WO2001084353A2 - Relationship discovery engine - Google Patents

Abstract

A system, method, and computer program product discover relationships among items and recommend items based on the discovered relationships. The recommendations provided by the present invention are based on user profiles that take into account actual preferences of users, without requiring users to complete questionnaires. An improved binomial log likelihood ratio analysis technique is applied, to reduce adverse effects of overstatement of coincidence and predominance of best sellers. The invention may be used, for example, to generate track lists for personalized radio station.

Description

RELATIONSHIP DISCOVERY ENGINE

Inventors:

Ted E. Dunning Bradley Kindig

Cross-Reference to Related Application

[0001] The present application claims priority from provisional U.S. Patent Application Serial No. 60/201,622, for "Recommendation Engine," filed May 3, 2000, the disclosure of which is incorporated herein by reference.

Background of the Invention

Field of the Invention

[0002] The present invention is related to systems, methods, and computer program products for relationship discovery, and more particularly to a system, method, and computer program product of discovering relationships among items such as music tracks, and making recommendations based on user preferences and discovered relationships.

Description of the Background Art

[0003] In many applications for the presentation and marketing of online content, personalization of the user's experience is desirable. Knowledge and application of user preferences permit online advertisers to more efficiently target their advertisements to those users who are more likely to respond. Electronic commerce sites are able to suggest products and services that are likely to be of interest to particular users, based on user profiles and preferences. Such suggestions may be made, for example, by sending e-mail to the user, or by presenting a list of recommended items in the context of a dynamically generated web page. Additional applications exist for such functionality, including both online applications (such as personalized radio stations, news delivery, and the like) and non-online applications (such as tar- geting of direct mail advertising, supermarket checkout coupons, and the like).

[0004] One particular application in which user-specific recommendations may be generated is personalized online radio stations. It is known to provide web pages for delivering selected music tracks to individual users, based on user selection. Compressed, digitized audio data is delivered to users in a streaming format (or al- ternatively in downloadable format), for playback at users' computers using conventional digital audio playback technology such as the Windows Media Player from Microsoft Corporation, or the RealPlayer from Real Networks. It would be desirable for such radio stations to be able to determine which music tracks are likely to be enjoyed by a particular user, even in the absence of, or as a supplement to, explicit selection of particular tracks by the user.

[0005] It is desirable, then, to provide accurate methods and systems for discovering user preferences in particular domains and with respect to particular types of products and services. Several prior art techniques exist for discovering user preferences. In one such technique, as described in U.S. Patent No. 6,064,980, Jacobi et al., "System and Methods for Collaborative Recommendations," issued May 16, 2000, collaborative filtering is employed. Users are asked to complete an online questionnaire specifying their preferences. Such a questionnaire may be presented to the user, for example, when he or she attempts to register for an online service or purchase an online product. The user's responses may then be stored as a user "profile" in a back-end database. The system correlates the profile to the profiles of other users in order to identify users having similar tastes; recommendations are then generated based on the preferences of the similar users.

[0006] However, many users may be reluctant to complete such online questionnaires, due to privacy concerns, or due to an unwillingness to take the time required to answer the questions. Furthermore, such questionnaires often fail to accurately collect user preference information, since they do not actually reflect the user's consumptive behavior; in other words, users may answer inaccurately because they are unaware of (or dishonest about) their own preferences. In addition, the accuracy of the results is limited by the quality of the designed questions. Finally, the stored user profile merely provides a description of the user's preferences at the par- ticular point in time when the questionnaire was completed, and may fail to take into account subsequent changes and/ or refinements to the preferences.

[0007] A second prior art technique for discovering user preferences is to observe user behavior. In online commerce environments, user behavior can be observed by tracking the particular pages visited, products ordered, files downloaded or accessed, and the like. Users may be prompted for login identifiers, providing a mechanism for identifying users. In addition to or instead of login, cookies may be stored on users' computers, as is known in the art, in order to recognize a user who has previously visited a site. Thus, user behavior can be tracked over multiple visits, without requiring the user to set up a login identifier or to even be aware that his or her behavior is being tracked.

[0008] For example, many online commerce sites keep track of user purchases, and, based on such purchases, make recommendations as to products and services that are likely to be of interest to a particular user. Such recommendations may be based on analysis of the purchases of other users who have purchased the same products and services. User browsing may also be monitored, so that recommendations may be based on products that the user has browsed, as well as those he or she has purchased.

[0009] The above-described technique for observing user behavior may lead to inaccurate results. Relatively few data points may be available, particularly when recommendations are based on user purchases. For example, a typical user may make four or five purchases annually from any particular online store, and may distribute his or her purchases among several stores, including online, conventional retail, and/ or other outlets. The relatively small number of purchases tracked by any particular store may be insufficient to develop a reasonably accurate user profile in a relatively short period of time. Thus, recommendations in such systems are often inaccurate since they are based on insufficient information. [0010] Furthermore, some purchases may be gifts, and may thus fail to accurately reflect personal preferences of the purchaser. In some cases, the purchaser may specify that an item is a gift (by requesting gift-wrapping, or a gift message for example), so that the item may be excluded from user behavior analysis; however in many cases the purchaser may not make the online merchant aware of the fact that the purchase is a gift, and there may be no way for the merchant to make this determination. Distortions and inaccuracies in the user profile may then result. In particular, when relatively few data points are available, each individual gift purchase may have a particularly powerful distorting effect on the user profile. [0011] Finally, distortions may result from the fact that, once a purchase is made, the merchant may not be able to easily determine whether the purchaser was satisfied with the product.. This is a particular problem in connection with products that are typically only purchased once, such as books, videos, and compact discs. A user may purchase a compact disc and listen to it only once, finding the music not to his liking. The user may purchase a second compact disc, by another artist, and enjoy it imrnensely, hstening to it hundreds of times. The user's behavior with respect to the online merchant is the same for the two cases— namely, a single purchase of a compact disc. The online merchant cannot determine, from the purchasing behavior, the musical tastes and preferences of the user, since the merchant is not aware of the post-purchase behavior of the user.

[0012] In addition to the above problems with data gathering for developing user profiles, there are additional limitations and shortcomings of conventional recommendation engines, with respect to the data analysis that is performed to generate recommendations. Conventionally, recommendations are made based on data analysis performed on the observed user behavior. Several types of data analysis are known in the art for developing recommendations based on observed behavior. One commonly used technique is to observe that people who buy a particular product X also tend to be more likely to buy a particular product Y. Thus, the system may suggest, to a user who is observed purchasing (or browsing) product X, that he or she may also be interested in product Y. The basis for the suggestion is an observed correlation between purchasers of product X and purchasers of product Y. [0013] Such a data analysis technique often leads to inaccurate results, particularly when the observed purchase is a relatively rare product. Relationships among such products often tend to be overstated, since relatively few data points are available for both the purchased product and the suggested product. Thus, the sig- nificance of a particular co-occurrence (i.e. an observed purchase of two products by the same individual) is given undue weight, when in actuality the co-occurrence may merely be a coincidence and may not provide an accurate indication of a relationship between the two products. In addition, certain products, such as "best sellers," tend to appeal to virtually all consumers, so that co-occurrence is seen between a best seller and nearly every other product. Conventional data analysis techniques often fail to yield meaningful results, because of both the overstated significance of coincidental co-occurrence, and the overpowering influence of best sellers.

[0014] The following is an illustration of ihe deficiencies of conventional data analysis techniques in situations involving a rare product and/or best sellers. Analysis of the co-occurrence of events A and B (e.g. a purchase of product A and a purchase of product B) involves construction of the following matrix:

A ~A

B k(B)

-B k(~B)

k(A) k(~A) k(*)

[0015] where: [0016] k(AB) is a count of the number of times A and B both occurred; [0017] k(~AB) is a count of the number of times A did not occur and B oc- curred;

[0018] k(A~B) is a count of the number of times A occurred and B did not occur;

[0019] k(~A~B) is a count of the number of times neither A nor B occurred;

[0020] k(A) is a count of the total number of times A occurred;

[0021 ] k(~A) is a count of the total number of times A did not occur;

[0022] k(B) is a count of the total number of times B occurred;

[0023] k(~B) is a count of the total number of times B did not occur; and [0024] k(*) is a count of the total number of events.

[0025] If p(B I A) = p(B), where p(B | A) is the probability of B given that A has occurred, and p(B) is the probability of B, then events A and B are considered to be independent. It also follows that if p(A)p(B) = p(AB), where p(A) is the probability of A, p(B) is the probability of B, and p(AB) is the probability of both A and B occurring, then A and B are independent.

[0026] It is assumed that probabilities can be estimated from observed event occurrences using the maximum likelihood principle, so that

[0027] ^ _{≤ /j(y3} | -4); and κ(A)

[0028] §- = p(B) k(*)

[0029] As discussed above, A and B are independent if p(B | A) = p(B). Accordingly, if — - > 1 , A and B are appearing together more than expected for in- P(B) dependent events. Substitution of the above equations yields the following test: k(AB)k(*) [0030] If — — > 1 , a co-occurrence relationship can be established. k(A)k(B) [0031] The above-described technique is deficient, in that quantization effects tend to overpower meaningful results. Particularly where event counts are small, coincidences often translate into perfect correlations, yielding misleading results.

[0032] Pearson's Chi-Squared test improves on the above-described tech- nique by introducing an estimate of significance. According to this technique, independence is assumed and a determination of how many k(AB) and k(A~B) would be expected. Expected k(AB) can be expressed as:

[0033] *(.«)= -2 -2 k( )

[0034] If k(AB) and all similar estimates are greater than five, the distribu- tion of the count of multinomially distributed events can be approximated using a normal distribution. Assuming a normal distribution, the difference between the ob- served k(AB) and the expected value can be determined and squared. The sum of the squared normal distribution is known to be χ² . Accordingly, the significance of the difference is then determined, and unexpected co-occurrence defined.

[0035] However, Pearson's Cm-Squared test yields misleading results when one of the events is relatively rare (such as when the expected count is less than 5). In such situations, the assumption of normal distribution tends to lead to an overstatement of the significance of the co-occurrence.

[0036] A second prior art data analysis technique for developing product recornmendations employs archetypal customers in order to categorize users accord- ing to observed behavior. Such techniques are employed, for example, in LikeMinds 3.1 from Macromedia Corporation. A set of customers is selected and denoted the archetype set. Prospective purchasers and users are compared with the archetype set in order to determine which archetypes they most resemble. However, such systems may also lead to inaccurate results, since the set of archetypes is often insufficient to accurately describe individual real-world users. In many situations, archetypes are non-orthogonal to one another, and the archetype set thus provides a poor basis space for modeling users. The system may thus fail to provide a concise description of a user (if too many archetypes are needed to provide an accurate description), or the description may not be accurate (if too few archetypes are used). [0037] In some variations, users may be presented with a list of archetypes and asked to select which archetype(s) they most resemble. Such an approach leads to similar disadvantages as described above with respect to questionnaires, and also may lead to inaccuracies as users have difficulty selecting a subset of archetypes that accurately reflects their own preferences. In such an approach, it rapidly becomes apparent that, no matter how many archetypes are available, the user cannot easily be defined as a sum of fixed archetypes.

[0038] The archetype approach also tends to yield recommendations that are dominated by a particular subgroup. Very popular items filter to the top of the list, since most archetypes are readers of bestsellers (as is most everyone; hence the definition of "bestseller"). This massive overlap of best sellers exacerbates the prob- lem of non-orthogonality of the archetype set. If bestsellers are removed from the set of items, results may be inaccurate because coincidental co-occurrences then dominate, as described above. This problem may be even more prevalent when this approach is employed, since the non-orthogonality of the archetype set tends to increase the noise sensitivity of the system, so that coincidental matches (as described above) become even more significant, leading to increased levels of distortion and unsatisfactory results.

[0039] Caid et al., U.S. Patent No. 5,619,709, for "System and method of context vector generation and retrieval" describes an approach that attempts to deal with this problem of non-orthogonality by expHcitly constructing an orthogonal basis space with relatively low dimensionality. However, such reduced-dimensionality systems suffer from the limitation that distinctions between words tend to be lost when reducing the dimensionality of the system. The loss of such distinctions can improve recall in an information retrieval system, but leads to a decrease in precision. Precision, expressed as the fraction of high scoring results that are correct, is the most useful figure of merit for a recommendation system.

[0040] What is needed is a system and method of generating and providing recommendations to users that avoids the above-described -imitations and disadvantages. What is further needed is a system and method of discovering relationships among items, that is not obtrusive to users and that leads to accurate recommendations based on user preferences. What is further needed is a recommendation engine that provides improved accuracy by reacting to user preferences that may change with time, and by collecting a larger number of data points so that more accurate profiles may be developed.

Summary of the Invention

[0041 ] The present invention provides a recommendation engine and application capable of discovering relationships among items and recommending items without requiring undue effort on the part of the user. The recommendations provided by the present invention are based on user profiles that take into account actual preferences of users, without requiring users to complete questionnaires. Problems of non-orthogonality, sparsity of data points, overstatement of coincidence, dominance of bestsellers, and flaws in the data source, as described above, are avoided. Thus, the present invention facilitates generation of recommendations that are likely to be of interest to the user, and leads to improved marketing and ad targeting, along with greater credibility and utility of the recommendation system.

[0042] The present invention provides improved data analysis by avoiding inaccurate assumptions regarding distribution of user preferences. In particular, the present invention employs a binomial log likelihood ratio to provide improved analy- sis of data points describing user preferences, and to avoid inaccurate assumptions inherent in a normal distribution analysis. The invention thus provides improved recommendation generation, while avoiding the problems of overstatement of coincidences and dominance of bestsellers, described above.

[0043] Furthermore, in one embodiment, the present invention operates in the domain of music, making recommendations as to music tracks (such as songs), based on analysis of music tracks previously selected by the user for listening. The invention may operate, therefore, in connection with a personalized radio station for playing songs over the Internet, based on user selection of tracks and based on recommendations derived from previously selected tracks. Conventional techniques for programming radio stations may be applied and combined with the techniques of the present invention. Thus, a plurality of programming "slots" may be specified in a given time period, to be filled alternately by explicit user selections (or requests), and by recommendations generated by the present invention based on the user's preferences. As the user makes additional selections of music tracks, the system is able to accumulate more information as to the user's preferences, so that more accurate recommendations may be made.

[0044] Since, in the context of a personalized radio station, a user specifies music tracks that he or she is interested in hearing, a finer granularity of user preferences can be recorded. By contrast to online commerce environments such as pur- chases of books, compact discs, and the like, in which a typical user may make four or five purchases annually, the present invention offers the opportunity to observe the user making selections several times per hour. The present invention thus facilitates more rapid data collection regarding user preferences, and thus provides more accurate profile generation.

[0045] In addition, repeated requests for a particular track may be noted, with the number of requests tending to indicate the level of satisfaction or enjoyment with regard to the requested music track. If a user aborts a track soon after it has begun, that may be an indication that the user does not like the track. Conventional user profile generation techniques, based on user purchases, do not include such a mechanism for deterrnining the degree of satisfaction of a user by observing the user's behavior, since a user does not tend to make repeated purchases of a particular item even if he or she enjoys the item. Thus, by contrast to conventional monitoring of online purchases, the present invention facilitates development of a user profile that indicates the degree to which various items are preferred. Negative, as well as positive, data points may be extracted, based on users aborting or repeating track playback, respectively. Finally, users' preferences are more accurately recorded, since the purchase of gifts for others ceases to be a factor in the context of an online radio station (a user does not listen to music "on behalf of" another person).

[0046] Based on recorded user preferences and data analysis as provided by the present invention, relationships among works are discovered, and recommenda- tions may be generated.

[0047] Additional applications of discovered relationships may also be provided. In one application, results of text-based searches (such as for albums by a particular artist, for example) may be enhanced by the discovered relationships of the present invention. Thus, in an online commerce environment, a user may search for artist A and be presented with works by artist B as well, based on a relationship between artists A and B that is discovered by analysis of user Hstening behavior. Such an application illustrates the utility of the present invention in discovering relationships based on user listening, and applying the relationships to generate recommendations in online commerce. [0048] In another application, the present invention may be employed in connection with conventional radio station prograrnn rng techniques, to implement an improved personalized radio station. As is known in the art, conventional radio stations typically divide a programming block (such as a one-hour period) into a number of segments. Each segment is assigned a programming category, such as "power hit," "new release," "recurrent hit," and the like. For a particular program- ming block, music tracks are assigned to each of the segments based on the particular programming format of the radio station. Music scheduling software, such as Selector® by RCS Sound Software, applies heuristic rules for repetition limits and classes of songs, to automatically generate track lists for use by radio stations. The present invention may be combined with such existing radio station prograrn ing tech- niques, to populate the defined segments with music tracks that are likely to appeal to a particular listener. Additional rules may be applied in generating track lists, so as to limit undesired repetition and to comply with limiting legislation (such as the Digital Millennium Copyright Act) and other restrictions.

[0049] In another application, the discovered relationships of the present invention may be employed to improve targeting of advertising. Once relationships between music tracks and/ or artists have been developed, users may be presented with ads that are most likely to be of interest to them. This provides another example of application of relationships discovered in one domain to content delivery in another domain, according to the present invention. [0050] As can be seen from the above examples, the present invention may be applied to many different domains, and is not limited to application to the domain of personalized online radio stations. In addition, relationship discovery according to the techniques of the present invention is not limited to observation of music listening habits. Many of the techniques of the present invention may be applied to observa- tion of user behavior in other domains, such as online or conventional purchases, viewing of web pages, viewing of television programs, movie ticket purchases, pay- per-view orders, and many others. In addition, the present invention may be applied to document-based systems, in order to detect relationships among documents based on co-occurrences of words and phrases therein. Brief Description of the Drawings

[0051] Fig. 1A is a block diagram of a functional architecture for one embodiment of the present invention.

[0052] Fig. IB is a block diagram of sequence construction flow according to one embodiment of the present invention.

[0053] Fig. IC is a block diagram of a sample history structure according to one embodiment of the present invention.

[0054] Fig. 2 is a data flow block diagram for one embodiment of the present invention. [0055] Fig. 3 is a block diagram showing an implementation of log and play history analysis according to one embodiment of the present invention.

[0056] Fig. 4 is a block diagram showing a technique for identifying related music tracks according to one embodiment of the present invention.

[0057] Fig. 5 is a block diagram showing a technique for identifying a map- ping between music tracks and artists according to one embodiment of the present invention.

[0058] Fig. 6 is a block diagram showing a technique for identifying a mapping between users and artists according to one embodiment of the present invention.

[0059] Fig. 7 is a block diagram showing a technique for identifying a ap- ing between users and music tracks according to one embodiment of the present invention.

[0060] Fig. 8A is a block diagram showing a technique for generating recommendations according to one embodiment of the present invention.

[0061] Fig. 8B is a block diagram showing a technique for generating notifi- cations according to one embodiment of the present invention.

[0062] Fig. 9 is a block diagram of a data model according to one embodiment of the present invention.

[0063] Fig. 10 A is a block diagram showing data flow for a browse function according to one embodiment of the present invention. [0064] Fig. 10B is a block diagram showing data flow for a recommendation function according to one embodiment of the present invention.

[0065] Fig. 11 is an example of a screen shot depicting sample artist-level relationships. [0066] Fig. 12 depicts main components for a sample user interface of a jukebox that implements the present invention.

[0067] Fig. 13 is a flow diagram of a method of initializing and ma taining a content index.

[0068] Fig. 14 is a flow diagram of a method of operation for a relationship discovery engine according to the present invention.

[0069] Fig. 15 is a flow diagram of a method of extracting significant information according to the present invention.

[0070] Fig. 16 is a block diagram of a conceptual architecture for one embodiment of the present invention. [0071] Figs. 17A, 17B, and 17C depict additional main components for a sample user interface of a jukebox that implements the present invention.

[0072] Fig. 18 depicts a series of menus for a sample user interface of a jukebox that implements the present invention.

[0073] Figs. 19A and 19B depict interface elements for File menu items of a sample user interface of a jukebox that implements the present invention.

[0074] Figs. 20A, 20B, and 20C depict interface elements for Edit menu items of a sample user interface of a jukebox that implements the present invention.

[0075] Figs. 21A through 21F depict interface elements for View menu items of a sample user interface of a jukebox that implements the present invention. [0076] Figs. 22A, 22B, and 22C depict interface elements for Option menu items of a sample user interface of a jukebox that implements the present invention. [0077] Figs. 23A through 23G depict interface elements for Option menu items of a sample user interface of a jukebox that implements the present invention. [0078] Figs. 24A through 24C depict interface elements for Music Library menu items of a sample user interface of a jukebox that implements the present invention. [0079] Figs. 25A and 25B depict interface elements for Recorder menu items of a sample user interface of a jukebox that implements the present invention.

[0080] Figs. 26A, 26B, 26C, and 26D depict interface elements for Radio menu items of a sample user interface of a jukebox that implements the present in- vention.

[0081] Figs. 27A and 27B depict examples of scalable coding according to one embodiment of the present invention.

Detailed Description of the Preferred Embodiments

[0082] The following description of preferred embodiments of the present invention is presented in the context of an online recommendation engine for music tracks, such as may be implemented in an Internet-based jukebox or personalized radio station. One skilled in the art will recognize that the present invention may be implemented in many other domains and environments, both within the context of musical recommendations, and in other contexts. Accordingly, the following descrip- tion, while intended to be illustrative of a particular implementation, is not intended to limit the scope of the present invention or its applicability to other domains and environments. Rather, the scope of the present invention is limited and defined solely by the claims.

Architecture [0083] Referring now to Fig. 16, there is shown a conceptual architecture of one embodiment of the present invention. In the architecture of Fig. 16, the invention is implemented in connection with a web-based "jukebox" 103, or personalized radio station, which accepts a user's selections of music tracks and makes additional recommendations as to music tracks the user is likely to enjoy. The user is able to search for particular tracks and/ or artists, and to control the playback of selected tracks. The system monitors the user's behavior with regard to searching, listening, and playback control, and generates and analyzes logs of such behavior in order to refine recommendations. Advertising, offers, and other information may be selected and pre- sented to the user based on observations of user behavior and analysis as to which material may be of interest to the user.

[0084] Stream delivery system 150 interacts with jukebox 103 to specify a sequence of audio files to deliver to jukebox 103. Jukebox 103 transmits requests to stream delivery system 150, and stream delivery system 150 delivers the audio files, as tracks, to jukebox 103. Stream delivery system 150 also communicates with realtime subscription authorization module 157, which includes real-time server 154 and database 156 that keep track of which user accounts are active and enforces global business rules about which accounts can listen to the radio at a given time. Within stream delivery system 150, there are a number of distinct software entities. Radio sequence generator 1613 receives requests from jukebox 103, receives format definitions 1611 and general constraints 1616, and receives recommendations from recommendation engine 107, to generate track selections to be transmitted to jukebox 103. The track selections generated by radio sequence generator 1613 specify which files to play according to estimated listener preferences as well as pre-determined station formats. Authorization and content server 1614 keeps a record of the files that are selected by radio sequence generator 1613; server 1614 is consulted by radio sequence generator 1613 when files are requested. If generator 1613 does not provide the necessary security information, server 1614 flags this anomaly and declines to provide the data.

[0085] Compressed signal files 1615 contain descriptions of music tracks, and in one embodiment contains digitized representations of the music tracks themselves. Compressed signal files 1615 are stored, for example, using conventional database storage means or in a conventional file system, and in one embodiment include several fields providing descriptive information regarding music tracks, such as title, album, artist, type of music, track length, year, record label, and the like.

[0086] Stream delivery system 150, real-time subscription authorization module 157, format definitions 1611, and general constraints 1616 are collectively designated as the radio sequence transmitter 121 of the present invention. [0087] Referring now to Fig. 1A, there is shown a block diagram of a functional architecture for one embodiment of the present invention. Content index 110 provides a concise index of content stored in database 102, and is generated by conventional index generation means, to enable more efficient searching and updating of database 102.

[0088] In one embodiment, relationship discovery engine 1604 uses a transient (non-persistent) TCL associative array, or hash table, (not shown) as is known in the art. The array includes a number of logical tables segmented by short prefixes on the keys. Track names are stored, for example, as lowercase strings, tracklDs as 32-bit integers. One example of a format for the array is as follows:

[0089] Index and search module 104 facilitates functionality for accepting user queries and searching database 102 for particular music tracks. In one embodiment, the user enters queries by accessing web site 106, which provides an interactive user interface for accessing the functions of the present invention. Web site 106 provides the main point of contact with users. A user interacts with web site 106 over a network, using a conventional web browser 105 (such as Microsoft Internet Explorer), running on a client computer. Module 104 accesses database 102 and index 110 in response to user queries. In addition, module 104 receives recommendations from recommendation engine 107, via web site 106. In one embodiment, module 104 also receives information from learned artist relationships 1605. Results are returned to the user via web site 106. In one embodiment, index and search module 104 also dy- namically updates content index 110 in order to provide improved efficiency for future searches. Such indexing techniques are well known in the art.

[0090] Index and search module 104 may provide fuzzy search capability to improve robustness and increase user satisfaction. Such capability detects imperfect matches between entered query terms and indexed content, so as to account for spelling errors or slightly incorrect titles or artist names in the entered query terms. Search capability includes, for example, searches for albums by artist, tracks by artist, text searches of lyrics, and the like. As described below, search results may be augmented by including secondary results that are si-milar to or related to the primary results, according to the relationship discovery techniques of the present invention. Thus, when a user searches for tracks by a particular artist, the invention may also present tracks by other artists that are musically related to the searched for artist. In one embodiment, module 104 presents a series of "browse pages", viewable via web site 106, for browsing through lists of related music tracks and artists. The user may follow links for particular tracks and artists, to either play the tracks, or continue browsing for additional related tracks. These related items are provided by recommendation engine 107.

[0091] In one embodiment, relationship discovery engine 1604 performs the following operations in developing and mamtaining learned artist relationships 1605: [0092] Add play logs

[0093] Calculate fixed parameters after indexing

[0094] Prune the index of tracks occurring in fewer than a threshold number of play logs

[0095] Read the index from a file [0096] Write the index to a file

[0097] Find the number of occurrences of a track in the corpus

[0098] Find the total number of tracks seen in the corpus

[0099] Find the number of unique tracks seen in the corpus

[0100] Find the set of play logs a track occurs in [0101] Find the number of occurrences of a track in a play log

[0102] Find the tracks that occur in a play log [0103] In one embodiment, the above operations are performed by creating and using a TCL associative array as described above in connection with the memory structures in relationship discovery engine 1604.

[0104] In addition, web site 106 offers the capability for suggesting tracks and artists that may interest the user, based on personal criteria 111, profiles 112, of track-level discovered relationships based on observed user listening behavior determined by log analysis 113 of play logs 114, as described in more detail below.

[0105] Personal criteria 111 is a database that stores demographic, contact, and other descriptive information concerning individual users. Personal criteria 111 may also include expressed preferences of particular artists, genres, and the like, which may be collected from the user by online surveys. The musical suggestions provided by web site 106 may be based in part on analysis of personal criteria 111, based on observations that certain types of music tend to appeal to users associated with certain profiles or demographic categories. [0106] Play log 114 is a database that monitors and stores information describing user behavior. Specifically, the user's interaction with jukebox 103, including track selection, repeats, aborts and skips, and the like, are recorded and stored in play log 114. Log analysis module 113 analyzes play log 114 in order to generate a profile of the user, which is stored in profile database 112. Profile database 112 contains user-level profiles that encode personal listening behavior of particular users. Log analysis module 113 periodically updates profile database 112 as new information becomes available, so as to refine the user profile over time.

[0107] In one embodiment, play log database 114 contains tables for storing forward and inverted indexes for play logs (play logs to tracks and tracks to play logs).

[0108] Tables in play log database 114 are implemented, for example, as

TCL associative arrays (hash tables) as are known in the art. Play log database 114 includes a number of logical tables segmented by short prefixes on the keys. In one embodiment, index tables in database 114 and in other databases and tables of the present invention use lists of track, album, or artist identifiers associated with a play log. [0109] Recommendation engine 107 provides suggestions for tracks and artists that are likely to appeal to a particular user. Suggestions provided by engine 107 are presented via web site 106 in the form of web pages, or via jukebox 103, or by some other output means. Recommendation engine 107 takes as input the user pro- file from profile database 112, as well as personal criteria database 111 containing demographic and other information describing the user. Thus, engine 107 uses a combination of explicit preferences and observed behavior to provide personalized music recommendations at any desired level, including for example tracks, artists, albums, genres, and the like. Details of the operation of recommendation engine 107 are provided below.

[0110] In one embodiment, the invention provides some music tracks for free, while others are only available upon receipt of payment. Payment may be collected via credit card or other means, as is known in the art. Suggestions provided by recommendation engine 107 and displayed via web site 106 may include both free and "for sale" music tracks. In addition, the user is able to preview tracks before deciding whether to purchase them. In one embodiment, suggestions made by recommendation engine 107 are augmented by additional information such as special offers or paid advertisements 109. Inventory 108 is a database of active advertisements, offers, promotions, and events that may be relevant to users that fit particular demo- graphic profiles and/ or expressed preferences.

[0111] Selected tracks are played via jukebox 103, which is implemented in one embodiment as a standalone application, or as a plug-in or bundled feature in browser 105. Jukebox 103 receives digitized representations of music tracks and plays the tracks over a speaker or headphones at the user's computer. In one embodiment, jukebox 103 can download and save music tracks in a compressed format, such as MP3, for playback on the user's computer or on a portable digital music listening device. A sample user interface for a jukebox application is described below in connection with Fig. 12.

[0112] Outbound notifier module 116 generates e-mail 119 or other com- munication that is sent to users in order to announce availabiUty of new tracks or other items, events, or promotions that may be of interest. For example, if a user has expressed interest in a particular artist, and that artist releases a new album or is touring the user's area, an e-mail 119 may be sent to the user. Notification criteria 115 are defined and provided to notifier module 116, in order to specify under what conditions such e-mail 119 should be generated and sent. User profile 112, based on log analysis, as well as personal criteria 111, and data from content index 110, may be used as input to notifier module 116 in deterrni ing the content of e-mails 119. In addition, third-party data 120 (such as touring information for artists), may be processed by a list generator 117 and filtered by targeting criteria 118 to be provided as further output to notifier module 116. In this manner, generated e-mails 119 are likely to be of value and interest to particular users. For example, tour information for an artist, as provided by a third party, may be sent to users whose preferences (whether observed or stated) indicate that the user would be interested in hearing about that artist.

[0113] In one embodiment, profile database 112 is augmented and en- hanced by data from user feedback. When users listen to music tracks, they may be offered the opportunity to provide feedback as to whether they enjoyed the tracks, and as to their opinions on other tracks and artists. Such feedback is processed and stored in profile database 112 and may be used as a basis for future recommendations provided by recommendation engine 107. In addition, such feedback may be used to generate and/ or refine discovered relationships among artists and tracks.

[0 14] One advantage of the present invention is that it provides recommendations that are responsive to particular tastes and preferences of individuals, so as to enable implementation of a personalized radio station that presents music tracks likely to be enjoyed by the individual user. As described below, the invention dis- covers relationships among artists and tracks in order to find musical selections that the user is likely to enjoy, based on observed behavior and profile information describing the user. These relationships can further be employed to serve as a basis for delivery of advertising, improved searches, customized promotions and offers, and the like. [0115] The present invention develops detailed behavior profiles based on observed user listening behavior. User track selections, made via jukebox 103, are monitored, along with user operations such as repeating, skipping, or scanning through tracks. Behavioral data is provided as input to a relationship discovery engine that operates as described herein. Relationship discovery takes place based on statistical analysis of track-to-track co-occurrences in observed user behavior. Rec- ommendation engine 107 uses discovered relationships to generate suggestions of additional artists and tracks. User profiles, as stored in profile database 112, contain descriptions of analyzed play logs, as well as additional track suggestions related to the tracks the user has demonstrated he or she likes. Profiles can be modified, enhanced, or filtered, to include second- or third-level related artists or track, or to in- elude only tracks the user does not aheady own. A randomization component may also be included in the development of profiles.

[0116] The architecture shown in Figs. 1A and 16 may be used, for example, for implementing a personalized radio station that takes into account learned relationships among artists and/ or tracks. Using the architecture of Figs. 1A and 16, the system of the present invention learns relevant relationships, and populates a learned relationships database 1605 with the results. In one embodiment, the system acquires information from a deployed population of jukeboxes 103.

[0117] Referring again to Figs. 1 A and 16, learned artist relationships

1605, along with user profiles describing characteristics of users, are provided to rec- ommendation engine 107, which operates as discussed above and transmits recommendations to radio sequence generator 1613, which is a component of radio sequence transmitter 121. Format definitions 1611, which includes descriptions of radio station formats (e.g. alternative rock, country/ western, etc.), and other general constraints 1616 such as, for example, track schedules (e.g. play a top-40 hit at the top of each hour), are also provided to radio sequence generator 1613.

[0118] Recommendation engine 107 generates track preferences based on user information. Radio sequence generator 1613 uses track preferences, along with general constraints 1616 and format definitions 1611, to generate a sequence of tracks to be played. General constraints 1616 include particular rules and restrictions on the sequence of tracks, as may be required by law or as may be determined to be desirable for marketing or aesthetic purposes or for other reasons. Examples of constraints 1616 include: "no more than one song per hour from a particular album," or "do not play a fast song immediately after a slow song." Radio sequence generator 1613 may also incorporate a randomization element, if desired, and may be configurable by a website operator. [0119] The track list is sent to jukebox 103 to be played to the user. A user activates jukebox 103 and selects music tracks for playback and/ or purchase, via a user interface including controls and selectors. Authorization and content server 1614 checks that the appropriate security measures are in place (in order to prevent the user from "hacking" jukebox 103 to request unauthorized tracks from content server 1614), obtains the actual music tracks from files 1615, and provides them to jukebox 103 for output.

[0120] In one embodiment, the connections among the various elements of Figs. 1A and 16 are implemented over the Internet, using known protocols such as HTTP and TCP/IP. Secure sockets layer (SSL) or other encryption techniques may be employed for added security.

[0121] In one embodiment, play logs representing the user's behavior are accumulated and stored in local storage at the user's computer. At periodic intervals, such as every one hundred songs, jukebox 103 transmits the locally stored play logs to centrally stored play log database 114. The transmission of play logs is accom- plished using any known network transmission protocol, such as FTP, HTTP, and the like. As described previously, play log database 114 includes play log data from all active jukeboxes 103 in operation, including those in use by all active users. In an alternative embodiment, play log database 114 may contain a subset of such information, based on geographic delimiters, storage limitations, or other factors. [0122] Relationship discovery engine 1604 mines database 114 to generate learned relationships, which are stored in database 1605. Discovery of relationships takes place according to techniques described in more detail below.

Sequence Construction

[0123] Audio files are selected by fusing estimated user preferences, radio station format requirements and general sequence constraints. Referring now to Fig. IB, there is shown a block diagram of sequence construction flow according to one embodiment of the present invention.

[0124] In this process, a human-designed "program clock" is used to specify a station format 161. Format 161 defines time slots that are filled sequentially. Each time slot has a class of songs that can be played in that time slot and each class has an associated set of audio files. The program clock specifies penalties for playing a song from a different class than the one specified. Station format 161 keeps track of the current time slot and outputs a list of all songs that can be played with associated (possibly zero) penalties. [0125] Listening preferences 162 for the listener of the current station are estimated either by analyzing the music that the listener's jukebox has reported that the listener has listened to or by asking the user to enter the names of a few favorite artists. In any case, these preferences are reduced to a list of bonus scores for each possible song that can be played. [0126] In order to decrease the predictability of the sequence of music played on a station, small random penalty scores 163 are associated with each song that can be played. This random penalty is small enough so that it does not outweigh the preference scores, but it is large enough to rearrange the order of the preferred songs. [0127] Candidate songs are scored to find violations of sequence constraints by rule engine 164 that has access to a list of all potentially playable songs as audio files 165 and a listener history 167 containing the songs that the current listener has heard on this station. The history structure is designed to allow songs to be scored very quickly and is customized for the sequence rules being used. History structure 167 and penalties are discussed in more detail below.

[0128] Score fusion 166 adds up all of the scores (bonuses as positive numbers, penalties as negative) for all possible songs. This is done using a standard heap data structure to si-mplify finding the song with the highest resulting score. Next song selection 168 identifies the "best" song to play next. The selected song is then inserted into the Hstener history structure 167 so that it affects future song selections. [0129] In one embodiment, the rules supported by the sequence constraint rule engine 164 are aU of the form: "Add a penalty of x whenever attribute y occurs more than n times in the most recent (m plays) or (t minutes)."

[0130] In one embodiment, attributes include the artist, album name and track name for songs that have been played by the radio for a particular listener.

Other candidate attributes include mood and tempo. This form of rule is sufficient to encode most of the desirable constraints for radio programming including both programmatic constraints as weU as legal constraints, such as those arising from the requirements of statutory Hcenses under the Digital MiUennium Copyright Act. One additional form of rule that is known to be useful is based on the combination of some attribute such as tempo from the last and current track. This additional rule form can be used to prevent huge variations in tempo or mood. One skilled in the art will recognize that many other rules and rule types could be employed.

[0131] In one embodiment, the data structure used to implement Hstener history 167 uses a number of cascaded queues with associated hash tables to maintain the necessary counts for attributes of aH past events. There is one hash table of counts associated with each rule. This hash table counts the number of times each unique value of the attribute associated with that rule has been seen in the time period associated with the rule. The counts in the hash table are incremented when a song is en- tered into history structure 167 and decremented when a song is removed from the associated queue. More than one hash table may be associated with each queue.

[0132] Referring now to Fig. IC, there is shown a sample history structure in connection with the sequence construction flow of Fig. IB. Two kinds of queues are maintained to retain the distinction between rules that are time based (last t hours) or ordinal (last m plays). The sample history structure includes hourly histories 171, 172, and 173; cumulative hourly counts 174, 175, 176, and 179, and ordinal queues 177, 178, 180, and 181.

[0133] There are two major operations on a history structure. These include the addition of a new event and testing a new event to determine if it would invoke any penalties. The addition of a new event involves the insertion of the event into the first of the time-based and ordinal queues and the incrementing of all tables according to the attributes values in the new event. Each queue must also be inspected to see if any events need to be moved to the next queue either due to the time or size Hmits on the starting queue. When an event is moved from one queue to another, aU of the hash tables associated with the source queue are decremented. In one embodiment, any entries decremented to zero are deleted to save space.

[0134] When a new event is tested, each hash table is probed to determine if any of the attributes of the putative new event would cause violation of a li-mit. For each limit found to be exceeded, the corresponding penalty is assessed. No structure modifications are needed for testing a new event and the process can be completed very quickly. Measurements on a typical central processing unit (CPU) appropriate for this purpose indicate that only a few microseconds are required to test each new event.

[0135] The following table contains a typical set of sequence rules. This table contains eight rules, which would result in eight hash tables in the resulting his- tory structure. Similarly, there are four distinct time limits (0.5, 2, 3 and 10 hours) and two distinct play sequence limits (3 and 4 plays). This means that there will be six queues in the history structure cascaded into two chains of length four and two.

Max

Attribute Count Period Unit Penalty artist 4 3 hours 2000 album 3 3 hours 2000 album 2 3 plays 2000 artist 3 4 plays 2000 track 1 2 hours 700 track 1 4 hours 100 track 1 10 hours 50 artist 1 30 minutes 90

Data Flow and Operation

[0136] Referring now to Fig. 2, there is shown a data flow block diagram for one embodiment of the present invention. Behavior of users 201 is monitored, including track selections, track repeating and skipping, and the like. Log server 202 collects user behavior information and stores the information in log database 114, as described above. Log analysis module 113 analyzes the stored behavior information to develop personal profiles, which are stored in profile database 112. Stored per- sonal profiles represent abstracted musical preferences as developed through the relationship discovery techniques of the present invention.

[0137] In one embodiment, a database 205 of Uniform Resource Locators

(URLs), or links, to music-related websites 203 is maintained. A music spider module 204 determines which of such links would be of interest to particular users, based on stored profiles in database 112, as weU as on discovered relationships to artists and tracks that the user has indicated he or she likes. If desired, such links may be presented to individual users, either on website 106 or via e-mails 119 that may be peri- odicaUy generated and transmitted. Such websites 203 may include, for example, e- commerce sites for the sale of compact discs or concert tickets, artist information sites, fan sites, and the like. [0138] In one embodiment, additional databases are provided for storage of event information 207 and offers 209. Administrators 206 and 208 maintain these databases. Based on stored profiles in database 112, as well as on discovered relationships to artists and tracks that the user likes, selected items are extracted from data- bases 207 and 209, and sent to users. Thus, users can be kept informed as to upcoming concerts, events, offers, and the like, for artists that match their personal profiles.

[0139] Entity indexing module 210 processes profile information from database 112 and provides processed information to matching index 211. Matching index 211, which may be implemented in recommendation engine 207, develops rela- tionships and matches among tracks and artists. Queries 213 (which may include any request for information, either from a user or from another module of the system) are provided as input, and results 212 are output, including related tracks and artists.

[0140] Referring now to Fig. 3, there is shown a block diagram depicting an implementation of log and play history analysis according to one embodiment of the present invention. User actions 301, including behavior as described above, are monitored and provided to play log database 114. Four analysis modules 302-305 are provided, for performing various types of analysis on stored information from database 114. Each of modules 302-305 develops a different type of mapping, including user-to-track mapping 302, user-to-artist mapping 303, track-to-artist mapping 304, and track-to-track mapping 305. Thus, user-to-track mapping module 302 discovers relationships between particular users and the music tracks they tend to enjoy the most, while user-to-artist mapping module 303 discovers si-trular relationships between users and artists. Track-to-artist mapping module 304 and track-to-track mapping module 305 discover relationships based on co-occurrence of particular tracks and artists in significant numbers of user track lists. The specific techniques of such relationship discovery will be described in more detail below.

[0141 ] In one embodiment, discovered relationships from modules 302-

305 are stored in profile database 112 (for describing user preferences) and in track profile database 306 (for describing track and artist relationships). In another em- bodiment, discovered relationships are stored in learned artist relationships 1605.

These stored relationships are then used for generating recommendations, and for other appHcations as described herein. For example, a track information window 308 may be provided as part of the user interface for jukebox 103 (or in any other desired format). Window 308 accepts as input a particular track information request, and provides as output a Hst of one or more related tracks, based on track profile database 306. Suggestions from the output Hst may then be used for programming of a personalized radio station, or for other appHcations.

[0142] In addition, a Net Music window 307 may be provided, for offering suggestions or personalized programming based on user profiles. When a request for a recommendation is made, window 307 retrieves user profile information from database 112 and provides recommendations for tracks and/ or artists based on user-to-track or user-to-artist mappings.

[0143] Referring now to Fig. 4, there is shown a block diagram depicting a technique for identifying related music tracks according to one embodiment of the present invention. The technique illustrated in Fig. 4 may be used, for example, in implementing module 305 of Fig. 3. In one embodiment, the steps of Fig. 4 are performed off-line, and results are saved in track profile database 306 for retrieval when needed.

[0144] Track Hst 401 contains aggregated information describing music tracks that have been downloaded by users (i.e., music Hbraries), play logs, repeats, skips, and the Hke. For a particular track, track list 401 can be consulted to determine which individual users have listened to that track the most as a fraction of aU of the music they listen to. The set of such users is indicated as "people who listen" 402 in Fig. 4. The system then determines which other tracks 403 tended to be popular among the users in list 402. Over-represented tracks 404 (i.e., best-seUers that appear on a high proportion of all user track lists) may be found 404 and pruned 405 according to a defined threshold, so that the resultant related tracks 406 captures music tracks that are distinctive and likely to be enjoyed by those who enjoy the tracks from track Hst 401. Related tracks 406 can then be stored in track profiles 306 for later reference in generating recommendations. In one embodiment, related tracks database 406 is implemented as part of learned artist relationships 1605. [0145] The particular techniques for performing the track-to-track association of Fig.4, as weU as refinements thereto, are described below in connection with the operation of the recommendation engine.

[0146] Referring now to Fig. 5, there is shown a block diagram depicting a technique for identifying a mapping between music tracks and artists according to one embodiment of the present invention. The technique iUustrated in Fig. 5 may be used, for example, in implementing module 304 of Fig. 3. In one embodiment, the steps of Fig. 5 are performed off-Hne, and results are saved in track profile database 306 for retrieval when needed. Track Hst 401 contains aggregated information de- scribing music tracks that have been downloaded by users (i.e., music Hbraries), play logs, repeats, skips, and the Hke. For a particular track, the technique of Fig. 4 is ap- pHed to find 501 related tracks 406. Artists for related tracks 406 are identified 502 and stored in related artists database 503 for later reference in generating recommendations. In one embodiment, related artists database 503 is implemented as part of learned artist relationships 1605.

[0147] Referring now to Fig. 6, there is shown a block diagram depicting a technique for identifying a mapping between users and artists according to one embodiment of

present invention. The technique illustrated in Fig. 6 may be used, for example, in implementing module 303 of Fig. 3. In one embodiment, the steps of Fig. 6 are performed off-Hne, and results are saved in user profile database 112 for retrieval when needed. User Hst 601 contains a Hst of users to be analyzed. For each user, tracks that the user has listened to are found 501. Artists for those tracks are identified 502 and stored in related artists database 503. In one embodiment, related artists database 503 is implemented as part of learned artist relationships 1605. As described below, artists may be scored with respect to particular users, in order to provide an indication of the degree of affinity between the user and the artist.

[0148] Referring now to Fig. 7, there is shown a block diagram depicting a technique for identifying a mapping between users and music tracks according to one embodiment of the present invention. The technique of Fig. 7 is used for generating music track recommendations for users, based on discovered relationships between tracks the user has listened to and other tracks with which the user may not be famil- iar. The technique illustrated in Fig. 7 may be used, for example, in implementing module 302 of Fig. 3. In one embodiment, the steps of Fig. 7 are performed off-Hne, and results are saved in user profile database 112 for retrieval when needed. One skilled in the art will note that artist and album recommendations can be made by a process analogous to the described technique for making track recommendations. Artist recommendations can be converted to track or album recommendations by noting which tracks or albums are the most popular for a given artist.

[0149] For a particular user, track information 403 is extracted from play log database 114. A list of tracks is thus obtained. The track Hst is augmented 701 by including additional tracks based on discovered relationships, determined for example using the technique of Fig. 4. Significance scores are associated with the listed tracks. Over-represented tracks may be identified 404 using a statistical test or other means. Low-frequency tracks may be pruned 405 if they have lower than a predefined number of listeners or plays. The resulting Hst is stored in track summary data- base 702. In one embodiment, track summary database 702 is implemented as part of learned artist relationships 1605. Personalized programming, advertising, music track suggestions, and the like, may be generated based on the stored Hst.

[0150] In one embodiment, the techniques depicted in Figs. 4, 5, 6, and 7 are implemented within relationship discovery engine 1604. [0151] Referring now to Fig. 8A, there is shown a block diagram depicting a technique for generating recommendations according to one embodiment of the present invention. The technique of Fig. 8 A may be used, for example, for generating recommendations in real time in response to requests for programming for a personalized radio station. A user ID 801 is obtained, either by user entry of a unique identi- fier (and password, if desired), or by retrieval of a cookie on a user's machine, or by other means. User information is then retrieved from profile database 112, and a profile 802 of recent behavior (including song selections) is obtained. The profile is used as a query to recommendation engine 107. An available inventory 108 of tracks (as weU as other related items) is provided as input to engine 107, along with learned art- ist relationships 1605. As described above, learned artist relationships 1605 is a database of discovered relationships among tracks and/ or artists, based on the relation- ship discovery techniques described herein. Recommendation engine 107 then generates output containing recommended items, including offers 804, events 805, tracks 806, links 807, and the Hke.

[0152] Referring now to Fig. 8B, there is shown a block diagram showing a technique for generating notifications according to one embodiment of the present invention. A Hst of users 601 is provided to notification criteria 115 for selecting which users should receive notifications. Criteria 115 may include, for example, user's stated preferences for receiving notifications, user's purchase threshold as may be determined from past purchasing behavior, length of time since most recent notifi- cation, physical location (e.g., for notification of location-specific events such as concerts), specified artists or related artists, and the like. Learned artist relationships 1605 are provided to recommendation engine 107, which determines which items to recommend to outbound notifier module 116. Current price offers 808, events 809, and the like are provided to outbound notifier module 116. Based on input from reeommendation engine 107, and based on notification criteria 115, module 116 generates e-mails 811 and transmits them to selected users from user Hst 601. E-mails 811 may include, for example, descriptions of special offers 804, events 805, news 810, related links 807, and the Hke. In one embodiment, e-mails 811 may even include selected music tracks or links thereto. [0153] Thus, using the technique illustrated in Fig. 8B, the present invention f aciHtates direct marketing via e-mail, which selectively targets users based on their impHcit and expHcit preferences, as processed through recommendation engine 107 to determine which items are likely to appeal to selected users.

[0154] Referring now to Fig. 9, there is shown a block diagram of a data model 900 according to one embodiment of the present invention. One skilled in the art will recognize that data model 900 is merely one example of an implementation of a data model for the present invention, and that many other organizational schemes and relationship among data files and records may be used without departing from the essential characteristics of the present invention. Accordingly, data model 900 of Fig. 9 is merely intended to be illustrative of a particular embodiment for implementing the invention. [0155] Each component of data model 900 contains fields that are maintained for records in a particular data table. Relationships between components are indicated by connecting lines, with both one-to-many relationships and many-to- many relationships being shown. One skilled in the art will recognize that such ta- bles and relationships can be implemented using any conventional relational database product, such as Oracle.

[0156] Fig. 9 shows the foHowing tables:

[0157] User table 901 for tracking individual users: Fields include user ID

(key field), last version downloaded, ZIP code, IP address, and e-mail address. [0158] Log segment table 902 (in log database 114): In one embodiment, fields include upload time (indicating when the log segment was uploaded) and estimated period (indicating the time period covered by the log segment).

[0159] Log element table 903 (in log database 114) for tracking user actions with regard to music tracks: Fields include action, count, last play, checksum, and track ID.

[0160] Audio source table 904 (in content database 102) for specifying locations of audio files: Fields include checksum (key field) and URL.

[0161] Audio file table 905 (in content database 102) for providing descriptive information regarding audio files: Fields include checksum, header infor- mation, and description.

[0162] Track table 906 (in content database 102) for providing specifics of tracks: Fields include track ID (key field), title, album ID, track number, genre, and description.

[0163] Artist table 907 (in content database 102) for providing artist in- formation: Fields include artist ID (key field) and name.

[0 64] Album table 908 (in content database 102) for providing information about albums: Fields include album ID (key field), publisher, genre, and description.

[0165] User profile table 909 for storing tracks related to users: Fields may include related track, weight, and whether the relationship was expHcitly provided by the user. In situations where user information cannot be extracted from ob- served behavior, such relationships may be provided expHcitly by the user (e.g. by feedback forms).

[0166] Artist expansion table 910 for storing related artists: Fields include related artist, weight, and whether the relationship was expHcitly provided by the user. This table is generated, for example, by relationship discovery engine 1604.

[0167] Track expansion table 911 for storing tracks related to other tracks:

Fields include related track, weight, and whether the relationship was expHcitly provided by the user. This table is generated, for example, by relationship discovery engine 1604. [0168] Album expansion table 912 for storing key tracks on albums:

Fields include related track, weight, and whether the relationship was expHcitly provided by the user. This table is determined by finding tracks that are played more than the average of aU tracks on an album.

[0169] In one embodiment, tables 909-912 are stored in profile database 112; in another embodiment, tables 909-912 are stored in learned artist relationships 1605.

[0170] In the present description of the invention, references to artists, tracks, and albums are interchangeable. Relationships among such entities can be determined and processed according to any desired degree of granularity and descrip- tion.

Indexing

[0171] Referring now to Fig. 13, there is shown a flow diagram of a method of initiaHzing and mamtaining an index in relationship discovery engine

1604. InitiaUy, play logs from database 114 are obtained 1302. Content index 110 is generated and maintained based on log analysis 113. Play log files and music Hbrary files are associated with particular users based on cross-referencing of User IDs

("MMUIDs"). An exemplary file naming convention is {<MMUID>} {<SEQ_NO>}<VERSION> .

[0172] For example: [0173] {00199CEO-8A7D-11D3 -AP7C-OOAOCC3C67B9}{0}4 .30.0058MMD [0174] A filtering program may also be appHed 1303 to the Hst of files to be indexed, in order to:

[0175] Filter files not corresponding to a version on the version "go" Hst

(so as to minimize the impact of users testing on development versions); [0176] Filter files from MMUIDs on a pre-specified "kill" Hst; and

[0177] Filter aU but the log with the largest sequence number for a particular user (to avoid using obsolete data).

[0178] The filtered list of files is read by an indexing script in index and search module 104, which reads each file and adds the play logs to content index 110. Several different indexes can be constructed by the indexing script, depending on whether artist, albums, or tracks are indexed.

[0179] The indexing subsystem is initialized using a command that instructs the subsystem to read initialization files from a directory. The subsystem reads 1304 stop files (artist.stop, album.stop, track.stop), index files (artist.index, album.index, track.index), and track tables (artist.tracks, album.tracks, track. tracks) from the specified location. The stop file contains a list of tracks that should be excluded from the index being initialized.

[0180] The indexing subsystem reads each play log as a file and parses it

1305 according to file type. For example, artist, album, track, and playCount fields are extracted for each record. Parser/ extractors return data in the same format to the indexing subsystem.

[0181] The stop lists are appHed 1306 to filter unwanted entries. Stop Hsts cascade, so that placing an artist on the artist stop Hst prevents all albums and tracks by that artist from indexing. For a finer grain of control, lower level stop Hsts may be used.

[0182] Fields are converted 1307 to aH lowercase and trimmed of leading and trailing white space. Leading "the " is stripped from artists, " & " is converted to

" and " and artists of the form "lastname, firstname" are transformed to "firstname lastname". Additional processing may also be performed, as appropriate. [0183] The output of parsing and cleaning a play log is a list of tracks for each of the indexes (artist, album, and track). [0184] The cleaned Hst of tracks for a play log is added 1308 to the appropriate index in relationship discovery engine 1604. Each track is added to the track table and its occurrence count talHed 1309. Adding a play log to the index includes the foUowing steps: [0185] Obtaining an integer trackID for each track;

[0186] Obtaining an integer play log ID for the play log;

[0187] Creating a Hst of track IDs and a paraUel Hst of occurrence counts for this play log, and storing the lists in the play log index, keyed by the play log ID;

[0188] For each track ID, adding the play log ID and the number of occur- rences of the track in the play log to the two Hsts, listing all play log IDs that a track appears in, and a paraUel Hst containing the occurrence count of the track in each play log; and

[0189] Updating track and play log total counts.

[0190] After aU logs have been added to the index, the indexing subsys- tem prunes 1310 indexes and calculates IDF and normalization weights. Pruning includes removing aU tracks that occur in fewer than a threshold number of play logs. Parallel indexing operations can be performed for albums and artists in addition to tracks.

User Interface [0191] Web site 106 of the present invention provides a front end for communicating recommendations and other results of the invention to users, and for accepting input from users and tracking their behavior. Particular displays and page designs may be implemented using known techniques of web development and database access, incorporating information and recommendations from the various da- tabases of the present invention. In one embodiment, web site 106 includes pages directed toward the foUowing functions and data:

[0192] Lists of new music (context-dependent, filtered and organized by recency of posting); [0193] Lists of "hot picks" (context-dependent, filtered and organized by popularity);

[0194] Browsing functionaHty to aUow the user to browse artists based on categories, discovered relationships, and other Hnks; [0195] Recommendations tailored to the individual user;

[0196] Search functionaHty;

[0197] Links to featured partner sites; and

[0198] Advertising (which may be targeted based on user preferences and discovered relationships). [0199] One skilled in the art will recognize that many other functions, web pages, and interfaces may be provided in connection with the present invention.

[0200] Referring now to Fig. 10A, there is shown a data flow diagram for a browse function according to one embodiment of the present invention. The browse function aUows users to traverse artists and genres by cHcking Hnks representing discovered relationships. Database 102 is populated from commerciaUy available entertainment information databases containing music/ artist/ album descriptions, such as available from Muze Inc. (www.muze.com) or the AU Media Guide (AMG) from Alliance Entertainment Group (www.allmusic.com). Such infor- mation may be provided, for example, in the form of updates 1009 using an import tool 1008 as provided by the database provider. Information for database 102 may also be provided by artist relationships import tool 1011 and content import tool 1007. Content is stored in database 102 in tables, as described above in connection with the data model of Fig. 9. Unmapped artist list 1006 and artist name equivalences 1005 are provided to content import tool 1007 to generate new records for database 102. Page btdlder 1003 queries database 102 for top-level genres, and buUds pages 1004, using HTML templates 1001 for each top-level genre, containing links to sub-genres. Page builder 1003 queries database 102 for each artist and builds a page or set of pages in 1004 for each, thus providing a linked set of pages for traversal by the user. [0201 ] Updates 1009 are provided to import tool 1008 for generating updates to stored data in database 102 in accordance with avaUable third-party software as provided by the database provider. In one embodiment, equivalencing is performed to account for different spellings and variations on artist names, track names, and album titles. In another embodiment, heuristic matching or other techniques are employed as weU. Artist-to-artist relationships 1010, as developed by relationship discovery techniques described herein, are provided to artist relationships import tool 1011 for storage in database 102.

[0202] Referring now to Fig. 10B, there is shown a data flow diagram for a recommendation function according to one embodiment of the present invention. Recommendations pages display selected items based on expHcit preferences or dis- covered relationships from 1605. Such pages thus include functionaHty for suggesting albums that may be purchased on compact disc as weU as downloadable music tracks. When play logs 1024 for the user are avaUable, suggestions are made based on the play logs, using the relationship discovery techniques described below. When play logs 1024 are not available, a user may be given an opportunity to upload a play log 1024 to receive recommendations, or alternatively to receive generic recommendations (such as those based on user demographics or overaU popularity of music tracks or albums). Recommendations may be refreshed and updated whenever a new play log 1024 is received. In addition, some randomness may be incorporated into the recommendations so as to increase variety and encourage repeat visits to the web site. [0203] Jukebox 103 periodicaUy uploads play logs to play log database

114. If jukebox 103 has obtained any additional relevant information regarding the user, this information may also be uploaded at this time. PeriodicaUy, the system retrieves a Hst of users from profile database 112 for which new play logs are avaUable, and module 1021 determines representative suggestions for each user. The represen- tative suggestions are stored in profile database 112. When the user accesses the suggestion page, representative items are fetched and used to formulate recommendations, using the relationship discovery techniques described herein. If no representative items are avaUable for the user, the play log for that user (if avaUable) is analyzed so that representative items may be determined. Based on the formulated recom- mendations, and using a format specified in HTML templates 1001, online page buUder 1003 generates output web pages 1004 for presentation to the user as part of web site 106.

Operation of Relationship Discovery Engine 1604

[0204] As described above, the present invention employs relationship discovery engine 1604, in connection with learned artist relationships 1605, to find related items for generation of suggestions, track Hsts, and the Hke. Referring now to Fig. 14, there is shown a flow diagram of a method of operation for relationship discovery engine 1604 according to the present invention. A query is formed 1402 using one or more tracks, artists, or albums, either from a user's play log or from another source. The query may specify tracks, artists, or any other relevant criteria.

[0205] Based on the suppHed query, a Hst of relevant users 1403 is obtained. In general, this Hst includes users that have played the specified tracks, or who have played music by the specified artist, and the Hst is ordered by the relative prominence of the track or artist in the user's play log. In one embodiment, step 1403 is performed by weighting the tracks in the query using one of several weighting strategies. A list of users having play logs that include one or more of the query tracks is obtained using an inverted index in play log database 114. The matching tracks from each play log are weighted according to the selected play log weighting scheme. If a query track is absent in the play log, its weight is zero. The score of the user with respect to the query is the sum across aU query tracks of the query weight multipHed by the user's play log weight for each track.

[0206] Play logs for the most significant users are obtained 1404. The tracks in the retrieved play logs are merged, ranked and filtered 1405 by statistical techniques to return the most relevant items. Alternatively, artists or albums for the tracks in the retrieved play logs are determined, and the artist Hst is merged, ranked and fUtered. The resulting Hst contains the related tracks, albums, or artists for the specified query.

[0207] Many types of music retrievals are possible using this system. By using the user's play log as the input for query in step 1402, the method of Fig. 14 dis- covers relationships based on the observed behavior of the user. [0208] In one embodiment, the present invention employs a binomial log

HkeHhood ratio analysis for finding significantly over-represented tracks, albums or artists in a set of retrieved play logs. The log HkeHhood ratio is a measure of how weU a nuU hypothesis fits the observed data. If the nuU hypothesis is the assumed inde- pendence of occurrence of two tracks, for example, the log HkeHhood ratio measures the likelihood that such independence is a vaHd assumption. It foUows, then, that the log HkeHhood ratio is a useful indicator of the relationship between the occurrences of the two tracks, if any.

[0209] The log HkeHhood ratio is based on a HkeHhood ratio. A HkeHhood ratio is the ratio of the maximum HkeHhood of the observed data for aU models where the nuU hypothesis holds to the maximum HkeHhood of the observed data for aU models where the nuU hypothesis may or may not hold. The log likelihood ratio is the logarithm of the HkeHhood ratio.

[0210] For the present invention, the log HkeHhood ratio is employed to determine whether a given track is more Hkely to appear in track Hsts of a first subset of users than in track Hsts of a second subset of users. Based on this measure, subsets of users are defined so as to identify those users most likely to enjoy the track, album, or artist.

[021 ] In one embodiment, the log HkeHhood ratio is appHed in the pre- sent invention to determine whether a particular track occurs more frequently than expected in the selections of a subset of users. Variables are defined as foUows:

[0212] N = the total number of users;

[0213] Ni = the number of users in the subset;

[0214] N₂ = the number of users not in the subset; [0215] kn = the number of users in the subset that selected the track;

[0216] kι₂ = the number of users not in the subset that selected the track;

[0217] k₂₁ = Ni - kn = the number of users in the subset that did not select the track; and

[0218] kz> = N₂ - kι₂ = the number of users not in the subset that did not select the track

[0219] The foUowing equations are appHed:

[0221] The log HkeHhood ratio is then given as:

[0222] LLR for the track = ∑ k_y log - ^πv

[0223] Referring now to Fig. 15, there is shown a flow diagram of a method of extracting significant information according to the present invention. The method illustrated in Fig. 15 is shown in terms of matching tracks in a music recommendation system. One skilled in the art wiU recognize that the method may be adapted and appHed to many other domains and techniques.

[0224] A total number of users Ν is determined 1502. A total number of tracks S is determined 1503. For each track, the system determines 1504 a track frequency (the number of times the track was played by aU users, or SE, = ∑n_y ) and a

Hstener frequency (the number of users that listened to the track at least once, or LF_j = ^T (n_tj > 0) ). The results are weighted 1505 according to a product of up to

three components: α = how many times the user has listened to the particular track; β = how rare the track is among aU users; and γ = a normalizing factor based on how many tracks the user has Hstened to, in total.

[0225] The first weighting factor, α, represents the frequency of the track within the user's play log. It may be represented and defined according to the foUowing alternatives: [0226] a_τ = k_y = Number of occurrences of the track in the user's play log; or

[0227] a_L = logk_y (or log(k_y + 1) ); or

[0228] a_χ = 1 (a constant, used if this weighting factor is not to be considered). [0229] α may be adjusted to account for repeat play, aborted play, high or low volume level, and the like. Other functions are also possible and are weU known in the Hterature describing information retrieval. [0230] The second weighting factor, β, represents the frequency of the track within aU users' play logs. It may be represented and defined according to the foUowing alternatives:

N + l [0231 ] β_j = log (inverse Hstener frequency, i.e. the log of the num-

LF_j + 1 ber of users divided by the number of users that listened to the track); or

[0232] βχ~ ^ (^a constant, used if this weighting factor is not to be considered).

[0233] β may be adjusted in a similar manner as is α.

[0234] The third weighting factor, γ, represents a normalizing factor, which serves to reduce the bias for scoring long play logs higher than short ones. Using a normalizing factor, a short relevant play log should score at least as weU as a longer play log with general relevance, γ may be represented and defined according to the foUowing alternatives:

[0235] γ_c = where S, = β and W„ = α; or

[0236] γ _x = 1 (a constant, used if this weighting factor is not to be considered).

[0237] By employing the above-described combination of three weighting factors in generating scores for tracks and artists, and then finding 1506 significantly over-represented elements using a test Hke the generalized log-likelihood ratio test, the present invention avoids the problems of overstating "best seUers" (i.e. those items that appeal to nearly aU users) and overstating coincidental co-occurrence. If a track is a best seUer, the second weighting factor wiU tend to diminish its overpowering effect. In addition, the effect of coincidental co-occurrence is lessened by the γ coefficient. [0238] In one embodiment, the system of the present invention generates scores as foUows. For each track of interest, a large m-dimensional vector is determined. For each Hstener, another m-dimensional vector is determined. The techniques of assigning meaning to such vectors and training the vector set to represent simUarities among vectors are weU known in the art, as described for example in Sal- ton et al., "The SMART information retrieval system," 1983. In such a scheme document weights can be defined as [0239] w^ βχ [0240] where k_y is as defined above, i is the document and j is the term.

[0241] Query weights _y. can defined where k_y now represents the word

counts. Given these document and query weights, the score for each user log is:

[0242] score_l = w_y. q_j j

[0243] A score can be generated for each listener's play logs relative to a query, and the highest-scoring listeners can be added to the Hstener Hst. A score for a Hstener with respect to a query is determined by taking the dot product of the query vector and the vector for a Hstener's play logs. In one embodiment of the present invention, the above-described weighting factors are appHed to the vector terms in order to improve the results of the scoring process. [0244] Once play logs have been scored for retrieval using weighting factors, play logs are retrieved, based on the relationships to the query. These play logs contain artists, albums, and/ or tracks. Over-represented artists, albums, and/ or tracks are extracted based on measured significance using the log HkeHhood ratio. These over-represented items are output as recommendations. [0245] Once the resultant tracks have had their significance measured, a subset of tracks, albums or artists in the resulting play-logs is output 1507 as recommendations. The subset may be determined by taking a fixed number of the top- scoring play logs and/ or by taking aU play-logs that have a higher score than a threshold value. In either case, the generalized log-likelihood ratio test can be used to find tracks, albums or artists that are significantly over-represented in this subset of play-logs relative to the entire set of aU play-logs. These over-represented items constitute a recommendation set. In this manner, the present invention is able to provide recommendations that are most Hkely to be of interest to the particular user. [0246] Further processing of the output of engine 1604 may be provided, in order to filter the results. For example, tracks that the user has aheady played may be omitted from recommended tracks. Alternatively, some tracks that have already been played may be included, so as to improve the credibUity (from the user's point of view) of the output results. Output may be ranked in order of score, or may be randomized and further ffltered, in order to obtain a desired level of variety in suggested tracks. Output may be proved to recommendation engine 107 for presentation to the user.

[0247] Referring now to Fig. 11, there is shown an example of a screen shot 1100 depicting sample artist-level relationships. Query term 1101 is shown, along with Hst 1102 of recommended artists, generated by engine 107. For each recommended artist, screen 1100 depicts a score as weU as the name of the artist; higher- scoring artists are those that have a closer discovered relationship to query term 1101. [0248] The present invention is able to refine the discovered relationships and user preferences as often as desired. For example, user behavior may be monitored after recommendations are made, so that play logs can be updated based on the user's selection of tracks, as weU as the user's skipping and/ or repeating of tracks. In one embodiment, more recent behavior may be assigned a greater weight than previous behavior. In this manner, the present invention provides a technique for continu- aUy updating user preference data, so as to take into account changing tastes or moods.

[0249] By making suggestions based on observed behavior with respect to music track selections, the above-described methods of the present invention avoid many of the limitations of the prior art. SpecificaUy, the user data may be dynami- caUy updated with each track selection, so that more data points are avaUable than in prior art schemes. By contrast to online commerce environments where user behavior may be monitored only when the user chooses to make a purchase (or, at best, when he or she browses a title), the present invention is able to monitor individual track selections and thus achieve a much greater degree of granularity. In other words, user preference data may be coUected at a higher bandwidth than in prior art systems. [0250] In addition, users' selection of music tracks is for their own personal enjoyment; such selections are not generaUy made on behaU of other people (as might be the case in online stores, where a user may purchase a gift for some other person). Thus, the developed user preferences, embodied in the user play logs, are more Hkely to accurately reflect the user's tastes.

[0251 ] FinaUy, play logs may include information as to which tracks were repeated, which were aborted or skipped, and at what volume level the tracks were played. Weights can be assigned to tracks in the log, based on such observations. For example, the system may assign a higher weight to a track that was repeated on the assumption that the user probably enjoyed that track, whUe a lower weight may be assigned to a track that was skipped halfway through, on the assumption that the user probably did not enjoy the track.

Applications

[0252] The above-described methods for implementing relationship dis- covery engine 1604 generate output that may be used for a variety of appHcations. In addition to generating artist and track recommendations based on a user's play log, the present invention may be employed for the foUowing appHcations as weU:

[0253] Recommendations based on explicit preferences: Input to engine

1604 may be presented in terms of the user's specified preferences, such as may be obtained via an online questionnaire. Such input may be employed to supplement data describing observed behavior, so as to dinunish the undesired effect of best seUers and other less-meaningful influences.

[0254] Improved text searches: Input to engine 1604 may be a text search term for a particular artist or track. Output may then include tracks and artists that engine 1604 deems Hkely to be of interest. Thus, a user may search for artist A and be presented with works by artist B as weU, based on a relationship between artists A and B that is discovered by analysis of user listening behavior. Alternatively, such relationships may be determined in advance and stored in database records, so that textual searches for tracks and artists can return information about related tracks and artists based on the stored fields in the database records. Such an appHcation may be particularly useful, for example, in an online commerce environment.

[0255] Improved text searches may alternatively be implemented by augmenting the pages to be searched by including tags for related artists or tracks. Conventional search engines wiU then automaticaUy include the pages in search results for the related artists or tracks, without any additional processing.

[0256] Personalized radio station programming: In another appHcation, the present invention may be employed in connection with conventional radio station programming techniques, to implement an improved personalized radio station. As is known in the art, conventional radio stations typicaUy divide a programming block into a number of segments. Each segment is assigned a programming category, such as "power hit," "new release," "recurrent hit," and the Hke. For a particular programming block, music tracks are assigned to each of the segments based on the particular programming format of the radio station. Music scheduling software, such as Selector® by RCS Sound Software, appHes heuristic rules for repetition limits and classes of songs, to automaticaUy generate track Hsts for use by radio stations. The present invention may be combined with such existing radio station programming techniques, to populate the defined segments with music tracks that are Hkely to appeal to a particular listener. Additional rules may be appHed in generating track Hsts, so as to Hmit undesired repetition and to comply with limiting legislation (such as the Digital MUlennium Copyright Act) and other restrictions.

[0257] To implement such an appHcation including a personalized radio station using suggestions from engine 1604, the present invention uses slot definitions (which may be generated manuaUy or by a software appHcation), together with de- scriptive information for each track, to generate a Hst of candidate tracks for each defined slot. Tracks are then ranked, based on several factors including the output of engine 1604. Ranked order may then be perturbed to a specified degree, in order to introduce a selected level of randomness to the resultant program. For each slot, a track from the ranked Hst is selected, either by strict rank-selection, or by a rank- weighted randomization. [0258] In one embodiment, selections for each defined slot are generated as foUows. A "penalty" value is associated with playing each track at a particular time. For example, playing a power track during a power slot might carry a penalty of zero, whUe playing a gold track during a power slot might carry a penalty of 1000 points. Other penalty values would simUarly be estabHshed. The penalty value would then be combined with track scores to generate a ranked list of preferred tracks.

[0259] Randomness can also be added so as to provide variety and unpredictability. A random number can be generated within the range [0,1). The score might then be adjusted by -μ log(l-u), where μ is a scale factor.

[0260] Additional constraints, restrictions, and rules might be added, in order to influence track selection and arrangement. For example, point values for a track might be reduced by 2000 if the track is played more than twice per hour, or if more than three tracks from a particular artist are played within an hour. Such con- straints may be appHed for aesthetic reasons, or to comply with Digital MUlennium Copyright Act requirements, or for any other reason.

[0261] Once the score is estabHshed, penalties appHed, and randomness appHed, the track having the smaUest penalty (or largest score) is selected and added to the track list. The above-described appHcation for implementing radio station pro- gramming provides distinct benefits over the prior art technology described previously. Traditional prograrruning techniques involving selection and placement of slots are combined with the advantages of user personalization, to implement an improved personalized radio Hstening experience.

[0262] Advertisement targeting: Once relationships between music tracks and/ or artists have been developed, users may be presented with ads that are most Hkely to be of interest to them. Particular ads may be associated with particular tracks, albums, or artists, and relationships among tracks, albums, or artists may then be exploited using the output of engine 1604 of the present invention. In one embodiment, such an appHcation may be implemented by generating keywords describ- ing user preferences (based on the output of engine 1604), and providing such key- words to conventional ad purchasers, so that the selected advertisements are selected based on the discovered keywords.

[0263] One skilled in the art wiU recognize that, in addition to the above appHcations, many other appHcations of the present invention may be contemplated. For example, in an online commerce environment where users may browse albums or other products, advertisements may be targeted to particular users based on discovered relationships between the products being browsed and other products that are Hkely to be of interest. In addition, user behavior regarding web surfing, volume levels of music tracks, repeats and skips, and/ or any other observable behavior, may be used as input to engine 1604. Weights can be assigned to different types of behaviors.

Sample User Interface

[0264] For Ulustrative purposes, a number of user interface elements, including menus, commands, dialog boxes, and screens, are described below. These user interface elements provide an example of an implementation of the present in- vention in the context of an online jukebox appHcation 103, as may be made avaUable over the Internet. One skiUed in the art wiU recognize that the particular functions, commands, layouts, and design of the illustrated user interface are merely exemplary of such an appHcation. Many other arrangements, features, and designs are possible. Accordingly, the foUowing description and accompanying drawings are in no way intended to limit the scope of the invention, which scope is defined solely by the claims herein.

[0265] Referring now to Fig. 12, there is shown a screen shot 1200 of main components for a jukebox 103 interface according to one embodiment. Jukebox 103 provides buttons for skipping and repeating tracks 1202, volume control 1201, track display information 1203, and track program Hst 1204. A listing of the tracks in the user's music Hbrary 1205 is also provided, along with controls 1206 for adding, deleting, and reorganizing the list. Media window 1207 may also be provided, for displaying current song visualizations 1207 A, album cover art 1207B for the currently playing or a related album, or other artwork 1207C. A miniaturized version 1208 of a player window may also be provided upon activation of a mini-player button 1209, to provide a subset of the features and controls of main screen 1200.

[0266] Referring now to Figs. 17A through 17C, there are shown additional main components for a sample user interface of a jukebox 103 that implements the present invention.

[0267] Fig. 17A depicts a "Now Playing" screen 1700, which provides information describing and related to a musical track that is currently playing. Information displayed within screen 1700 may be provided from a web page, for example. Such information may include, for example, a track Hsting 1701 for the currently play- ing album, a Hsting of the most popular tracks 1702 for the currently playing album, a Hst of album recommendations 1703 (as may be determined using the above- described techniques of the present invention), and a link 1704 to an online radio station that may be personalized according to the preferences of the individual Hstener, using techniques described above. Additional information, advertisements, and con- trols may also be displayed in various areas of screen 1700.

[0268] Fig. 17B depicts a detached track Hsting 1711 that shows an exploded view of the information in track Hsting 1701. Detached Hsting 1711 may be activated by a user control 1710. Radio station screen 1712 provides access to a number of online radio stations, as Hsted 1713. The user can activate any selected online radio station, or may create (i.e. configure) a new station by activating control 1714 and interacting with broadcast radio window 1715 for selecting parameters for a new station.

[0269] Fig. 17C depicts a music guide screen 1720. Screen 1720 contains additional information related to the currently playing track or to other musical selec- tions that the user may be interested in. Information may include articles 1721 as weU as access to personalized recommendations 1722 that may be determined using the above-described techniques of the present invention. Articles 1721 may be selected by reference to artists, albums, or tracks that the system of the present invention determines are Hkely to be of interest to the user. [0270] Referring now to Fig. 18, there is shown a series of menus 1801-

1805 for a sample user interface of a jukebox that implements the present invention. Menus 1801-1805 of Fig. 18 may be avaUable, for example, in a menu bar as part of screen 1200 of the user interface. The user may select items from menus 1801-1805 to activate various commands and functions of the online jukebox, including those related to the present invention. The particular menus of Fig. 18, which are merely ex- emplary, include FUe menu 1801, Edit menu 1802, View menu 1803, Options menu 1804, and Help menu 1805. Several commands and screens related to menus 1801- 1805 wUl be described in more detaU below, for Ulustrative purposes.

[0271] Referring now to Figs. 19A and 19B, there are shown various interface elements for FUe menu 1801 items. Open command 1901 activates an Open Mu- sic screen 1901A for navigating among and selecting files containing music tracks, such as may be located on the user's hard drive, or on a compact disc, or the Hke. Convert command 1902 activates a FUe Format Conversion screen 1902A for converting files from one format to another, using techniques that are known in the art. Add New Track(s) to Music Library command 1903 activates an Add Tracks to Music Li- brary screen 1903A for adding music tracks, found on hard drives, compact discs, and the Hke, to the user's Hbrary as shown in 1205.

[0272] Open Music Library command 1904 activates Open Music Library screen 1904A for navigating among and selecting music Hbrary fUes. Music Hbrary fUes may be selected and opened by the user to provide a set of music tracks. Print command 1905 activates Print screen 1905A for printing various Hsts, tracks, and H- braries. Export PlayHst Tracks command 1906 activates Export Playlist Tracks screen 1906A for converting and/ or exporting tracks from playHsts to other formats and locations. Create CD from Playlist command 1907 activates Create CD from Playlist screen 1907A for providing access to features for creating compact discs from selected playlists. Exit command 1908, 1908A exits the appHcation.

[0273] Referring now to Figs. 20A, 20B, and 20C, there are shown various interface elements for Edit menu 1802 items of a sample user interface of a jukebox that implements the present invention. PlayHst Track Tag(s) command 2001 activates Edit Track Tag(s) screen 2001 A that aUows a user to view and edit descriptive infor- mation concerning a particular track. Screen 2001A contains tabs 2031, 2032, 2033,

2034, and 2038 for accessing various subscreens as shown in Figs. 20B and 20C. Gen- eral tab 2031 provides access to subscreen 2031 A, which provides fields and controls for entering general information concerning the track, including track title, track number, artist, album, genre, and the Hke. Genre field 2036 is presented as a pulldown menu 2035A for selecting among genres. Preference field 2037 is presented as a puU-down menu 2036A for selecting the user's degree of Hking of the track. Find Art FUe button 2011 A activates Open screen 2021A for browsing a hard drive or other sources for artwork related to the track. The user may select an artwork file using screen 2021 A, and the software then associates the selected artwork with the track. The artwork may then appear in media window 1207, if desired. Copy to CHpboard command 2004, which is accessible from Edit menu 1802 as weU as from subscreen 2031 A, copies the artwork to the operating system cHpboard, so that it may be pasted in other appHcations as desired. Paste from CHpboard command 2005, which is accessible from Edit menu 1802 as weU as from subscreen 2031 A, pastes artwork that was previously stored in the operating system cHpboard to subscreen 2031 A, thereby associating the artwork with the track. Remove Art button 2006 removes the artwork from association with the track. Load Album button 2011C loads an entire album into screen 2001 A. Select AU in Playlist command 2002 selects aU the tracks in the current playlist, as shown in screen 2002A. Clear PlayHst command 2003 removes aU tracks from the current playlist, as shown in screen 2003A. [0274] Lyrics tab 2032 provides access to subscreen 2032A, which provides a field for viewing and editing lyrics for the track. Notes tab 2033 provides access to subscreen 2033 A, which provides a field for viewing and editing notes for the track. Bios tab 2034 provides access to subscreen 2034A, which provides a field for viewing and editing biographical information for the track. [0275] More tab 2038 provides access to subscreen 2038A, which provides fields for viewing and editing additional information and characteristics describing the track. Subscreen 2038A contains Tempo field 2040 which aUows selection from menu 2040A, Mood field 2041 which aUows selection from menu 2041A, and Situation field 2042, which aUows selection from menu 2042A. [0276] Referring now to Figs. 21A, 21B, 21C, 21D, 21E, and 21F, there are shown various interface elements for View menu 1803 items of a sample user inter- face of a jukebox that implements the present invention. SmaU Player View command 2101 activates miniaturized version 1208 of the player window. FuU Player View command 2102 activates fuU-sized player window 1200. My Library command 2103 shows the user's music Hbrary 1205. [0277] MusicMatch Radio command 2104 activates radio screens 2104A and 2104B for operating and confroUing a personalized online radio station. Music Guide command 2105 activates Music Guide screen 2105A that displays information, offers, and recommendations related to the currently playing track. Now Playing command 2106 activates Now Playing screen 2106A showing track listing and other information related to the currently playing track. Recorder command 2107 activates Recorder screen 2107A providing controls for making recordings of tracks and track Hsts. Media Window command 2108 activates Media Window screen 2108A containing media window 1207 for displaying artwork, graphics, and other material. Buy CD Site command 2109 provides access to e-commerce web page 2109A where the user may purchase music related to the currently playing track.

[0278] Visualizations command 2110 provides access to functionaHty for presenting visual accompaniments to tracks being played (not shown). Sound Enhancement command 2111 provides access to controls for altering and enhancing the sound presentation (not shown). Auto Arrange Components command 2112 toggles between free-form arrangement 2112A of windows and structured arrangement

2112B. Always on Top command 2113 keeps the jukebox appHcation on top of other windows, as shown in 2113A.

[0279] Referring now to Figs. 22A, 22B, and 22C, there are shown various interface elements for Option menu 1804 items of a sample user interface of a jukebox that implements the present invention. Player command 2201 provides access to various commands described below in connection with Fig. 23D. Playlist command 2202 provides access to various commands described below in connection with Figs. 23E and 23F. Music Library command 2203 provides access to various commands described below in connection with Figs. 24A through 24C. Recorder command 2204 provides access to various commands described below in connection with Figs. 25A through 25E. Add New Features command 2205 activates or provides access to screen 2205A for downloading and installing plug-ins providing additional functionaHty for the jukebox appHcation.

[0280] Get Music Recommendations command 2206 activates music recommendations screen 2206A, which provides recommendations based on observation of user behavior, as described above. Update Software command 2207 activates Software Update screen 2207A, which provides functionaHty for downloading and instaUing the latest release of the cHent software in response to user instructions.

[0281 ] Change Skin command 2208 activates Change Skin screen 2208A, which provides alternatives for "skins," or themes for decorative user interface ele- ments for selection by the user, as is known in the art. Download Skins command 2209 activates Download Skins screen 2209A, which aUows the user to access, download, and instaU additional "skins" as desired.

[0282] Change Text Size command 2210 activates Change Text Size screen

(not shown), which provides functionaHty for changing the size of text displayed in various user interface screens. Settings command 2211 provides access to Settings screens 2211 A-2211E, which aUow the user to specify various settings and preferences for operation of the software appHcation.

[0283] General Settings screen 2211A aUows the user to specify various general settings. In 2301, the user may specify which file types are to be played by the software appHcation. In 2302, the user may specify the result of a double-cHck action. In 2303, the user may specify settings for dow oading music fUes. In 2304, the user may specify whether a QuickPlay function is enabled in the System Tray. In 2305, the user may specify permission settings for communication with the central server.

[0284] Player Settings screen 2211B aUows the user to specify various set- tings concerning the player appHcation. In 2306, the user may specify seek increments and song skip increments. In 2307, the user may specify whether the media window appears on first play. In 2308, the user may specify the mixer to be used. In 2309, the user may enable and configure a waUpaper function that converts album art to background waUpaper. [0285] Recorder Settings screen 2211C aUows the user to specify various settings concerning recording of music. In 2310, the user may specify and configure the recording quaHty. Button 2311 activates a navigation screen (not shown) for accessing a songs directory. Button 2312 activates a screen (not shown) for specifying advanced features. Referring also to Fig. 23G, button 2313 activates Delayed Recording screen 2313A for specifying delayed recordings. Button 2314 activates Digital Rights Management screen 2314A for configuring security attributes. In 2315, the user may enable and configure the creation of song cHps. In 2316, the user may specify the recording mode for compact disc recording.

[0286] Music Library screen 2211D aUows the user to specify various settings concerning the music Hbrary. In 2318, the user may specify display settings. In 2319, the user may specify tag updates. In 2320, the user may specify which tag is to be used when conflicts occur. In 2321, tag conversion may be enabled.

[0287] CDDB/ Connectivity screen 2211E aUows the user to specify various settings concerning compact disc database connectivity. In 2322, the user can enable the CDDB album lookup service. In 2323, the user can specify and configure the connection to the central server.

[0288] Referring now to Fig. 23D, there are shown various screens and menus associated with Player command 2201 of Options menu 1804. Player command 2201 provides access to Player submenu 2201F, which contains Play Control command 2201 A, Play Cycle command 2201B, Play Reordering command 2201C, EquaHzer command 2201D, and Settings command 2201E. Play Control command 2201A provides access to Play Control submenu 2201G, which contains commands related to the operation of the player appHcation. Play Cycle command 2201B provides access to Play Cycle submenu 2201H, which aUows the user to select between single play ("once") and repeated play ("repeat"). Play Reordering command 2201C provides access to Play Reordering submenu 2201 J, which aUows the user to select how tracks are to be reordered. EquaHzer command 2201D activates Equalizer screen 2201K containing controls for a graphic equalizer. Settings command 2201F provides access to Settings screen 2211 A as described above in connection with Fig. 23 A. [0289] Referring now to Figs. 23E and 23F, there are shown various screens and menus associated with PlayHst command 2202 of Options menu 1804.

Playlist command 2202 provides access to PlayHst submenu 2202E, which contains Open Music command 2202A, AutoDJ command 2202B, Save PlayHst command 2202C, and Clear PlayHst command 2202D. Open Music command 2202A activates Open Music screen 2202F, which allows the user to open files containing music, located on a hard drive, remote server, compact disc, and the Hke. AutoDJ command 2202B activates AutoDJ screen 2202G, which aUows the user to specify various criteria for adding musical selections to the music Hbrary. As seen in Fig. 23F, screen 2202G includes entry fields for specifying total play time, album preference, artist preference, genre preference, tempo preference, and the Hke. The software appHcation retrieves tracks corresponding to the specified preferences. Save PlayHst command 2202C activates Save PlayHst screen 2202H, which aUows the user to specify a name and location for the saved playHst file. Clear PlayHst command 2202D clears the user's playlist.

[0290] Referring now to Figs. 24A through 24C, there are shown various screens and menus associated with Music Library command 2203 of Options menu 1804. Music Library command 2203 provides access to Music Library submenu 2203Q, which contains commands 2203A through 2203P, as described below.

[0291 ] New Music Library command 2203 A activates screen 2203R for specifying the name and location of a new music Hbrary to be created. Open Music Library command 2203B activates Open screen 2203S for navigating among stored f es and folders and indicating a music Hbrary file to be opened. Save Music Library As command 2203C activates Save Music Library screen 2203T for specifying a name and location for a music Hbrary to be saved. Clear Music Library command 2203D presents confirmation screen 2203U aUowing the user to confirm that the currently open music Hbrary is to be cleared. [0292] Export Music Library command 2203E activates Export screen

2203W for specifying the name, location, and file type for an exported copy of the music Hbrary. This command thus provides functionaHty for generating, storing, and fransmitting music Hbrary files in any of a number of fUe formats. FUe excerpt 2203Y illustrates an example of a line of an exported file in a text format, as may be gener- ated and saved in connection with Export Music Library command 2203E. Import

Music Library command 2203F activates Import screen 2203V for specifying the name, location, and file type for a fUe to be imported as a music Hbrary. This command thus provides functionaHty for accessing music Hbrary fUes in any of a number of fUe formats. Add New Track(s) to Music Library command 2203G activates Add Tracks to Music Library screen 2203X, which provides functionaHty for identifying individual tracks, as may be stored on a hard drive, server, compact disc, or the like, to be added to the music Hbrary.

[0293] Delete Track(s) command 2203H presents confirmation screen

2203Z aUowing the user to confirm that the selected track or tracks are to be deleted from the user's database. The user may also specify whether the associated song fUe or files should be removed from the user's computer. Edit Track Tag(s) command 2203J activates Edit Track Tag(s) screen 2403 providing functionaHty similar to screen 2001A described above in connection with Fig. 20B. Find Track(s) in Music Library command 2203K activates Find screen 2401 providing functionaHty for keyword searches in the user's music Hbrary. Search and Add Track(s) from AU Drives com- mand 2203L activates Search for Music screen 2402 providing functionaHty for searching the user's computer for digital music files so that the files may be added to the user's music Hbrary.

[0294] Preview Track command 2203M plays a track in a preview mode.

Add Track(s) to PlayHst command 2203N adds selected tracks to the user's current playHst. Music Library Settings command 2203P activates Music Library screen 2211D as described above in connection with Fig. 23B.

[0295] Referring now to Figs. 25A and 25B, there are shown various screens and menus associated with Recorder command 2204 of Options menu 1804. Recorder command 2204 provides access to Recorder submenu 2204F, which includes Control command 2204A, Source command 2204B, QuaHty command 2204C, Send Album info to CDDB command 2204D, and Settings command 2204E. Control command 2204A provides access to submenu 2204G containing various commands related to control of the recorder. Source command 2204F provides access to submenu 2204J containing commands for selecting the source to be recorded, including for ex- ample a CD, line in, microphone in, and the Hke. QuaHty command 2204F provides access to submenu 2204H containing commands for specifying the format and quaHty level of the recording to be made.

[0296] Send Album info to CDDB command 2204D activates screen

2204K, which displays results of a search for database records matching the track be- ing recorded. The user is given an opportunity to confirm the match, and, in Submit screen 2204L, to modify the information being transmitted. Settings command 2204E activates Recorder Settings screen 2211C described above in connection with Fig. 23B. [0297] Referring now to Figs. 26A through 26D, there are shown various screens and user interface elements for implementing a personalized radio station ac- cording to the techniques of one embodiment of the present invention. Screen 2600 provides controls for initializing a personalized radio station by accepting three favorite artists from the user. Alternatively, the user may initialize a personalized radio station based on the user's Hstening profile; this option may be specified in section 2305 of General Settings screen 2211A, as described above in connection with Fig. 23 A. One advantage to this alternative method is that the user's history of music selections provides a more accurate profile of the user's preferences.

[0298] Create New Station screen 2601 provides functionaHty for configuring the personalized radio station. The user can select a Station Match function 2602, which aUows the user to match existing predefined radio stations and to mix genres from two or more predefined stations. The user can also select an Artist Match function 2603, which provides musical selections based on the user's input regarding his or her favorite artist, as determined using the above-described techniques of the present invention. Input controls are also provided for naming the station 2604, launching the station 2605, and deleting the station 2606. [0299] The user may also e-maU a Jink to the newly created station to another user, such as a friend. Screen 2104B provides various controls related to the operation of the personalized radio station. Send to Friend button 2609 activates screen 2607 for providing an e-maU address and message. The software appHcation sends an e-mail message 2608 to the specified recipient, and includes a link to the personalized radio station. The recipient can then Hsten to the personalized radio station by cHcking on the link. [0300] Screens 2610 and 2611 provide functionaHty for selecting among predefined radio stations. The user can browse among various formats, as shown in screen 2610, or may view search results in screen 2611, based on a keyword search. The functionaHty of screens 2610 and 2611 may be used by the user to select two or more predefined radio stations to be combined to generate a personalized radio station.

Stream Delivery

[0301] As described above, the relationship discovery engine of the present invention may be implemented in conjunction with a personalized online radio station. In one embodiment, music is deHvered to users in a streamed audio format. For example, radio sequence transmitter 121 may deHver units of data to jukebox 103 in a format wherein each unit encodes a period of music. Since radio stations typicaUy repeat their programming several times, it is beneficial to cache the data units in order to reduce the amount of transmitted data. In addition, if a sufficiently large time scale is used, different channels of the radio station may have considerable overlap among currently playing selections that are being deHvered to various users. By identifying these common units, transmitter 121 can take advantage of further economies of transmission, so as to provide more efficient deHvery of audio data. [0302] Using known compression methods, FM-quaHty music deHvery can be provided with a bandwidth of approximately 32,000 bits per second, and AM- quaHty music deHvery can be provided with a bandwidth of approximately 20,000 bits per second. CD-quaHty music deHvery can be provided with a bandwidth of approximately 128,000 bits per second. Conventional channel capacities for users' Internet connections range from approximately 14,400 to 56,000 bits per second for dial-up modems, to one miUion (or more) bits per second for cable modems and

ADSL connections. Channel capacities can vary from moment to moment, depending on current network conditions. VariabUity is particularly evident in shared access environments, such as LAN-based or cable modem connections. Thus, audio deHvery as provided by transmitter 121 is, in one embodiment, designed to function despite such variations in channel capacities from user to user and from moment to moment. [0303] In one embodiment, transmitter 121 employs scalable coding to increase the quaHty of audio output despite limitations in channel capacity. Audio data is categorized so that low-quaHty audio can be produced using the primary information, whUe secondary information can be combined with the primary information to enhance output quaHty. In one embodiment, additional levels of information may also be provided, each of which can be combined with the lower levels to further enhance output quaHty. Thus, by caching lower-quaHty audio and later combining it with subsequently received secondary information, jukebox 103 is able to increase the quaHty of the audio output. [0304] SpecificaUy, the first time an audio track is transmitted, transmitter

121 provides jukebox 103 with the primary information first. Secondary (and additional) information is transmitted as time permits. Jukebox 103 outputs the audio track with whatever level of information it has received at the time output is to commence. If only primary information has been received, jukebox 103 outputs lower- quaHty audio. If secondary information has been received, it is combined with the primary information and jukebox 103 outputs higher-quaHty audio.

[0305] In addition, jukebox 103, in one embodiment, caches the received information. If the same audio track is requested at a later time, transmitter 121 provides jukebox 103 with the next level of information. Therefore, even if jukebox 103 was unable to provide higher-quaUty audio during the first listening, it may be able to provide higher-quaHty audio during subsequent listenings, by combining secondary (and/ or additional) information with the previously cached primary information to generate the higher-quaHty audio output. Such a technique f ac itates the output of high quaHty audio even when network transmission capacities are limited. [0306] Referring now to Fig. 27 A, there is shown an example of a transfer sequence for a channel with moderate bandwidth. InitiaUy, tracks A and B are requested. Primary information for track A 2701 is downloaded. As primary information 2701 is downloaded, a low-quaHty version of track A 2705 is played, according to conventional sfreaming audio techniques. Downloaded primary information 2701 is cached. [0307] Once the download of primary information for track A 2701 is complete, jukebox 103 begins to download primary information for track B 2702. This download may begin even though frack A is stUl playing 2705. In the example shown in Fig. 27 A, the download of primary information for track B 2702 is completed whUe track A is stUl playing 2705. Therefore, jukebox 103 begins to download secondary information for track B 2703. Then, when playback 2705 of track A is finished, jukebox 103 is able to output a high quaHty version of frack B 2706, by combining secondary information 2703 with previously downloaded primary information 2702. The output of the high quaHty version 2706 may take place whUe secondary information 2703 is stUl being downloaded, again using streaming techniques.

[0308] In the example of Fig. 27 A, a request to play track A a second time is received. Therefore, once secondary information 2703 has been downloaded, jukebox 103 begins to download secondary information for track A 2704. Once the high quaHty version of track B 2706 is finished playing, jukebox 103 outputs a high quaHty version of track A 2707, by combining secondary information 2704 with previously downloaded primary information 2701.

[0309] Referring now to Fig. 27B, there is shown another example of a transfer sequence for a channel with a lower bandwidth than that of Fig. 27A. Here, the secondary information for frack B 2703 is not downloaded, because it would not arrive in time to improve the output of track B. Accordingly, a lower quaHty version of track B 2708 is output in Heu of the higher quaHty version 2706 of Fig. 27A. However, the higher quaHty version of track A 2707 can stUl be presented, since there is sufficient time to download secondary information for track A 2704 before the second playback of track A commences. [0310] One skiUed in the art wiU recognize that the tracks depicted in

Figs. 27A and 27B may refer to individual songs, or song segments, or any other unit of information. One skiUed in the art wUl further recognize that the scalable coding techniques described herein may be appHed to video data, or to any other type of data, and are not limited to audio data. [0311 ] The scalable coding techniques of the present invention thus facilitate the trading off of quaHty in bandwidth-limited situations, without requiring complex bandwidth estimation and determination. If insufficient bandwidth exists for the deHvery of higher-quaHty versions, the system simply continues playing lower quaHty versions of tracks. No skipping, pausing, or other interruption of the audio stream is necessary. Jukebox 103 can determine whether to continue any particular transfer to improve the avaUable quaHty or to download the next requested track, based on upcoming track selections. At any given moment, the next data segment to request can be determined by requesting the highest priority data segment from the next few audio segments. In one embodiment, priorities are defined to either play audio at a maximum short-term quaHty level or at a consistent quaHty level.

[0312] In one embodiment, jukebox 103 requests data for downloading according to the foUowing order of priorities:

[0313] One skiUed in the art wiU recognize that any desired priority Hst may be provided. For example, if item 5 in this table is moved up to the third rank, the system will give more priority to high quaHty presentation at the possible expense of inconsistent quaHty on lower bandwidth connections.

[0314] In one embodiment, locaUy-cached downloaded data is stored in an encrypted or otherwise protected form, so as to prevent its abuse and to inhibit copyright infringement. In another embodiment, primary information is stored in an encrypted or otherwise protected form, but secondary and subsequent information is not, since the secondary and subsequent information is unusable without access to the primary information. [0315] In one embodiment, jukebox 103 downloads audio files when the user is not actuaUy listening to music, so as to f acUitate improved usage of an otherwise idle network connection. Jukebox 103 determines which items are Hkely to be requested by a user, so that at idle times it can transfer data that is likely to be useful for rendering audio segments in the future. Such determination may be made, for example, using the learned artist relationships described above, in order to "guess" which tracks the user is most likely to request in the future. In one embodiment, secondary information for such "predicted" audio segments is downloaded first, so that encryption is not required unless and until the user actuaUy requests the tracks and the primary information is to be downloaded.

[0316] Scalable coding may also be used to process, a signal of a conventional broadcast radio station that plays music. An audio recognition device, as is conventional, pre-processes the signal in order to identify individual songs. Those portions of audio information that are not music are compressed and stored, and a transfer sequence is sent to jukebox 103 that references these recently encoded non- music segments as weU as previously known and cached musical segments. The recently encoded segments can be encoded at a lower quaHty level in order to aUow a jukebox 103 connected by a low speed line to transfer the recently encoded segments in real-time whUe stUl playing the cached musical segments at a higher quaHty level.

[0317] From the above description, it wUl be apparent that the invention disclosed herein provides a novel and advantageous system and method for relationship discovery. The foregoing discussion discloses and describes merely exemplary methods and embodiments of the present invention. As wiU be understood by those famUiar with the art, the invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. For example, the invention may be appHed to other domains and environments, and may be employed in connection with additional appHcations where personalized recommendations are desirable. Accordingly, the disclosure of the present invention is intended to be illus- trative, but not limiting, of the scope of the invention, which is set forth in the foUowing claims.

Claims

[0318] What is claimed is:

1. A computer-implemented method of discovering relationships between items, comprising: accepting item selections from a pluraHty of users; generating a log for each user, each log containing identifiers for the user's item selections; accepting a query including at least one query item identifier; scoring the user logs, responsive to a degree of occurrence of the at least one query item identifier in the user logs, to generate user log scores; and deterrruning at least one result item, responsive to a degree of occurrence in at least a subset of ihe scored user logs.

2. The computer-implemented method of claim 1, wherein a significance of the occurrence is determined by a log HkeHhood ratio analysis and the determined result is responsive to the determined significance.

3. The computer-implemented method of claim 1, wherein a significance of the occurrence is determined by a substantial equivalent of a log HkeHhood ratio analysis and the determined result is responsive to the determined significance.

4. The computer-implemented method of claim 1, wherein each item is a video track and wherein accepting item selections comprises determining which tracks are selected for playback.

5. The computer-implemented method of claim 1, wherein each item is a music track and wherein accepting item selections comprises deterrnining which tracks are selected for playback.

6. The computer-implemented method of claim 5, further comprising: generating a track Hst containing an identifier for each determined result item comprising a music track.

7. The computer-implemented method of claim 6, further comprising: deleting from the track Hst at least one identifier corresponding to a music track aheady selected by the user.

8. The computer-implemented method of claim 6, further comprising: playing the music tracks specified by the generated track Hst.

9. The computer-implemented method of claim 5, further comprising: accepting a format schedule specifying music track categories for time periods; and generating a frack Hst coriforming to the format schedule and containing an identifier for each determined result item comprising a music track.

10. The computer-implemented method of claim 5, wherein scoring the user logs comprises determining a degree of occurrence in each user log of at least one music frack identified by the query item identifier.

11. The computer-implemented method of claim 5, wherein scoring the user logs comprises determining a degree of occurrence in each user log of at least one music track associated with an artist identified by the query item identifier.

12. The computer-implemented method of claim 1, wherein accepting item selections comprises receiving input provided by a user via a web page.

13. The computer-implemented method of claim 1, wherein accepting item se- lections comprises receiving input specifying an item purchase by a user.

14. The computer-implemented method of claim 1, further comprising, prior to determining the at least one result item, defining the subset of the scored user logs responsive to the user log scores.

15. The computer-implemented method of claim 1, further comprising: monitoring user behavior with respect to the selected items; and adjusting the user log responsive to the monitored user behavior.

16. The computer-implemented method of claim 15, wherein monitoring user behavior comprises at least one selected from the group consisting of: detecting user input requesting that a selected item be repeated; detecting user input requesting that a selected item be skipped; detecting user input specifying a volume change; and detecting user input specifying that a selected item be muted.

17. The computer-implemented method of claim 1, wherein accepting item se- lections comprises receiving input provided by a user via an appHcation for playing tracks.

18. The computer-implemented method of claim 1, wherein accepting a query comprises receiving a user log containing identifiers for a user's item selections.

19. The computer-implemented method of claim 1, wherein accepting a query comprises receiving a first search term, the method further comprising: generating a second search term containing an identifier for each determined result item.

20. The computer-implemented method of claim 19, further comprising at least one of: providing the second search term as input for a search engine; and adding the second search term to a searchable portion of a document associated with the first search term.

21. The computer-implemented method of claim 1, further comprising: periodicaUy uploading the generated log.

22. The computer-implemented method of claim 1, further comprising: outputting an advertisement relating to the determined at least one result item.

23. The computer-implemented method of claim 22, wherein outputting an advertisement comprises displaying at least one selected from the group consisting of: a web page; a banner; a portion of a web page; and an animation.

24. The computer-implemented method of claim 1, further comprising: outputting a notification relating to the determined at least one result item.

25. The computer-implemented method of claim 24, wherein outputting a notification comprises displaying a web page.

26. The computer-implemented method of claim 24, wherein outputting a notification comprises sending a communication to a user.

27. The computer-implemented method of claim 26, wherein sending a communication to a user comprises at least one selected from the group consisting of: fransmitting an electronic maU message to the user; telephoning the user; and sending a direct maU item to the user.

28. The computer-implemented method of claim 1, wherein the determined result is responsive to a significance of the occurrence of the item in at least a subset of the scored user logs, and wherein the significance is determined by a log HkeHhood ratio analysis submethod comprising: determining a total number of user logs N; detern ining a number of user logs Ni in a subset of user logs; determining a number of user logs N₂ not in the subset of user logs; determining a number of user logs kn in the subset that include the item; determining a number of user logs kι₂ not in the subset that include the item; deterrnining a number of user logs k₂ι = Ni - kn in the subset that do not include the item; determining a number of user logs h.22 = N₂ - kι₂ not in the subset that do not include the item; and determining a log HkeHhood ratio for tire item.

29. The computer-implemented method of claim 28, wherein the log HkeHhood ratio is defined as:

where: π„ = —,μ, = y,— . N. ^J X^* N

30. The computer-implemented method of claim 29, further comprising: adjusting at least one of the kij values responsive to at least one selected from the group consisting of: the number of occurrences of the item in a user log; the logarithm of the number of occurrences of the item in a user log; the number of occurrences of the item in aU user logs; the logarithm of the total number of users divided by the number of users who have selected the item; and a normalizing factor.

31. The computer-implemented method of claim 30, wherein the normalizing

factor is where S, is a weight based on the number of occurrences of

the item in aU user logs and W is a weight based on the number of occurrences of the item in a particular user log.

32. The computer-implemented method of claim 1, further comprising: deleting from the determined at least one result item any result items already selected by a user associated with the query.

33. The computer-implemented method of claim 1, further comprising: ranking the at least one result item responsive to the degree of significance.

34. A computer-implemented method of discovering a relationship between a first item and a second item, comprising: determining a total number of item groups N; determining a number of item groups Ni in a subset of item groups, the subset of item groups being defined as including those item groups that contain a second item; determining a number of item groups N₂ not in the subset of item groups; determining a number of item groups kn in the subset that contain the first item; determining a number of item groups kι₂ not in the subset that contain the first item; determining a number of item groups kn = Ni - kn in the subset that do not contain the first item; determining a number of item groups k2_ = N₂ - kι₂ not in the subset that do not contain the first item; and determining a log HkeHhood ratio.

35. The computer-implemented method of claim 34, wherein the log HkeHhood ratio is defined as:

ii V"* ii where: π„ - —^L-,μ , = — .

" N_j ¹ -Y N

36. The computer-implemented method of claim 35, wherein each item group comprises a document.

37. The computer-implemented method of claim 35, further comprising: adjusting at least one of the kij values responsive to at least one selected from the group consisting of: the number of occurrences of the item in a document; the logarithm of the number of occurrences of the item in a document; the number of occurrences of the item in aU documents; the logarithm, of the total number of documents divided by the number of documents that include the item; and a normalizing factor.

38. The computer-implemented method of claim 37, wherein the normalizing

factor is where S, represents the number of occurrences of the item in

aU documents and W_y represents the number of occurrences of the item in a particu- lar document.

39. A system for discovering relationships among items, comprising: a user interface for accepting item selections from a pluraHty of users; at least one log database, coupled to the user interface, for storing a log for each user, each log containing identifiers for the user's item selec- tions; a query input device for accepting a query including at least one query item identifier; and a recommendation engine, coupled to the log database and to the query input device, for scoring the user logs, responsive to a degree of occur- rence, to generate user log scores, and for determining at least one result item, responsive to a degree of occurrence in at least a subset of the scored user logs.

40. The system of claim 39, wherein the significance of the occurrence is determined by a log HkeHhood ratio analysis and the recommendation engine determines the at least one result item responsive to the determined significance.

41. The system of claim 39, wherein the significance of the occurrence is deter- mined by a substantial equivalent of a log HkeHhood ratio analysis and wherein the recommendation engine determines the at least one result item responsive to the determined significance.

42. The system of claim 39, wherein each item is a video track and wherein the user interface accepts item selections by determining which tracks are selected for playback.

43. The system of claim 39, wherein the user interface accepts item selections by determining which tracks are selected for purchase.

44. The system of claim 39, wherein each item is a music frack and wherein the user interface accepts item selections by determining which tracks are selected for playback.

45. The system of claim 44, wherein the user interface comprises an online jukebox.

46. The system of claim 45, wherein the online jukebox monitors user behavior with respect to the selected items and adjusts the user log scores responsive to the monitored user behavior.

47. The system of claim 46, wherein the online jukebox monitors user behavior by detecting at least one selected from the group consisting of: user input requesting that a selected item be repeated; and user input requesting that a selected item be skipped; and user input specifying a volume change; and user input specifying that a selected item be muted.

48. The system of claim 47, further comprising: a track Hst generator, coupled to the recommendation engine, for generating a track list containing an identifier for each determined result item comprising a music track.

49. The system of claim 44, further comprising: a music player, coupled to the frack list generator, for playing the music fracks specified by the generated frack Hst.

50. The system of claim 44, further comprising: a format scheduler, for accepting a format schedule specifying music track categories for time periods; and a track Hst generator, coupled to the recommendation engine and to the format scheduler, for generating a frack Hst conforming to the format schedule and containing an identifier for each determined result item comprising a music track.

51. The system of claim 39, wherein the query input device receives a user log containing identifiers for a user's item selections.

52. The system of claim 39, wherein the query input device receives a first search term, the system further comprising: a search term generator, coupled to the recommendation engine, for generating a second search term containing an identifier for each determined result item and for providing the second search term as input for a search engine.

53. The system of claim 39, wherein the query input device receives a first search term, the system further comprising: a search term generator, coupled to the recommendation engine, for generating a second search term containing an identifier for each determined result item and for providing the second search term to be added to a searchable portion of a document associated with the first search term.

54. The system of claim 39, further comprising: an advertisement output device, coupled to the recommendation engine, for outputting an advertisement relating to the determined at least one result item.

55. The system of claim 54, wherein the advertisement output device displays at least one selected from the group consisting of: a web page; a banner; a portion of a web page; and an animation.

56. The system of claim 39, further comprising: a notification output, coupled to the recommendation engine, for outputting a notification relating to the determined at least one result item.

57. The system of claim 56, wherein the notification output device displays at least one selected from the group consisting of: a web page; a banner; a portion of a web page; and an animation.

58. The system of claim 56, wherein the notification output device sends a communication to a user.

59. A computer-readable medium comprising computer-readable code for discovering relationships between items, comprising: computer-readable code adapted to accept item selections from a pluraHty of users; computer-readable code adapted to generate a log for each user, each log containing identifiers for the user's item selections; computer-readable code adapted to accept a query including at least one query item identifier; computer-readable code adapted to score the user logs, responsive to a degree of occurrence of the at least one query item identifier in the user logs, to generate user log scores; and computer-readable code adapted to determine at least one result item, responsive to a degree of occurrence in at least a subset of the scored user logs.

60. The computer-readable medium of claim 59, wherein a significance of the occurrence is determined by a log HkeHhood ratio analysis and the determined result is responsive to the determined significance.

61. The computer-readable medium of claim 59, wherein a significance of the occurrence is determined by a substantial equivalent of a log HkeHhood ratio analysis and the determined result is responsive to the determined significance.

62. The computer-readable medium of claim 59, wherein each item is a video track and wherein the computer-readable code adapted to accept item selections comprises computer-readable code adapted to determine which tracks are selected for playback.

63. The computer-readable medium of claim 59, wherein each item is a music frack and wherein the computer-readable code adapted to accept item selections comprises computer-readable code adapted to determine which tracks are selected for playback.

64. The computer-readable medium of claim 63, further comprising: computer-readable code adapted to generate a track Hst containing an identi- fier for each determined result item comprising a music track.

65. The computer-readable medium of claim 64, further comprising: computer-readable code adapted to delete from the track Hst at least one identifier corresponding to a music track aheady selected by the user.

66. The computer-readable medium of claim 64, further comprising: computer-readable code adapted to play the music fracks specified by the generated track Hst.

67. The computer-readable medium of claim 63, further comprising: computer-readable code adapted to accept a format schedule specifying music track categories for time periods; and computer-readable code adapted to generate a track Hst conforming to the format schedule and containing an identifier for each determined result item comprising a music frack.

68. The computer-readable medium of claim 63, wherein the computer- readable code adapted to score the user logs comprises computer-readable code adapted to determine a degree of occurrence in each user log of at least one music frack identified by the query item identifier.

69. The computer-readable medium of claim 63, wherein the computer- readable code adapted to score the user logs comprises computer-readable code adapted to determine a degree of occurrence in each user log of at least one music track associated with an artist identified by the query item identifier.

70. The computer-readable medium of claim 59, wherein the computer- readable code adapted to accept item selections comprises computer-readable code adapted to receive input provided by a user via a web page.

71. The computer-readable medium of claim 59, wherein the computer- readable code adapted to accept item selections comprises computer-readable code adapted to receive input specifying an item purchase by a user.

72. The computer-readable medium of claim 59, further comprising, computer- readable code adapted to, prior to determine the at least one result item, define the subset of the scored user logs responsive to the user log scores.

73. The computer-readable medium of claim 59, further comprising: computer-readable code adapted to monitor user behavior with respect to the selected items; and computer-readable code adapted to adjust the user log scores responsive to the monitored user behavior.

74. The computer-readable medium of claim 73, wherein the computer- readable code adapted to monitor user behavior comprises at least one selected from the group consisting of: computer-readable code adapted to detect user input requesting that a selected item be repeated; computer-readable code adapted to detect user input requesting that a selected item be skipped; computer-readable code adapted to detect user input specifying a volume change; and computer-readable code adapted to detect user input specifying that a selected item be muted.

75. The computer-readable medium of claim 59, wherein the computer- readable code adapted to accept item selections comprises computer-readable code adapted to receive input provided by a user via an appHcation for playing tracks.

76. The computer-readable medium of claim 59, wherein the co puter- readable code adapted to accept a query comprises computer-readable code adapted to receive a user log containing identifiers for a user's item selections.

77. The computer-readable medium of claim 59, wherein the computer- readable code adapted to accept a query comprises computer-readable code adapted to receive a first search term, the computer-readable medium further comprising: computer-readable code adapted to generate a second search term containing an identifier for each determined result item.

78. The computer-readable medium of claim 77, further comprising at least one of: computer-readable code adapted to provide the second search term as input for a search engine; and computer-readable code adapted to add the second search term to a searchable portion of a document associated with the first search term.

79. The computer-readable medium of claim 59, further comprising: computer-readable code adapted to periodicaUy upload the generated log.

80. The computer-readable medium of claim 59, further comprising: computer-readable code adapted to output an advertisement relating to the determined at least one result item.

81. The computer-readable medium of claim 80, wherein the computer- readable code adapted to output an advertisement comprises computer-readable code adapted to display at least one selected from the group consisting of: a web page; a banner; a portion of a web page; and an animation.

82. The computer-readable medium of claim 59, further comprising: computer-readable code adapted to output a notification relating to the determined at least one result item.

83. The computer-readable medium of claim 82, wherein the computer- readable code adapted to output a notification comprises computer-readable code adapted to display a web page.

84. The computer-readable medium of claim 82, wherein the computer- readable code adapted to output a notification comprises computer-readable code adapted to send a communication to a user.

85. The computer-readable medium of claim 84, wherein the computer- readable code adapted to send a communication to a user comprises at least one se- lected from the group consisting of: computer-readable code adapted to transmit an electronic maU message to the user; computer-readable code adapted to telephone the user; and computer-readable code adapted to send a direct maU item to the user.

86. The computer-readable medium of claim 59, wherein the determined result is responsive to a significance of the occurrence of the item in at least a subset of the scored user logs, and wherein the computer-readable code adapted to determine a binomial log HkeHhood ratio for an item comprises computer-readable code adapted to determine the result by a log HkeHhood ratio analysis submethod.

87. The computer-readable medium of claim 86, wherein the computer- readable code adapted to determine the result by a log HkeHhood ratio analysis sub- method comprises: computer-readable code adapted to determine a total number of users N; computer-readable code adapted to determine a number of users Ni in a subset of users; computer-readable code adapted to determine a number of users N₂ not in the subset of users; computer-readable code adapted to determine a number of users kn in the subset that selected the item; computer-readable code adapted to determine a number of users k₁₂ not in the subset that selected the item; computer-readable code adapted to determine a number of users k₂ι = Ni - kn in the subset that did not select the item; computer-readable code adapted to determine a number of users kn = N₂ - kι₂ not in the subset that did not select the item; and computer-readable code adapted to determine a log HkeHhood ratio for the item.

88. The computer-readable medium of claim 87, wherein the log HkeHhood ra- tio is defined as:

∑V°^§-

where: π„ = —,//, = Y!— .

^J N. ^^J N

89. The computer-readable medium of claim 59, wherein the computer- readable code adapted to determine the result by a log HkeHhood ratio analysis sub- method further comprises: computer-readable code adapted to adjust at least one of the r j values responsive to at least one selected from the group consisting of: the number of occurrences of the item in a user log; the logarithm of the number of occurrences of the item in a user log; the number of occurrences of the item in aU user logs; the logarithm of the total number of users divided by the number of us- ers who have selected the item; and a normalizing factor.

90. The computer-readable medium of claim 89, wherein the normalizing factor is where S, is a weight based on the number of occurrences of the

item in aU user logs and W.. is a weight based on the number of occurrences of the item in a particular user log.

91. The computer-readable medium of claim 59, further comprising: computer-readable code adapted to delete from the determined at least one result item any result items aheady selected by a user associated with the query.

92. The computer-readable medium of claim 59, further comprising: computer-readable code adapted to rank the at least one result item responsive to the degree of significance.

93. A computer-readable medium comprising computer-readable code for discovering a relationship between a first item and a second item, comprising: computer-readable code adapted to determine a total number of item groups

N; computer-readable code adapted to determine a number of item groups Ni in a subset of item groups, the subset of item groups being defined as including those item groups that contain a second item; computer-readable code adapted to determine a number of item groups N₂ not in the subset of item groups; computer-readable code adapted to determine a number of item groups ku in the subset that contain the first item; computer-readable code adapted to determine a number of item groups k₁₂ not in the subset that contain the first item; computer-readable code adapted to determine a number of item groups k₂ι =

Ni - kn in the subset that do not contain the first item; computer-readable code adapted to determine a number of item groups k^ =

N₂ - ki2 not in the subset that do not contain the first item; and computer-readable code adapted to determine a log HkeHhood ratio.

94. The computer-readable medium of claim 93, wherein the log HkeHhood ratio is defined as:

where: π„ - — , μ , = Y — .

" N_j ^J N

95. The computer-readable medium of claim 93, wherein each item group comprises a document.

96. The computer-readable medium of claim 93, further comprising: computer-readable code adapted to adjust at least one of the -ky values responsive to at least one selected from the group consisting of: the number of occurrences of the item in a document; the logarithm of the number of occurrences of the item in a document; the number of occurrences of the item in aU documents; the logarithm of the total number of documents divided by the number of documents that include the item; and a normalizing factor.

97. The computer-readable medium of claim 96, wherein the normalizing factor is where S, represents the number of occurrences of the item in aU

documents and W represents the number of occurrences of the item in a particular document.