for example, individual networked communication devices. Example embodiments include Bluetooth® devices and other devices communicating in an ad-hoc network of Internet or other network-connected devices using, for example, the Apple Bonjour protocol. Users with individual communication devices may be added to and/or removed from the group by the system as they implicitly or explicitly connect and disconnect from the communication network, which links the individual devices to the computational means for building the group playlist.
As a user enters the group, the system may retrieve the user's taste data. Taste data may be retrieved by, for example, accessing a database of taste data from users known to the system or by requesting taste data directly from the users' communication devices and adding it to the pool of taste data for the group. As a user leaves the group, the system may also be configured to remove that user's taste data from the pool of taste data for the group. In some embodiments in which the media items available for inclusion in a playlist are provided by users and not centrally maintained by the system, the system may maintain a pool of media items available for the current group.
Using the pool of taste data for the group, and the pool of media items available for inclusion on a playlist, the first process may derive compositional goals for the playlist, such as requiring that the values for the selected media items of a specific attribute have a specified distribution. In some embodiments, the first process may also involve selecting a subset of media items from the total pool of media items to be used to build the playlist. This may be accomplished using a media item recommender, such as that described in U.S. patent application Ser. No. 11/346,818 titled “Recommender System for Identifying a New Set of Media Items Responsive to an Input Set of Media Items and Knowledge Base Metrics,” which is incorporated herein by reference in its entirety. The aforementioned media item recommender may be used to recommend media items responsive to an input set of media items.
The second process may build the composite playlist by selecting media items from the total pool of media items in a manner which causes the evolving playlist to more closely approximate the specified compositional goal as the selected items are added to the playlist. Some implementations of the system may therefore be responsive to the constantly changing group membership. In particular, as users in the group continuously enter and leave the group, the compositional goals and/or the pool of media items available to achieve those goals may continually change. Some embodiments may remove media items from the dynamic playlist as users depart from the group, particularly in those situations where the group members actually contribute the media items to the pool. In such situations, the media items in the pool could be removed from the pool as users leave, such as by physically leaving a proximity or by logging out of a system. It should be understood, however, that such a feature is not necessary in all implementations, since a media item can be skipped if it is no longer available when it is to be performed.
The third process may involve diversifying the group playlist. The diversification process may involve shuffling media items on the list as necessary to ensure that no segment of the playlist is dominated by media items representative of the taste of one or more group members. In some embodiments, additional information about aesthetic properties of the media items might be used to rearrange the order of the media items in the group playlist to achieve specific aesthetic goals. Finally, in cases where there are few users, and therefore for each user the playlist includes a large number of items responsive to the taste of just that user, some embodiments may replace some media items with additional media items. These additional or supplemental media items may not be provided by any of the members in the group, and may be selected according to a diversifying criteria to bring more variety to the group playlist.
One embodiment of the system may implement the processes detailed in FIGS. 18-20. This embodiment may be analogized to the automated evolution of a jukebox. Whereas a jukebox supplies local music and requires explicit user interaction from the user to pick desired music, these embodiments automatically build playlists based on previously collected data about users that are proximal (either physically or virtually) to the system.
In describing the embodiment of FIGS. 18-20 below, a process of adding users to the system will be described first, then a process for removing users, and finally the overall playlist construction process is described, including how the process uses information to build a targeted playlist that leverages knowledge of proximal users for building a playlist.
In order for the system to “narrowcast” (i.e., to target content to a specific set of users) a playlist for proximal users, a mechanism may be provided to allows for discovery/detection of proximal users. In one embodiment, a server process 2400 (as shown in FIG. 24) may be provided using, for example, Bluetooth® technology to allow users to “log in” to the system. Users who are “logged in” to the system are considered proximal users, regardless of their physical proximity to the other users or server. In some embodiments, all proximal users immediately begin to affect the playlist being constructed, which will be discussed in greater detail later. It will be obvious to one of ordinary skill in the art that Bluetooth® is just one of many data discovery/entry mechanisms that could be used to add users to the system. Other embodiments may include explicit text entry from users, or even be a zero-configuration protocol such as Apple's BonJour protocol.
In step 1 of one implementation of a Bluetooth® user addition process, a Bluetooth® server with a predetermined service UUID (Universally Unique Identifier) is provided. The Bluetooth® specification uses UUIDs to identify services uniquely across many devices. By using a UUID, a Bluetooth® client is able to detect a specific service on a remote server.
In step 2, once a client Bluetooth® process has connected with the system's Bluetooth® server 2400, the client transmits user information to the server process 2400, as shown in FIG. 25 at 2500. The server process 2400 will then attempt to validate the user information, as indicated at 2501. If the user validation process 2501 fails, then the user is rejected and is not added to the proximal user list 2503 or the proximal device list 2504 by the user list modification process 2502. If the user validation process 2501 succeeds, then the user is added to the proximal user list 2503 and the user device is added to the proximal device list 2504. The means for entry of the user and the user device to the appropriate lists may vary, as would be understood by one of ordinary skill in the art.
In order to provide real-time narrowcasted playlist content, the system preferably updates and maintains the contents of the proximal user list 2503 and proximal device list 2504 on a regular basis. At the same time, the system may be configured to reduce the chance that users are erroneously moved from either list. These concerns may be addressed by using a User/Device Proximity Detection Process 2600, as shown in FIG. 26, to identify which users are proximal to the system. Process 2600, like various other processes, may be implemented using Bluetooth® technology. The system may be able to determine when users are no longer proximal by, for example, detecting proximal devices and mapping the proximal device list 2504 to the proximal user list 2503.
After a user has been discovered to no longer be proximal, a User/Device Inactivity Detection Process 2601 may be used to begin considering whether a user should be removed from the proximal user list 2503 and the user's device removed from the proximal device list 2504. The User Device User/Device Inactivity Detection Process 2601 may be implemented as a time-based process. The system may be configured to remember the time when a user first became classified as non-proximal. Then, for example, if a specific time limit has been reached and the non-proximal user is identified as still being non-proximal, the user may be removed from the proximal user list 2503, along with the user's device from the proximal device list 2504, by the User/Device Removal Process 2602. If a user is found to be proximal by the User/Device Proximity Detection Process 2600 before the time limit of the User/Device Inactivity Detection Process 2601 is reached, then the user and their device will be left on the appropriate lists 2503/2504. It will be apparent to one skilled in the art that specific implementations of the aforementioned system may, but need not, rely on the use of Bluetooth® or a time-based user removal strategy.
The Composite Playlist Builder process may rely on the User Addition/Removal processes described above in order to determine which users to which a playlist is to be narrowcasted. In one implementation, the Composite Playlist Builder process may first generate a list of media items that define a user taste, and then repeat this step for all proximal users in the system.
As shown in FIG. 18, the process of generating a list of media items that define user taste may be accomplished by examining the users 1800 logged into the system and using information—i.e., media item taste data—to synthesize user taste, as shown at 1802. The media item taste data may be gathered by the system contemporaneously or, alternatively, it may be compiled into a list or database from previously gathered media item taste data. Either of the foregoing may be represented by step 1801 in FIG. 18. A list of media items and/or media item metadata (collectively “media data items”) that contains information about the music that should be narrowcasted to user 1800 may then be generated, as indicated at 1803. It should be noted that it is not necessary to have any users logged into the system in order for a group playlist to be generated. Preconfigured goals can be used when no/few users or otherwise insufficient data is available to the system. This process may then be repeated for each user 1800 who is logged into the system.
The data in one embodiment may be cached to enable quick lookup. As will be demonstrated in other steps, the data which represents user taste may also provide the foundation from which other steps derive information. In some embodiments, user taste may be synthesized by first obtaining a set of the tracks that a user has listened to recently and/or those that a user has ranked highest. Of course, it will be obvious to those skilled in the art that there are many different ways to synthesize user taste based on collected user taste data.
As a second process of one implementation, aggregate playlist goals may be computed based on results of the first process and/or on a list of desired categories. Note that there will be typically be one set of input media data items for each user.
In order to compute aggregate playlist goals, each set of media items and/or metadata that indicates user taste 1900, and/or a predetermined or computed set of categories of interest 1901, may be used as input to a User Taste Aggregation Process 1902, as shown in FIG. 19. User Taste Aggregation Process 1902 may be used to create a set of optimal playlist characteristics 1903 for a group playlist that is intended to be narrowcasted to the proximal users. In certain embodiments, media items and/or metadata that are representative of more than one user's taste may not be viewed as duplicative. This feature may allow media items and/or metadata that are common between multiple users to be given increased levels of importance by the User Taste Aggregation Process 1902. The User Taste Aggregation Process 1902 may accept both user taste 1900 and desired categories of interest 1901 as input. Categories of interest 1901 are often (but not required to be) metadata indicators that indicate the set of metadata to be used in computing the optimal playlist characteristics 1903 for a narrowcast playlist targeted at a particular group of users.
One embodiment uses a category set 1901 that comprises a genre. Such a system may compute the frequency distribution of genres in the play histories retrieved for each connected user. The percentage of each genre may then be used as the optimal genre distribution for the generated group playlist. Of course, a genre is not the only category which could be used for generating a playlist. It should also be understood that a raw frequency distribution is not the only method for computing statistics about any given category. Other similar implementations are not limited to, but could employ, a weighting strategy or voting strategy to determine desired levels of each feature.
After the aggregate playlist goals have been computed, relevant media items for a user may be selected from the pool of available media items. This process may then be repeated for each proximal user in the system.
For example, in FIG. 20, the set of relevant media items 2002 that are associated with a given user may be computed/selected. To do this, a process 2001 for each user may be executed. Process 2001 may accept as input the set of media items and/or metadata that indicates a user's taste 1803 and/or the set of all available media items 2000. Process 2001 may also produce a subset of available media items that are relevant media items for a single user.
In one embodiment, process 2001 may be implemented by using a media item recommender, such as those described in U.S. patent application Ser. No. 11/346,818 titled “Recommender System for Identifying a New Set of Media Items Responsive to an Input Set of Media Items and Knowledge Base Metrics,” which was previously incorporated by reference. Process 2001 may also be augmented by providing the media item recommender with the complete scope of recommended media items from which to recommend a subset of media items. In practice, this may be used to make sure that the recommended media items are available for use by the composite playlist builder application. For example, if there are one-thousand available media items 2000 to choose from, process 2001 may ensure that the relevant media items 2002 are within the one-thousand available media items 2000. This is analogous to a jukebox that has a limited set of media from which to produce a playlist. It should be apparent that the media items available to process 2001 do not need to be resident on the same machine that is executing process 2001. Any media item which is programmatically obtainable via any protocol may be considered an available media item.
After the relevant media items for each user have been selected from the pool of available media items, the media items may be categorized according to a set of desired attributes. For example, as shown in FIG. 21, the relevant media items 2100 may be categorized via a categorization process 2101 according to the categories of interest 1901 used previously in the user taste aggregation process 1902. Categorization process 2101 may be used to produce a categorized list of relevant media items 2102. This list may enable the overall system to pick one or more media items based on one or more categories of interest 1901.
After the media items have been categorized according to a set of desired attributes, the status of the current group playlist's achievement of goals may be assessed/computed. For example, with reference to FIG. 22, a goal assessment process 2201 may be used, along with categorization process 2101, to produce a prioritized list of goals needing to be fulfilled 2202. The categorization process 2101 may accept the current playlist 2200 and categories of interest 1901 as input. The categorization process 2101 may use the categories of interest 1901 to categorize the media items on the current playlist 2200. The output of the categorization process 2101 may be sent to the goal assessment process 2201 that may be used to determine how close the current playlist 2200 characteristics are to the optimal playlist characteristics 1903. The output of the goal assessment process 2201 may be a prioritized list of goals needing to be fulfilled 2202.
In one embodiment, a frequency distribution of genres may be computed for the categorization process. The goal assessment process may comprise subtracting the achieved genre percentages for the current playlist from the target percentages for the optimal playlist. It can then take the largest value difference as the needed upcoming genre for the playlist. Of course, a variety of other approaches can be used for computing playlist needs.
After the status of the current group playlist's achievement of goals has been assessed/computed, a set of tracks may be selected to add to the group playlist based on assessing current needs for the playlist according to the goal achievement assessment process previously performed. For example, with reference to FIG. 23, a set of one or more media items 2301 are selected for addition to the current playlist 2200 by a playlist addition process 2302. The media item selection process 2300 may accept as input the prioritized list of goals needing fulfilled 2202, the categorized list of relevant items 2102, and the current playlist 2200. The media item selection process 2300 may attempt to fulfill the goals in a prioritized manner in some embodiments, such that the most important goal is fulfilled by picking a media item from the list of relevant items 2102 that best fulfills the goal. If multiple media items could help fulfill a goal, then a media item may be selected at random from the list or in another fashion, such as by some form of tiebreaker process.
The selected media item may also be subjected to a diversification step, which may be used to ensure that the media items being added to the playlist are not too similar to the current playlist contents. In some embodiments, the diversification step may comprise shuffling media items in the group playlist to diversify at least one segment of the group playlist that includes media data items that are overly representative of the tastes of one or more users. The system may be configured such that, in response to determining that the group playlist is dominated by media items representative of the taste of one or more users, removing at least one media data item from the group playlist that corresponds with the taste of the one or more users.
In some embodiments, in order for a media item to be selected and added to the current playlist 2200, it must pass through the diversification process (unless the diversification process eliminates all media items in the pool). Once a sufficient set of media items have been selected, the set of media items 2301 may be sent to a playlist addition process 2302 for addition to the active playlist. The playlist addition process 2302 may have the capability of creating/updating the current playlist 2200 and, in some implementations, initiating playing the media items in the current playlist 2200.
In one embodiment, a media item may be selected from the most needed genre for the current playlist by choosing at random from relevant tracks categorized in the appropriate genre. The diversification step may be used to ensure that no media item is repeated in the group playlist for a predetermined period of time. In other embodiments, the diversification step may also, or alternatively, be used to ensure that no subset of media items—such as artists, albums, genres, etc.—are repeated within a given subset of the group playlist. For example, the diversification step may be used to prevent media items from a particular artist from being repeated within a seven track window within the group playlist. If it is impossible to fulfill the requirements of the diversification process, then a media item may be chosen at random from all media items within the appropriate genre. If multiple genres are identified with equal priority, then the selected media item may be selected from any one of the identified genres, or by some other tie-breaking selection procedure.
The above description fully discloses the invention including preferred embodiments thereof. Without further elaboration, it is believed that one skilled in the art can use the preceding description to utilize the invention to its fullest extent. Therefore the examples and embodiments disclosed herein are to be construed as merely illustrative and not a limitation of the scope of the present invention in any way.
It will be obvious to those having skill in the art that many changes may be made to the details of the above—described embodiments without departing from the underlying principles of the invention. Therefore, it is to be understood that the invention is not to be limited to the specific embodiments disclosed and that modifications and other embodiments are intended to be included within the scope of the appended claims. For example, one of ordinary skill in the art will understand various aspects of the embodiments disclosed herein could be used in any system for building and sharing a composite playlist from collective group tastes on multiple media playback devices.
The scope of the present invention should, therefore, be determined only by the following claims.

Claims

1. A method for generating a group playlist from a plurality of individual mediasets of users within a group, each user mediaset comprising a respective plurality of media data items, the method comprising:

performing a taste analysis on a first one of the user mediasets to generate a first sub-mediaset from the first user mediaset;

performing a taste analysis on a second one of the user mediasets to generate a second sub-mediaset from the second user mediaset; and

aggregating media data items from the first and second sub-mediasets to generate a group playlist.

2. The method of claim 1, wherein the media data items comprise playable media files.

3. The method of claim 1, wherein the media data items each comprise metadata that identifies a playable media file.

4. The method of claim 1, further comprising:

generating a recommended mediaset for the first user based on the first user taste analysis; and

generating a recommended mediaset for the second user based on the second user taste analysis.

5. The method of claim 4, wherein each of the recommended mediasets comprises metadata that identifies a playable media file.

6. The method of claim 1, wherein the step of aggregating the media data items comprises ranking individual media data items by the number of sub-mediasets in which they appear.

7. The method of claim 1, wherein the group playlist includes media data items comprising songs.

8. The method of claim 1, wherein the step of aggregating the media data items comprises selecting a set of media data items that optimizes a utility function for the group.

9. The method of claim 8, wherein the utility function is based on weights associated with individual media data items.

10. The method of claim 9, wherein the step of selecting a set of media data items that optimizes a utility function for the group comprises selecting a set of media data items so as to maximize a sum of selected media data item weights.

11. The method of claim 10, wherein the step of selecting a set of media data items so as to maximize a sum of selected media data item weights comprises selecting the media data items that maximize a sum of the highest weights for each of a plurality of different media data items in a plurality of user mediasets.

12. The method of claim 9, wherein the step of selecting a set of media data items that optimizes a utility function for the group comprises selecting the media data items that maximize a sum of average media data item weights.

13. The method of claim 1, wherein each of the media data items in the group playlist can be found in a user mediaset within the group.

14. A method for generating a dynamic, composite playlist of media data items for a user group, comprising:

retrieving media item taste data for a plurality of users, wherein the user taste data for a given user is retrieved in response to the given user joining the user group;

aggregating the media item taste data in the user group to create a set of optimal playlist characteristics for the user group; and

adding media data items to a group playlist based at least in part on the set of optimal playlist characteristics.

15. The method of claim 14, wherein the user taste data for a given user is automatically retrieved as the given user joins the user group.

16. The method of claim 14, wherein media data items associated with a given user are retrieved in response to the given user joining the user group.

17. The method of claim 14, wherein the media data items added to the group playlist are selected from a central database of media data items, and wherein the content of the database is independent of the users in the user group.

18. The method of claim 14, wherein the user taste data for a given user is removed from the aggregated user taste data in response to the given user exiting the user group.

19. The method of claim 14, wherein media data items in the group playlist that were retrieved from a given user are removed from the group playlist in response to the given user exiting the user group.

20. The method of claim 14, further comprising subjecting the media data items in the group playlist to a diversification process.

21. The method of claim 20, wherein the diversification process comprises ensuring that no media data item is repeated in the group playlist for a predetermined period of time.

22. The method of claim 20, wherein the diversification process comprises shuffling media data items in the group playlist to diversify at least one segment of the group playlist that includes media data items that are overly representative of the tastes of one or more users.

23. The method of claim 20, wherein the diversification process comprises determining whether the group playlist is dominated by media data items representative of the taste of at least one user and, in response to determining that the group playlist is dominated by media data items representative of the taste of the at least one user, removing at least one media data item from the group playlist that corresponds with the taste of the at least one user.

24. The method of claim 14, wherein the user taste data for a given user is retrieved in response to the given user logging into a network of connected devices to join the user group.

25. The method of claim 24, wherein the step of logging into the network is accomplished implicitly as the network-connected device of a given user enters a given proximity.

26. A system for building a composite playlist from the collective tastes of a plurality of users in a session, each having a separate media device, comprising:

a plurality of session managers, each associated with one of a plurality of media devices, wherein each session manager is configured to manage the status of the media device with which it is associated;

a playlist builder for receiving media item data from the plurality of media devices and for building a composite playlist of media data items from the media item data; and

a plurality of playlist managers, each associated with one of a plurality of media devices, wherein each playlist manager is configured to communicate with the playlist builder to send the media item data of the media device with which it is associated to the playlist builder, and wherein each playlist manager is further configured to play the media data items on the composite playlist on the media device with which it is associated.

27. The system of claim 26, wherein the media item data comprises media item taste data.

28. The system of claim 26, wherein the media item data comprises metadata that identifies a playlist of playable media files.

29. The system of claim 26, wherein the playlist builder resides on a central server.

30. The system of claim 29, wherein the plurality of session managers reside on the server.

31. The system of claim 26, wherein the session managers are configured to verify the eligibility of media devices to participate in the session.

32. The system of claim 26, wherein each playlist manager is configured to stream media files to other media players in the session.

33. The system of claim 32, wherein the system is configured such that each media data item in the composite playlist is located on at least one of the media devices in the session, and wherein each individual media data item on the composite playlist is played at least substantially simultaneously on each of the media devices in the session by causing each of the media data items in the composite playlist to be streamed from its corresponding media device to each of the other media devices in the session.