JP2007233515A - Information processor and information processing method, information providing apparatus and information providing method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To generate taste information accurately showing user's taste. <P>SOLUTION: A viewing processing section 57 generates viewing information that is information relating to viewing of a music piece. A numerical value information gain calculation section 61 calculates an information gain of numerical information that is information to be indicated by a numerical value relating to the music piece on the basis of the viewing information. A non-numerical value information gain calculation section 62 calculates information gain of non-numerical information that is information to be indicated by a non-numerical value relating to the musical piece on the basis of the viewing information. A decision tree generation section 63 generates a decision tree showing the user's taste on the basis of the information gain of the numerical value and that of the non-numerical value information. The present invention is applicable, for example, to a personal computer recommending music content. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an information processing device, an information processing method, a providing device, a providing method, and a program, and in particular, an information processing device and an information processing method capable of accurately predicting and recommending content that a user likes. , A providing apparatus, a providing method, and a program.

  In recent years, with the spread of networks such as the Internet, the types and number of contents sold on the networks are increasing. On the other hand, user preferences are also diversified, making it difficult for users to search for and purchase content they like from a vast amount of content sold on the network. . Note that the content is data such as images and music.

  Traditionally, rankings such as hit charts have been used as leading information sources for content purchases. However, as described above, the types and numbers of content sold are increasing, so it can be found in hit charts. It ’s becoming difficult to find and buy content you like. Therefore, there is a need for a recommendation device that recommends content that the user likes.

  As a method for recommending user-preferred content, for example, there is a method for predicting a preference level that represents a user's preference for content that is a candidate for recommendation and recommending the content based on the preference level.

  In this method, for example, the preference level of the recommendation candidate is predicted based on the keyword included in the character string in the EPG information of the program to be the recommendation candidate and the preference level calculated in advance for the keyword. A recommendation list in which recommendation candidates are arranged in descending order is created (for example, see Patent Document 1).

  Note that information representing the user's preference for content includes the decision tree shown in FIG. 1 in addition to the preference level. FIG. 1 shows a decision tree that represents a user's preference for music.

  As shown in FIG. 1, the decision tree has a tree structure in which the discrimination nodes 1-1 and 1-2 are connected to the class nodes 3-1 to 3-3 via the branches 2-1 to 2-4. ing. The discriminating nodes 1-1 and 1-2 are nodes that ask questions about roots and nodes, and the branches 2-1 to 2-4 represent individual answers (cases) to the question. Further, class nodes 3-1 to 3-3 correspond to leaves and represent classes into which discrimination targets are classified.

  The recommendation device can determine (classify) which class the music belongs to by applying the attribute (or attribute value) of the music to the decision tree.

  In FIG. 1, a class node “like” 3-1 is connected to a discrimination node “genre?” 1-1 through a branch “jazz” 2-1. Accordingly, music pieces whose attribute value of the attribute “genre” is “jazz” are classified into a class “like” which is a class of music that the user likes.

  The discrimination node “genre?” 1-1 is connected to the discrimination node “artist?” 1-2 via the branch “classic” 2-2, and is connected to the discrimination node “artist?” 1-2. Is connected to the class node “like” 3-2 via the branch “winding chord”. Therefore, among the music pieces having the attribute value “genre” of “classic”, the music piece having the attribute value “artist” of “winding one” is classified into the class “like”.

  Furthermore, a class node “I dislike” 3-3 is connected to the discrimination node “Artist?” 1-2 through a branch “Hanako”. Accordingly, among the music pieces having the attribute value of “genre” and “classic”, the music piece having the attribute value of “artist” is “Hanako” and is a class that is a class of music that the user does not like (dislikes). Classified as "I hate".

  Next, a decision tree generation process in which the recommendation device generates the decision tree of FIG. 1 will be described with reference to the flowchart of FIG.

The recommendation device includes an attribute value v A _′ (in this case, jazz, classic, orchestra, and song Hanako) for each song attribute A ′ (in this case, genre, artist) stored in itself. ) And information indicating whether or not each piece of music is viewed is stored.

  In step S1, the recommendation device determines whether or not all the songs included in the set C ′ are the same. In the first step S1, the set C ′ is a set of all music stored in the recommendation device.

When it is determined in step S1 that the presence or absence of viewing of all the songs included in the set C ′ is not the same, the recommendation device calculates the information gain _GA ′ of each attribute A ′ and proceeds to step S3. In step S3, the recommendation device selects an attribute A′max corresponding to the maximum information gain G _A′max among the attributes A ′ corresponding to each information gain _{GA ′} calculated in step S2, and step S4. Proceed to

  In step S4, the recommendation device creates a discrimination node “attribute A′max?”. For example, when the attribute A′max is the attribute “genre”, the recommendation device creates the discrimination node “genre?” 1-1 as shown in FIG.

After the processing in step S4, the process proceeds to step S5, where the recommendation device generates a branch for each attribute value v _{A′max of the} attribute A′max, and determines the determination node “attribute A′max?” Created in step S4. Connecting. For example, as illustrated in FIG. 1, the recommendation apparatus has a branch “jazz” 2-1 with an attribute value “jazz” of an attribute “jazz” having an attribute A′max and a branch “classic” with an attribute value “classic”. 2-2 is generated and connected to the discrimination node “genre?” 1-1.

In step S6, the recommendation device sets a subset C _vA′max for each attribute value v _A′max of the attribute _A′max to the set C _′ . After the process of step S6, the process returns to step S1, and the processes of steps S2 to S7 are performed for each subset C _vA′max . That is, the algorithm of the decision tree generation process is a recursive algorithm.

  On the other hand, if it is determined in step S1 that all the songs included in the set C ′ are viewed, the process proceeds to step S7, and the recommendation device generates a class node based on the presence / absence of viewing. , Connect to the branch corresponding to the set C ′.

For example, in the case where viewing of all the songs in the subset C _vA′max of the attribute value “jazz” of the attribute “genre” that is the attribute A′max is possible (all the songs are viewed), the recommendation device is As shown in FIG. 1, the node “like” is generated and connected to the branch “jazz” corresponding to the subset C _vA′max . In addition, when the viewing of all the songs in the subset C _vA′max of the attribute value “jazz” of the attribute “genre” that is the attribute A′max is not viewed (all the songs are not viewed), the recommendation device As shown in FIG. 1, the node “dislikes” is generated and connected to the branch “Hanakako” corresponding to the subset C _vA′max .

  In the process of step S7 after the process of the first step S1, there is no branch corresponding to the set C ′, so the class node is not connected anywhere.

Japanese Patent Application Laid-Open No. 2004-7502

  However, in the method described in Patent Document 1, the preference level for the information represented by numerical values among the information on the content is not calculated, and it is difficult to accurately predict the preference level of the recommendation candidate. As a result, there are cases where content that is preferred by the user is not recommended.

  The present invention has been made in view of such circumstances, and makes it possible to accurately predict and recommend content that a user likes.

  An information processing apparatus according to a first aspect of the present invention is viewing information generation means for generating viewing information, which is information regarding viewing of content, and information represented by numerical values regarding the content based on the viewing information. Calculation means for calculating information gain of numerical information, calculation means for calculating information gain of non-numeric information, which is information represented by non-numeric values, based on the viewing information, and information of the numerical information Preference information generating means for generating preference information representing the user's preference based on the gain and the information gain of the non-numeric information;

  The information processing apparatus is transmitted from the providing apparatus based on the preference information and a transmission unit that transmits the preference information to a providing apparatus that provides content information that is information related to the user's favorite content. Further, receiving means for receiving the content information and presentation means for presenting the content information to a user can be further provided.

  The information processing method according to the first aspect of the present invention generates viewing information that is information related to viewing of content, and based on the viewing information, information gain of numerical information that is numerical information related to the content And calculating an information gain of non-numeric information, which is information expressed in non-numeric values, based on the viewing information, and an information gain of the numerical information and an information gain of the non-numeric information And generating the preference information representing the user's preference.

  The program according to the first aspect of the present invention generates viewing information, which is information relating to viewing of content, and calculates an information gain of numerical information, which is information represented by numerical values, regarding the content, based on the viewing information. Then, based on the viewing information, an information gain of non-numeric information that is non-numeric information related to the content is calculated, and based on the information gain of the numerical information and the information gain of the non-numeric information Then, the computer is caused to execute a process including a step of generating preference information representing the user's preference.

  The providing device according to the second aspect of the present invention is non-numeric information relating to content and receiving means for receiving the preference information from an information processing device that generates preference information representing the user's preference. Selection means for selecting content preferred by the user based on numerical information, numerical information about the content expressed by numerical values, and the preference information; and content information that is information related to the content selected by the selection means. And providing means for providing to the information processing apparatus.

  The providing device further includes storage means for storing ranking information, which is information related to content ranking, and the selection means is configured to select a user's preference among the contents ranked in the ranking based on the ranking information. Content to be selected.

  According to a second aspect of the present invention, there is provided a non-numeric information that is information expressed by non-numeric values regarding information received from the information processing apparatus that generates the preference information that represents user preferences. A step of selecting content preferred by the user based on numerical information represented by numerical values and the preference information relating to the content, and providing the information processing device with content information that is information relating to the selected content including.

  The program according to the second aspect of the present invention receives the preference information from an information processing device that generates preference information representing the user's preference, and includes non-numeric information that is information represented by a non-numeric value regarding the content, A step of selecting content that the user likes based on numerical information about the content expressed by numerical values and the preference information, and providing content information that is information about the selected content to the information processing apparatus. Causes the computer to execute the processing that includes it.

  In the first aspect of the present invention, viewing information, which is information relating to viewing of content, is generated, and based on the viewing information, an information gain of numerical information, which is information represented by numerical values, regarding content is calculated. Based on the information, the information gain of the non-numeric information, which is information expressed in non-numeric values, is calculated based on the information, and the preference of the user is expressed based on the information gain of the numerical information and the information gain of the non-numeric information. Preference information is generated.

  In the second aspect of the present invention, preference information is received from an information processing device that generates preference information that represents a user's preference. Based on the numerical information represented by numerical values and the preference information, the content that the user likes is selected, and the content information that is information about the selected content is provided to the information processing apparatus.

  As described above, according to the first aspect of the present invention, preference information representing user preferences can be generated. Moreover, according to the 1st side surface of this invention, the preference information which represents a user's preference correctly can be produced | generated.

  Furthermore, according to the 2nd side surface of this invention, the content which a user likes can be recommended. Further, according to the second aspect of the present invention, it is possible to accurately predict and recommend the content that the user likes.

  Embodiments of the present invention will be described below. Correspondences between the constituent elements of the present invention and the embodiments described in the specification or the drawings are exemplified as follows. This description is intended to confirm that the embodiments supporting the present invention are described in the specification or the drawings. Therefore, even if there is an embodiment which is described in the specification or the drawings but is not described here as an embodiment corresponding to the constituent elements of the present invention, that is not the case. It does not mean that the form does not correspond to the constituent requirements. Conversely, even if an embodiment is described here as corresponding to a configuration requirement, that means that the embodiment does not correspond to a configuration requirement other than the configuration requirement. It's not something to do.

The information processing apparatus according to the first aspect of the present invention (for example, the client shown in FIG. 5)
Viewing information generating means (for example, the viewing processing unit 57 in FIG. 5) for generating viewing information that is information related to content viewing;
Calculation means (for example, the numerical information gain calculation unit 61 in FIG. 5) that calculates the information gain (for example, the numerical information gain Gn) of the numerical information that is the numerical information related to the content based on the viewing information. When,
Based on the viewing information, calculation means (for example, the non-numeric information gain calculation unit 62 in FIG. 5) that calculates information gain of non-numeric information that is information represented by non-numeric values regarding the content;
And information gain of the numerical information, the information gain of a non-numerical information (e.g., non-numeric information gain G _A) based on the preference information generation means for generating preference information indicating preferences of the user (e.g., decision trees) (For example, the decision tree generation unit 63 in FIG. 5).

Secondly, the information processing apparatus according to the first aspect of the present invention includes:
Transmission means (for example, the transmission unit 54 in FIG. 5) that transmits the preference information to a providing device (for example, the server 13 in FIG. 13) that provides content information that is information related to the content that the user likes;
Receiving means for receiving the content information transmitted from the providing device based on the preference information (for example, the receiving unit 56 in FIG. 5);
Presenting means for presenting the content information to the user (for example, the display unit 58 in FIG. 5) is further provided.

An information processing method or program according to the first aspect of the present invention includes:
Viewing information that is information related to viewing of content is generated (for example, step S111 in FIG. 12),
Based on the viewing information, an information gain (for example, a numerical information gain Gn) of numerical information, which is information expressed by numerical values, regarding the content is calculated (for example, step S74 in FIG. 11),
Based on the viewing information, an information gain (for example, non-numeric information gain G _A ) of non-numeric information, which is information represented by non-numeric values, is calculated (for example, step S72 in FIG. 11).
Based on the information gain of the numerical information and the information gain of the non-numeric information, preference information (for example, a decision tree) representing the user's preference is generated (for example, step S76 in FIG. 11).
Includes steps.

The providing device according to the second aspect of the present invention (for example, the server 13 in FIG. 13)
Receiving means (for example, the receiving unit 101 in FIG. 13) that receives the preference information from an information processing apparatus (for example, the client 11) that generates preference information (for example, a decision tree) that represents user preferences;
Non-numeric information relating to content, which is information represented by non-numeric values, numerical information relating to the content, numerical information represented by numerical values, and the preference information, a selection means (for example, FIG. 13 discriminating units 102),
Providing means (for example, the transmission unit 106 in FIG. 13) for providing content information (for example, a recommendation list), which is information related to the content selected by the selection means, to the information processing apparatus;

Storage means for storing ranking information that is information related to content ranking (for example, the ranking information storage unit 132 in FIG. 16).
Further comprising
Based on the ranking information, the selection unit selects content that the user likes among the content ranked in the ranking (for example, processing in step S164 in FIG. 17).

The providing method or program according to the second aspect of the present invention includes:
The preference information is received from an information processing device (for example, the client 11 in FIG. 5) that generates preference information (for example, a decision tree) representing the user's preference (for example, step S121 in FIG. 14).
Based on the non-numeric information that is non-numeric information related to the content, the numerical information represented by the numeric values related to the content, and the preference information, the user-preferred content is selected (for example, FIG. 14). Step S124),
Content information, which is information about the selected content, is provided to the information processing apparatus (for example, step S125 in FIG. 14).
Includes steps.

  Hereinafter, specific embodiments to which the present invention is applied will be described in detail with reference to the drawings.

  FIG. 3 is a block diagram showing a configuration example of an embodiment of the content recommendation system 1 to which the present invention is applied.

  The content recommendation system 1 in FIG. 3 includes a client 11, a network 12, and a server 13, and recommends content that the user likes to the user.

  In the following description, the content is described as music data, but the content may be image data instead of music data.

  The client 11 generates a decision tree that represents the user's preference for music, and transmits the decision tree to the server 13 via the network 12 such as the Internet or a local area network. In addition, the client 11 transmits a list of attribute information indicating attribute values of the attributes of the music, which is information related to the music of the music recommended to the user, transmitted from the server 13 according to the decision tree (hereinafter referred to as a recommendation list). ).

  When the attribute information of the desired music is selected as the viewing target attribute information from the recommendation list by the user, the client 11 requests the server 13 to transmit the music to be viewed through the network 12. To do. In response to the request, the client 11 receives the music transmitted from the server 13 via the network 12 and outputs the music corresponding to the music. Thereby, the user can view desired music.

  Based on the decision tree transmitted from the client 11 via the network 12, the server 13 selects the attribute information of the music predicted to be preferred by the user from the attribute information of the music stored by itself. To do. Then, the server 13 recommends the music to the user by providing the list of attribute information as a recommendation list to the client 11 via the network 12.

  Further, the server 13 selects the music to be viewed from among the music stored by itself in response to a request for transmission of the music to be viewed from the client 11, and via the network 12, Send to client 11.

  FIG. 4 is a block diagram illustrating a hardware configuration example of the client 11 of FIG.

  A CPU (Central Processing Unit) 31 in FIG. 4 executes various processes according to a program stored in a ROM (Read Only Memory) 32 or a recording unit 38. A RAM (Random Access Memory) 33 appropriately stores programs executed by the CPU 31 and data. The CPU 31, ROM 32, and RAM 33 are connected to each other by a bus 34.

  An input / output interface 35 is also connected to the CPU 31 via the bus 34. Connected to the input / output interface 35 are an input unit 36 including a keyboard, a mouse, a microphone, a receiving unit that receives a command transmitted from a remote controller (not shown), and an output unit 37 including a display and a speaker. . The CPU 31 executes various processes in response to commands input from the input unit 36. Then, the CPU 31 outputs the processing result to the output unit 37.

  The recording unit 38 connected to the input / output interface 35 includes, for example, a hard disk, and stores programs executed by the CPU 31 and various data. The communication unit 39 communicates with an external device such as the server 13 via the network 12. The communication unit 39 may acquire the program via the network 12 and record it in the recording unit 38.

  The drive 40 connected to the input / output interface 35 drives a removable medium 41 such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, and drives the program or data recorded therein. Get etc. The acquired program and data are transferred to and stored in the recording unit 38 as necessary.

  Although not shown, the server 13 in FIG. 3 is configured in the same manner as the client 11 in FIG. However, the communication unit 39 of the client 11 communicates with an external device such as the client 11 via the network 12.

  FIG. 5 is a block diagram illustrating a functional configuration example of the client 11.

  5 includes an attribute information storage unit 51, a viewing information storage unit 52, a preference extraction unit 53, a transmission unit 54, a command unit 55, a reception unit 56, a viewing processing unit 57, and a display unit 58.

  The attribute information storage unit 51 and the viewing information storage unit 52 correspond to, for example, the recording unit 38 of FIG. 4, and the preference extraction unit 53, the command unit 55, and the viewing processing unit 57 correspond to, for example, the CPU 31. The transmission unit 54 and the reception unit 56 correspond to the communication unit 39, for example, and the display unit 58 corresponds to the output unit 37, for example.

  The attribute information storage unit 51 stores attribute information included in the recommendation list received from the server 13 and supplied from the receiving unit 56. In addition to the attribute information included in the recommendation list, attribute information acquired from a CDDB (Compact Disc Date Base) or the like via the receiving unit 56 may be stored in the attribute information storage unit 51.

  The viewing information storage unit 52 stores viewing information that is supplied from the viewing processing unit 57 and is information related to the viewing of the music.

  The preference extraction unit 53 includes a numerical information gain calculation unit 61, a non-numeric information gain calculation unit 62, and a decision tree generation unit 63.

  The numerical information gain calculation unit 61 responds to the decision tree transmission request supplied from the command unit 55, and the attribute information stored in the attribute information storage unit 51 is represented by numerical values (hereinafter referred to as numerical values). , Referred to as numerical information), and viewing information is read from the viewing information storage unit 52. The numerical information gain calculation unit 61 calculates a numerical information gain Gn that is an information gain of the numerical information based on the numerical information and the viewing information, and supplies the numerical information gain Gn to the decision tree generation unit 63.

The non-numeric information gain calculation unit 62 is information whose attribute value is represented by a non-numeric value among the attribute information stored in the attribute information storage unit 51 in response to a decision tree transmission request supplied from the command unit 55. (Hereinafter referred to as non-numeric information) is read, and viewing information is read from the viewing information storage unit 52. Non-numeric information gain calculation unit 62, based on the non-numerical information and viewing information, non-numeric information gain G _A which is the information gain of a non-numerical information is calculated and supplied to the decision tree generator 63.

Decision tree generator 63, and numerical information gain Gn from numerical information gain calculation unit 61, based on the non-numeric information gain G _A, it generates a decision tree, and supplies the decision tree in the transmitting unit 54. The transmission unit 54 transmits the decision tree supplied from the decision tree generation unit 63 to the server 13 via the network 12.

  When nothing is stored in the attribute information storage unit 51 or the viewing information storage unit 52, the decision tree generation unit 63 does not generate a decision tree and displays information indicating that no decision tree has been generated. The data is transmitted to the server 13 via the transmission unit 54 and the network 12. In this case, the server 13 transmits a predetermined piece of music stored by itself to the receiving unit 56 of the client 11.

  The transmission unit 54 also requests the server 13 to transmit the music via the network 12 in response to the acquisition request for the music to be viewed supplied from the command unit 55.

  The command unit 55 receives a command input from the input unit 36 (FIG. 4) by the user, and supplies a request corresponding to the command to the preference extraction unit 53, the transmission unit 54, or the viewing processing unit 57.

  Specifically, the command unit 55 receives a recommendation list acquisition command from the user, and supplies the preference extraction unit 53 with a transmission request for a decision tree corresponding to the command. In addition, the instruction unit 55 receives a command for selecting a music to be viewed from the user, and supplies a request for acquiring a music to be viewed corresponding to the command to the transmission unit 54. Furthermore, the command unit 55 supplies a viewing request corresponding to the command to the viewing processing unit 57 in response to a command (eg, playback, fast forward, rewind, stop, etc.) related to viewing from the user. In response to a recommendation list display command, a recommendation list display request corresponding to the command is supplied to the viewing processing unit 57.

  The receiving unit 56 receives the recommendation list or the music transmitted from the server 13 via the network 12. The receiving unit 56 supplies the received recommendation list to the attribute information storage unit 51. The receiving unit 56 supplies the received music to the viewing processing unit 57.

  The viewing processing unit 57 holds the music from the receiving unit 56 in, for example, a built-in memory. In response to a request for viewing supplied from the command unit 55, the viewing processing unit 57 performs processing corresponding to the request (for example, playback, fast forward, rewind, stop) for the music stored in the built-in memory. Etc.). In addition, the viewing processing unit 57 creates viewing information according to the processing and supplies the viewing information to the viewing information storage unit 52.

  Furthermore, in response to a recommendation list display request from the command unit 55, the viewing processing unit 57 reads attribute information included in the recommendation list from the attribute information storage unit 51 and supplies the attribute information to the display unit 58. The display unit 58 displays (presents) the attribute information from the viewing processing unit 57.

  The user operates the input unit 36 in FIG. 4 to select the desired song attribute information from the attribute information included in the recommendation list displayed on the display unit 58, so that the music to be viewed Command to select. The instruction unit 55 receives the instruction and supplies the transmission unit 54 with an acquisition request for the music to be viewed.

  FIG. 6 shows an example of attribute information.

  As shown in FIG. 6, the attribute information is configured by associating an attribute value of each attribute of the song with an ID unique to the song (hereinafter referred to as a song ID) and a song name (song name). The

  In the example of FIG. 6, the song IDs “song # 1”, “song # 2”, “song # 3”, “song # 4”, etc., and the song names “song #A”, “song #B”, “Song #C”, “Song #D”... Have attributes “Genre”, “Artist”, “Release Year”, “Tempo”, “Volume Change”, and “Sound Pressure Change”. By associating values, attribute information is configured.

  Among the attribute information, the attribute values of the attributes “genre”, “artist”, and “release year” are non-numeric values that are not numerical values calculated for each music piece, that is, non-numeric information. In the example of FIG. 6, the attribute value of the attribute “genre” is “jazz”, “classic”, “jazz”, “classic”, etc., which are the genres of the music, and the attribute value of the attribute “artist” The names of the artists are “Taro Odede”, “Hanako Kyuka”, “Hanako Kyuka”, “One of the Orchestra”. Further, the attribute value of the attribute “release year” is “2005”, “2004”, “2001”, “2002”, and so on, which are the years when the music was released.

  On the other hand, the attribute values of the attributes “tempo”, “volume change”, and “sound pressure change” are numerical information represented by numerical values calculated for each music piece. In the example of FIG. 6, the attribute value of the attribute “tempo” is “100”, “90”, “75”, “80”,... The attribute values are “90”, “50”, “80”, “100”,..., Which are numerical values indicating the degree of change in volume. Also, the attribute value of the attribute “sound pressure change” is “40”, “30”, “45”, “50”,..., Which are numerical values indicating the degree of change in sound pressure (volume).

  The attribute is not limited to the example of FIG. 6, and for example, “chord progression”, “album”, and the like may be used. In this case, the attribute value of the attribute “chord progression” is a numerical value indicating the progression (change) of the chord of the music, and the attribute value of the attribute “album” is the name of the album in which the music is recorded, and is a non-numeric value. It is.

  FIG. 7 shows an example of viewing information.

  As shown in FIG. 7, the viewing information is configured by associating information related to the viewing of the music corresponding to the music ID with the same music ID as the music ID included in the attribute information of FIG. 6.

  In the example of FIG. 7, a process related to viewing (for example, playback, etc.) is performed on the music ID “music # 1”, “music # 2”, “music # 3”, “music # 4”. Viewing information is configured by associating the ratio of the length (hereinafter referred to as viewing ratio) of the portion where fast-forwarding, rewinding, and the like) are performed with the determination result of viewing success / failure.

  In the example of FIG. 7, when the ratio of the length of the portion where the reproduction is performed to the total length of the music in the viewing ratio is 70% or more, the viewing processing unit 57 If it is less than 70%, viewing information including a determination result “viewing determination” is created, and if it is less than 70%, it is determined that the song is “viewing failure”, and the determination result “viewing determination failure” is included. Viewing information shall be created.

  Therefore, the music corresponding to the determination result “viewing determination” is the user's favorite music (favorite music) that is almost all played by the user, and the music corresponding to the determination result “viewing determination failure” is the user Therefore, the music (which is disliked by the user) that is hardly played by the user is not played.

  In the example of FIG. 7, the music ID “music # 1” is associated with the viewing ratio “play 100%” and the determination result “viewing determination”. That is, all of the songs with the song ID “Song # 1” have been played back, and the viewing processing unit 57 determines that “passing the viewing” for the song.

  In addition, the music ID “music # 2” is associated with the viewing ratio “reproduction 30%”, “fast-forward 70%”, and the determination result “viewing determination failure”. That is, in the song with the song ID “Song # 2”, only 30% of the total length is played back, and 70% of the total length is fast-forwarded (flyed). Therefore, the viewing processing unit 57 determines that “viewing failure” for the song.

  The song ID “Song # 3” is associated with the viewing ratio “fast forward 15%”, “playback 80%”, “fast forward 5%”, and the determination result “viewing determination”. That is, in the song with the song ID “Song # 3”, 80% of the total length is reproduced, and 20% of the total length is fast-forwarded. Therefore, the viewing processing unit 57 determines that “viewing is passed” for the music.

  Also, the music ID “music # 4” is associated with the viewing ratio “playback 80%” and the determination result “viewing determination”. That is, 80% of the total length is played in the song with the song ID “Song # 4”. Therefore, the viewing processing unit 57 determines that “viewing is passed” for the music.

  Next, FIG. 8 shows an example of a decision tree generated by the decision tree generator 63 of FIG.

  In FIG. 8, the same components as those in FIG. 1 are denoted by the same reference numerals, and description thereof will be omitted because it will be repeated.

  In the example of FIG. 8, a discrimination node “sum of numerical information preference degree?” 81 is connected to the branch “winding one” 2-3, and the discrimination node “sum of numerical information preference degree?” 81 is connected. Is connected to the class node “like” 83-1 via the branch “greater than the threshold Zc”.

  Accordingly, among the music pieces having the attribute “artist” having the attribute value “1”, the music having a sum of numerical information preference degrees (details will be described later in FIG. 10) larger than the threshold value Zc is the music that the user likes. It is classified into the class “like” which is a class.

  Also, the class node “dislike” 83-2 is connected to the discrimination node “total preference of numerical information?” 81 via a branch “below threshold Zc” 82-2.

  Accordingly, among the music pieces having the attribute “artist” having the attribute value of “one orchestra”, the music whose sum of the preference degrees of the numerical information is equal to or less than the threshold value Zc is a class that is a class of music that the user does not like (dislikes). Classified as "I hate".

  Next, a decision tree transmission process in which the client 11 in FIG. 5 transmits a decision tree will be described with reference to the flowchart in FIG. This decision tree transmission process is started, for example, when the user instructs acquisition of a recommendation list.

  In step S <b> 31, the numerical information gain calculation unit 61 responds to the decision tree transmission request supplied from the command unit 55 in response to a user's recommendation list acquisition command, and the preference level for the numerical information of each song Perform a preference calculation process to calculate the sum of This preference degree calculation process will be described with reference to FIG.

  After the process of step S31, it progresses to step S32 and the preference extraction part 53 performs the decision tree production | generation process which produces | generates the decision tree of FIG. This decision tree generation process will be described with reference to FIG.

  After the processing in step S32, the process proceeds to step S33, in which the transmission unit 54 transmits the decision tree supplied from the decision tree generation unit 63 to the server 13 via the network 12, and ends the processing.

  As described above, the transmission unit 54 does not transmit personal information such as a user's viewing history to the server 13 but transmits a decision tree to the server 13, so that the privacy of the user can be protected. Moreover, since the transmission unit 54 transmits a decision tree that generally has a smaller amount of data than the viewing history, the bandwidth of the network 12 can be saved.

  Next, the preference calculation processing in step S31 in FIG. 9 will be described with reference to the flowchart in FIG.

In step S51, the numerical information gain calculation unit 61 reads the attribute value _Xij of the numerical information of all pieces of music from the attribute information storage unit 51, and normalizes the attribute value _Xij . Note that i is 1 representing the attribute “tempo”, 2 representing the attribute “volume change”, or 3 representing the attribute “sound pressure change” in the numerical information of the music. J is 1 representing the music ID “music # 1”, 2 representing “music # 2”, 3 representing “music # 3”, 4 representing “music # 4”, and so on.

Specifically, the numerical information gain calculation unit 61 normalizes the attribute value X _ij according to the following equation (1) to obtain the normalized attribute value Y _ij .

・・・（１）

... (1)

In Equation (1), Xm _i represents the average of the attribute values X _ij of the attribute i, and Xs _i represents the standard deviation of the attribute value X _ij of the attribute i.

After the processing of step S51, the process proceeds to step S52, and the numerical information gain calculation unit 61 selects a song as a “favorite song” based on the determination result of the viewing information (FIG. 7) stored in the viewing information storage unit 52. And “dislike music”, and using the normalized attribute values Y _ij classified into the “favorite music” and “dislike music”, for each attribute i, “favorite music” classified as being average Gm _i and the standard deviation Gs _i of music piece, and determining the average Bm _i and the standard deviation Bs _i of songs categorized into the "dislike music." As a result, the distribution for each attribute i of the normalized attribute value Y _ij corresponding to the music classified as “favorite music” and “dislike music” is predicted.

After the process of step S52, the process proceeds to step S53, and the numerical information gain calculation unit 61 uses the average Gm _i and the standard deviation Gs _i of each attribute i of “favorite music” according to the following equation (2). The range R _i of each attribute i is obtained.

・・・（２）

... (2)

In the formula (2), Cr is a constant representing the width of the range R _i . For example, a value of 1.5 to 2.0 is used as the constant Cr.

After the processing of step S53, the process proceeds to step S54, and the numerical information gain calculation unit 61 uses the range R _i and follows the following equation (3) based on Bayes' theorem, and the conditional probability P _i of each attribute _i. Find (Good | R _i ).

・・・（３）

... (3)

In Equation (3), Good represents the probability of being a “favorite music”, and Bad represents the probability of being a “dislike music”. In other words, the conditional probability P _i (Good | R _i ) represents the probability that a song whose attribute value Y _ij after normalization of the attribute i is in the range R _i is a “favorite song”.

The average Gm _i , standard deviation Gs _i , and conditional probability P _i (Good | R _i ) obtained as described above represent the power (discriminating power) for discriminating the class of music for each attribute i. Yes.

After the process of step S54, the process proceeds to step S55, and the numerical information gain calculation unit 61 uses the conditional probability P _i (Good | R _i ) of each attribute i for each piece of music according to the following equation (4). Then, the preference degree Z _ij of each attribute i is obtained.

・・・（４）

... (4)

  In Expression (4), φ (t) represents a cumulative distribution function of Gaussian distribution.

After the processing of step S55, the process proceeds to step S56, and the numerical information gain calculation unit 61 performs all the processing for each piece of music according to the following formula (5) from the preference Z _ij of each attribute i obtained for each piece of music. The sum Zall _j of the preference degree Z _ij of the attribute i is obtained.

・・・（５）

... (5)

After the processing in step S56, the process proceeds to step S57, the numerical information gain calculation unit 61, based on the sum Zall _j of preference Z _ij of each song, the respective music as "favorite songs", "dislike music" Classify. Specifically, when the numerical information gain calculation unit 61 determines that the total sum Zall _j is larger than the predetermined threshold value Zc, the music piece ID j is classified as “favorite music”, and the total sum Zall _j is the threshold value. When it is determined that it is equal to or less than Zc, the music with the music ID j is classified as “dislike music”.

  The threshold value Zc may be set in advance, or may be obtained by a predetermined calculation so that the correct answer rate of the classification based on the threshold value Zc is increased.

  As described above, the numerical information gain calculation unit 61 classifies each music piece into “favorite music” and “dislike music”, and then stores the classification result in a built-in memory (not shown). Return to step S31 of step 9.

In FIG. 10, the conditional probability P _i (Good | R _i ) is obtained in step S54. However, for example, a numerical value indicating a difference in distribution such as the amount of KL information may be obtained.

  Next, the decision tree generation process in step S32 in FIG. 9 will be described with reference to the flowchart in FIG.

  In step S71, the decision tree generation unit 63 reads the viewing information from the viewing information storage unit 52, and determines whether the determination results of all the songs included in the set C are the same among the determination results of the viewing information. To do. Note that in the first step S71, the set C is a set of songs having all the song IDs included in the attribute information and the viewing information.

If it is determined in step S71 that the determination results of all the songs included in the set C are not the same, the process proceeds to step S72, and the non-numeric information gain calculation unit 62 performs the non-numeric information gain of each attribute A of the non-numeric information. to calculate the G _a.

Specifically, the non-numeric information gain calculation unit 62 firstly sets a ratio of a subset C _k of music pieces whose viewing information determination result is k in the set C (a ratio of the subset C _{k to} the set C). ) Using P _k , the information amount I (C) of the set C is _{obtained according} to the following equation (6). Note that the determination result k is, for example, 0 indicating “viewing determination result” or 1 indicating “viewing determination failure”.

・・・（６）

... (6)

Next, non numerical information gain calculation unit 62, the ratio of the attribute value v in the subset C _vA per _A, subset C _k subsets C _k occupying the proportion (subset C _vA of attributes A non-numeric information ) Using P _vA , information amount I (C _vA ) for each subset C _{vA is obtained} according to the following equation (7). The attribute value v _A represents a number assigned to the attribute value in order from 1 as a number indicating the attribute value of each attribute A. For example, when the attribute A is the attribute “genre”, the attribute value v _A is 1 representing the attribute value “jazz” or 2 representing the attribute value “classic”.

・・・（７）

... (7)

Then, non numerical information gain calculation unit 62 uses the information amount I (C _vA), subject to the following equation (8), to calculate a non-numeric information gain G _A for each attribute A.

・・・（８）

... (8)

In Expression (8), | C _vA | is the size of the subset C _vA , and | C | is the size of the set C. In Expression (8), m is the maximum value of the attribute value v _A (for example, 2 when the attribute A is the attribute “genre”).

The non-numeric information gain G _A calculated as described above is based on the attribute value v _A of the attribute _A , and the classification and viewing information determination when the music is classified into the class “like” or “dislike” The degree of coincidence with the “favorite song” or “dislike song” represented by the result k, that is, the degree of accuracy of the classification.

After the step S72, the process proceeds to step S73, non-numeric information gain calculation unit 62, among the non-numeric information gain G _A calculated in step S72, the attribute Amax corresponding to the highest non-numeric information gain G _Amax The non-numeric information gain G _Amax and attribute Amax are supplied to the decision tree generator 63.

  After the process of step S73, the process proceeds to step S74, where the numerical information gain calculation unit 61 calculates the information amount I (C) calculated in step S72, the viewing information stored in the viewing information storage unit 52, and FIG. Based on the classification result in step S57, the numerical information gain Gn is obtained according to the following equation (9).

・・・（９）

... (9)

In Equation (9), P _kl represents the proportion of music pieces whose classification result is determined to be 1 in step S57 among the music pieces included in the subset C _k . The classification result l is 0 representing “favorite music” or 1 representing “dislike music”.

The numerical information gain Gn calculated as described above is obtained when the music is classified into the class “like” or “dislike” based on whether the preference degree Zall _j is greater than the threshold value Zc or less than the threshold value Zc. Represents the degree of coincidence between the classification and the “favorite music” or “dislike music” indicated by the judgment result k of the viewing information, that is, the degree of accuracy of the classification.

After the process of step S74, the process proceeds to step S75, and the decision tree generation unit 63 determines whether or not the maximum non-numeric information gain G _Amax is greater than or _equal to the numerical information gain Gn. If it is determined in step S75 that the maximum non-numeric information gain G _Amax is greater than or equal to the numerical information gain Gn, the process proceeds to step S76, and the decision tree generation unit 63 creates a discrimination node “attribute Amax?”.

After the processing of step S76, the process proceeds to step S77, where the decision tree generation unit 63 creates a branch for each attribute value v _{Amax of the} attribute Amax and connects to the discrimination node “attribute Amax?” Created in step S76. The process proceeds to step S78.

In step S78, the decision tree generation unit 63 sets the subset C _vAmax for each attribute value v _Amax of the attribute Amax to the set C. After the process of step S78, the process returns to step S71, and the processes of steps S72 to S83 are performed for each subset C _vAmax .

On the other hand, when it is determined in step S74 that the non-numeric information gain G _Amax is not equal to or smaller than the numerical information gain Gn (YES in step S74), the processing proceeds to step S79, where the decision tree generation unit 63 determines the preference level of the numeric information Is created (FIG. 8).

  After the processing of step S79, the process proceeds to step S80, and the decision tree generation unit 63 creates branches “greater than threshold value Zc” 82-1 and “below threshold value Zc” 82-2. To the total preference degree? "81.

  After the process of step S80, the process proceeds to step S81, where the decision tree generation unit 63 creates a class node “like”, connects it to the branch “greater than the threshold Zc”, and proceeds to step S82.

  In step S82, the decision tree generating unit 63 creates a class node “dislike”, connects to the branch “below threshold Zc”, returns to step S32 in FIG. 9, and proceeds to step S33 in FIG.

  When it is determined in step S71 that the determination results k of all the songs included in the set C are the same, the process proceeds to step S83, and the decision tree generation unit 63 selects a class node corresponding to the determination result k. create.

  Specifically, when the determination result k of all the songs included in the set C is “viewing determination result”, that is, when all the songs are “favorite songs”, the decision tree generation unit 63 determines whether the class Create the node “like”. On the other hand, when the viewing determination result k of all the songs included in the set C is “viewing determination not possible”, that is, when all the songs are “disliked music”, the decision tree generating unit 63 uses the class node “ I hate it.

  Then, the decision tree generation unit 63 connects the created class node to the branch corresponding to the set C, and ends the process. In the process of step S82 after the process of the first step S71, since there is no branch corresponding to the set C, the class node is not connected anywhere.

  By applying the song attribute information to the decision tree generated as described above, it is possible to determine whether the class of the song is the class “like” or the class “dislike”.

  As described above with reference to FIG. 12, the decision tree generation unit 63 creates both a determination node for determining non-numeric information and a determination node for determining numerical information, and generates a decision tree by combining both of the determination nodes. Thus, the decision tree can accurately represent user preferences.

  As a result, when the attribute information of the music is applied to this decision tree, the music is classified into classes “like” or “compared to the case where it is applied to the decision tree constituted only by the discrimination node for discriminating non-numeric information. It can be accurately classified as “dislike”. For example, according to the user's preference, songs with fast jazz tempo can be classified as class “like”, and songs with slow jazz tempo can be classified into class “dislike”. Further, the attribute information can be applied to the same decision tree regardless of whether the attribute information of the music is numerical information or non-numeric information.

Further, in FIG. 12, when the maximum non-numeric information gain G _Amax is equal to or greater than the numerical information gain Gn, that is, the music classification based on the attribute value v _Amax of the attribute Amax is compared with the classification based on the preference level Zall _j. If it is correct, the decision tree generator 63 creates a discrimination node “attribute Amax?”. On the other hand, when the maximum non-numeric information gain G _Amax is smaller than the numerical information gain Gn, that is, the classification of the music based on the preference Zall _j is more accurate than the classification based on the attribute value v _Amax of the attribute Amax. The decision tree generation unit 63 creates a discrimination node “total preference of numerical information?”.

  Therefore, by applying the decision tree generated by the decision tree generation process of FIG. 12 to the song attribute information, the song can be accurately classified into the class “like” or “dislike”.

  In FIG. 11, in step S71, the decision tree generation unit 63 determines whether the determination results k of all the songs included in the set C are the same, but not all the songs included in the set C. Further, it may be determined whether or not the determination results k of the music pieces of a certain ratio or more among all the music pieces included in the set C are the same.

  With reference to the flowchart of FIG. 12, the viewing process in which the client 11 of FIG. 5 causes the user to view the music will be described. This viewing process is started, for example, when the display of the recommendation list is instructed by the user.

  In step S <b> 101, the viewing processing unit 57 reads the attribute information from the attribute information storage unit 51 in response to the viewing list display request supplied from the command unit 55 and causes the display unit 58 to display the attribute information. Here, the user operates the input unit 36 of FIG. 4 to select the attribute information of the desired music from the attribute information displayed on the display unit 58, thereby instructing the selection of the music to be viewed. To do. The command unit 35 receives the command and supplies a request for acquiring a music to be viewed to the transmission unit 54 as a request corresponding to the command.

  After the processing of step S101, the process proceeds to step S102, and the transmission unit 55 determines whether or not the user has selected a music to be viewed, that is, whether or not a transmission request for the music to be viewed has been supplied from the command unit 55. If it is determined that the viewing target is not selected by the user, the process waits until the viewing target is selected by the user.

  If it is determined in step S102 that the user has selected a song to be viewed, the process proceeds to step S103, and the transmission unit 55 requests the server 13 to transmit the song to be viewed via the network 12. At this time, the viewing processing unit 57 also transmits the music ID of the music to be viewed.

  In step S104, the receiving unit 56 determines whether or not the music transmitted from the server 13 has been received in response to the request in step S103, and determines that the music has not been received yet. stand by.

  On the other hand, if it is determined in step S104 that the music has been received, the receiving unit 56 supplies the received music to the viewing processing unit 57, and the process proceeds to step S105. In step S105, the viewing processing unit 57 starts playback of the music supplied from the receiving unit 56, and proceeds to step S106.

  In step S <b> 106, the instruction unit 55 determines whether or not an instruction regarding viewing has been performed by the user, that is, whether or not an instruction regarding viewing input from the input unit 36 has been received, and when determining that an instruction regarding viewing has been performed, A request corresponding to the command is supplied to the viewing processing unit 57.

  After the processing in step S106, the process proceeds to step S107, and the viewing processing unit 57 determines whether or not the instruction regarding viewing by the user is a reproduction stop command, that is, the request supplied from the command unit 55 is a stop command. It is determined whether the request is a corresponding stop request, and if it is determined that the request is not a stop command, the process proceeds to step S108.

  In step S <b> 108, the viewing processing unit 57 performs a process according to a command from the user in response to a request from the command unit 55. For example, when the user commands fast-forwarding (fast-forward playback) of music and a request for fast-forwarding is supplied from the command unit 55 to the viewing processing unit 57, the viewing processing unit 57 performs fast-forward playback of the music.

  After the process of step S108, the process proceeds to step S109, where the viewing processing unit 57 determines whether or not the music has been played to the end, and if it is determined that the music has not been played to the end, the process returns to step S106 and the above-described processing is performed. repeat.

  If it is determined in step S107 that the instruction regarding viewing by the user is an instruction to stop reproduction, or if it is determined in step S109 that the music has been reproduced to the end, the process proceeds to step S110 and the viewing processing unit 57 is performed. Ends the playback.

  After the processing in step S110, the process proceeds to step S111, and the viewing processing unit 57 performs viewing information based on the status of the viewing-related processing (for example, playback, fast forward, stop, rewind, etc.) performed in steps S105 to S110. And the viewing information is supplied to the viewing information storage unit 52 to update the viewing information of the music to be viewed.

  FIG. 13 is a block diagram illustrating a functional configuration example of the server 13 of FIG.

  13 includes a receiving unit 101, a determining unit 102, an attribute information storage unit 103, a music storage unit 104, a music reading unit 105, and a transmission unit 106.

  The receiving unit 101 receives a decision tree transmitted via the network 12 from the transmitting unit 54 of the client 11 in FIG. 5 and supplies the decision tree to the determination unit 102. The receiving unit 101 receives a transmission request transmitted from the transmission unit 54 of the client 11 and receives a music ID of a music to be viewed. The receiving unit 101 supplies the received music ID to the music reading unit 105.

  The determination unit 102 reads the attribute information from the attribute information storage unit 103 and applies the attribute information to the decision tree from the reception unit 101 to determine the music class of the music ID included in the attribute information. Then, the determination unit 102 creates a list of attribute information of music pieces determined to have the class “like” as a recommendation list, and supplies the list to the transmission unit 106.

  The attribute information storage unit 103 stores a list of music stored in the music storage unit 104. The music storage unit 104 stores music. The music reading unit 105 reads the music corresponding to the music ID from the music storage unit 104 based on the music ID from the receiving unit 101 and supplies the music to the transmission unit 106.

  The transmission unit 106 transmits the recommendation list from the determination unit 102 to the reception unit 56 (FIG. 5) of the client 11 via the network 12. The transmission unit 106 transmits the music from the music reading unit 105 to the reception unit 56 via the network 12.

  Next, a list transmission process in which the server 13 in FIG. 13 transmits a recommendation list will be described with reference to the flowchart in FIG. This list transmission processing is started, for example, when a decision tree is transmitted from the transmission unit 54 of the client 11 in step S33 of FIG.

  In step S121, the reception unit 101 receives the decision tree, supplies the decision tree to the determination unit 102, and proceeds to step S122. In step S122, the determination unit 102 reads the attribute information from the attribute information storage unit 103, and proceeds to step S123. In step S123, the determination unit 102 determines the class of the song with the song ID included in the attribute information by applying the attribute information read in step S122 to the decision tree.

  For example, when the determining unit 102 determines the class of the song # 4 included in the attribute information by applying the attribute information shown in FIG. 6 to the decision tree shown in FIG. Since the attribute value of the attribute “genre” is “classic”, the determination process by the determination unit 102 is performed from the determination node “genre?” 1-1 through the branch “classic” 2-2 to the determination node. Proceed to "Artist?"

Then, since the attribute value of the attribute “artist” of the song # 4 is “winding one”, the determination processing is performed from the determination node “artist?” Through the branch “winding one” 2-3 and the determination node. Proceed to “Total preference of numerical information?”. Thereafter, the determination unit 102, a preference degree Zall ₄ of music of music # 4, determined by performing the same processes as steps S51 to S56 in FIG. 10, the preference degree Zall ₄ obtained as a result of the threshold Zc Determine if it is larger.

If it is determined that the preference level Zall ₄ is greater than the threshold value Zc, the determination process passes through the branch “greater than the threshold value Zc” 82-1 to the class node “like” 83-1, and the determination unit 102 The class of the song # 4 is determined as the class “like”. On the other hand, when it is determined that the preference level Zall ₄ is equal to or less than the threshold value Zc, the determination process proceeds to the class node “dislike” 83-2 through the branch “is equal to or less than the threshold value Zc” 82-2, and the determination unit 102 determines that the class of the song # 4 is class “dislike”.

  After the process of step S123, the process proceeds to step S124, and the determination unit 102 selects a song that has been determined as the class “like”, that is, a song that the user is expected to like, and recommends a list of attribute information of the song. Create as a list.

  The discriminating unit 102 uses the number of discriminating nodes through which discrimination processing has passed before discriminating a song as a class “like” as the suitability indicating the degree of fit between the song and the decision tree. A predetermined number of songs may be selected in descending order of suitability from the determined songs, and a list of attribute information of the selected songs may be created as a recommendation list. In addition, the determination unit 102 may add a fitness to each attribute information of the recommendation list. The determination unit 102 supplies the created recommendation list to the transmission unit 106.

  After the process of step S124, the process proceeds to step S125, and the transmission unit 106 transmits the recommendation list from the determination unit 102 to the client 11 via the network 12, and ends the process.

  The recommendation list transmitted in this way is received by the receiving unit 56 (FIG. 5) of the client 11. The receiving unit 56 supplies the recommendation list to the attribute information storage unit 51 and causes the attribute information storage unit 51 to store the attribute information included in the recommendation list. When the display of attribute information is instructed by the user, the attribute information read from the attribute information storage unit 51 is displayed on the display unit 58 as described in step S101 of FIG.

  As a result, the display unit 58 displays only the attribute information of the music classified into the class “like”, that is, the music that the user likes (and is predicted to be). Therefore, the user can easily find the attribute information of the desired music from the attribute information displayed on the display unit 58 and view the music.

  Next, the music transmission process in which the server 13 in FIG. 13 transmits music will be described with reference to the flowchart in FIG. This music transmission process is started, for example, when a transmission request for a music to be viewed and a music ID of the music are transmitted from the transmission unit 54 of the client 11 in step S103 of FIG.

  In step S <b> 141, the receiving unit 101 receives a transmission request for a music to be viewed, which is transmitted from the client 11 via the network 12, and receives a music ID of the music. The receiving unit 101 supplies the received music ID to the music reading unit 105.

  After the process of step S141, it progresses to step S142, and the music reading part 105 reads the music corresponding to the music ID from the music memory | storage part 104 based on music ID from the receiving part 101, and transmits the music. 106.

  After the process of step S142, it progresses to step S143, and the transmission part 106 transmits the music from the music reading part 105 to the client 11 via the network 12, and complete | finishes a process. The music transmitted in this way is received in step S104 in FIG.

  FIG. 16 is a block diagram showing another functional configuration example of the server 13 of FIG.

  In the server 130 of FIG. 16, a determination unit 131 is provided instead of the determination unit 102 of the server 13 of FIG. 13, and a ranking information storage unit 132 is newly provided.

  The determination unit 131 reads information (hereinafter referred to as ranking information) in which the ranking in the hit chart and the music ID are associated with each other from the ranking information storage unit 132. Further, the determination unit 131 reads, from the attribute information storage unit 103, attribute information of the music ranked in the hit chart based on the ranking information.

  Further, the discriminating unit 131 is ranked in the music of the music ID included in the attribute information, that is, the hit chart, by applying the attribute information read from the attribute information storage unit 103 to the decision tree from the receiving unit 101. Determine the class of the song. Then, the determination unit 102 creates a list of songs determined as the class “like” as a recommendation list and supplies the list to the transmission unit 106.

  Note that the determination unit 131 may add the ranking of the corresponding music to each attribute information of the recommendation list.

  The ranking information storage unit 131 stores ranking information. For example, the ranking information is acquired and stored (updated) from other devices at any time via the receiving unit 101.

  Next, a list transmission process in which the server 130 of FIG. 16 transmits a recommendation list will be described with reference to the flowchart of FIG. This list transmission process is started, for example, when a decision tree is transmitted from the transmission unit 54 (FIG. 5) of the client 11 in step S33 of FIG.

  In step S161, the reception unit 101 receives a decision tree, supplies the decision tree to the determination unit 131, and proceeds to step S162, as in step S121 of FIG. In step S162, the determination unit 131 reads the ranking information from the ranking information storage unit 132, and proceeds to step S163.

  In step S <b> 163, the determination unit 131 reads the attribute information of the music corresponding to the ranking information read from the ranking information storage unit 132 from the attribute information storage unit 103. Specifically, based on the music ID included in the ranking information, the determination unit 131 reads the attribute information of the music corresponding to the music ID from the attribute information storage unit 103.

  The processing in steps S164 to S166 is the same as the processing in steps S123 to S125 in FIG.

  The recommendation list transmitted as described above is received by the receiving unit 56 (FIG. 5) of the client 11, and when the display of attribute information is instructed by the user, the display is performed as described in step S101 of FIG. In part 58, attribute information of the recommendation list is displayed.

  As a result, the display unit 58 displays only the attribute information of the music that is classified into the class “like” among the popular music ranked in the hit chart, that is, the music that the user likes (and is predicted). Is done. Therefore, the user can easily find the attribute information of the desired music from the attribute information of the music ranked in the hit chart, compared to the case where the attribute information of all the music ranked in the hit chart is displayed. , You can watch the song.

  In the above description, the server 13 has transmitted the attribute information of the music determined as the class “like” to the client 11 as a recommendation list, but the server 13 is a part of the attribute information of the music ( For example, a list of music IDs may be transmitted as a recommendation list. In this case, the client 11 acquires attribute information from another device based on a part of the attribute information and stores it.

  In the above description, the attribute information acquired from another device is stored in the attribute information storage unit 51 in FIG. 5, but meta information (for example, MPEG (Moving Picture) added to the music is added. Information included in the ID3 tag of Experts Group) 3) may be stored as attribute information. Further, an analysis unit (not shown) provided in the client 11 may acquire the numerical information of the music by analyzing the music, and store the numerical information in the attribute information storage unit 51.

  Furthermore, in the above description, the numerical information gain calculation unit 61 calculates the numerical information gain Gn using the entire numerical information as one attribute. However, the numerical information attribute i is divided into several groups, The numerical information gain Gn may be calculated.

  Further, the preference extraction unit 53 may generate a decision tree using other criteria using the concept of information amount, for example, an information gain ratio, a gini index, or the like.

  Furthermore, in the present embodiment, a technique using a decision tree has been described as a technique for determining a music class. However, as a technique for determining a music class, for example, an information gain such as ILP (Inductive Logic Programming) is used. Other methods for performing clustering may be used.

  In this embodiment, the music is classified into two classes, “like” and “dislike”, but the type of class is not limited to this, and by using viewing information for each person. For example, it can be classified into classes of likes and dislikes for each person. In this case, for example, the classes are classes “I like Mr. A”, “I hate Mr. A”, “I like Mr. B”, “I hate Mr. B”, and so on. Furthermore, by adding the weather information of the day you watched to the viewing information, you can classify into classes such as “Songs that fit on a rainy day”, “Songs that fit on a sunny day”, and “Songs that fit on a cloudy day” You can also.

  In the present specification, the step of describing the program stored in the program recording medium is not limited to the processing performed in time series in the order described, but is not necessarily performed in time series. Or the process performed separately is also included.

  Further, in this specification, the system represents the entire apparatus constituted by a plurality of apparatuses.

  The embodiment of the present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the gist of the present invention.

It is a figure which shows an example of the conventional decision tree. It is a flowchart explaining the decision tree production | generation process which produces | generates the decision tree of FIG. It is a block diagram which shows the structural example of the content recommendation system to which this invention is applied. It is a block diagram which shows the hardware structural example of a client. It is a block diagram which shows the functional structural example of a client. It is a figure which shows the example of attribute information. It is a figure which shows the example of viewing information. It is a figure which shows the example of a decision tree. It is a flowchart explaining a decision tree transmission process. It is a flowchart explaining a preference degree calculation process. It is a flowchart explaining a decision tree generation process. It is a flowchart explaining a viewing-and-listening process. It is a block diagram which shows the functional structural example of a server. It is a flowchart explaining list transmission processing. It is a flowchart explaining a music transmission process. It is a block diagram which shows the other functional structural example of a server. It is a flowchart explaining another list transmission process.

Explanation of symbols

  11 client, 13 server, 31 CPU, 32 ROM, 33 RAM, 38 recording unit, 41 removable media, 54 transmitting unit, 56 receiving unit, 57 viewing processing unit, 58 display unit, 61 numerical information gain calculating unit, 62 non-numeric value Information gain calculator, 63 decision tree generator, 101 receiver, 102 discriminator, 106 transmitter

Claims (8)

Viewing information generating means for generating viewing information that is information related to content viewing;
Calculation means for calculating an information gain of numerical information, which is information represented by numerical values, regarding the content based on the viewing information;
Calculation means for calculating an information gain of non-numeric information, which is information represented by a non-numeric value, on the content based on the viewing information;
An information processing apparatus comprising: preference information generating means for generating preference information representing a user preference based on the information gain of the numerical information and the information gain of the non-numeric information. Transmitting means for transmitting the preference information to a providing device that provides content information that is information relating to the user-preferred content;
Receiving means for receiving the content information transmitted from the providing device based on the preference information;
The information processing apparatus according to claim 1, further comprising: presentation means for presenting the content information to a user. Generate viewing information that is related to content viewing,
Based on the viewing information, the information gain of the numerical information, which is information expressed by numerical values, regarding the content is calculated,
Based on the viewing information, calculate an information gain of non-numeric information, which is information represented by non-numeric values, regarding the content;
An information processing method including a step of generating preference information representing a user preference based on an information gain of the numerical information and an information gain of the non-numeric information. Generate viewing information that is related to content viewing,
Based on the viewing information, the information gain of the numerical information, which is information expressed by numerical values, regarding the content is calculated,
Based on the viewing information, calculate an information gain of non-numeric information, which is information represented by non-numeric values, regarding the content;
A program that causes a computer to execute a process including a step of generating preference information representing a user's preference based on an information gain of the numerical information and an information gain of the non-numeric information. Receiving means for receiving the preference information from an information processing device that generates preference information representing user preferences;
Non-numeric information relating to content, which is information represented by non-numeric values, numerical information relating to the content, numeric information represented by numerical values, and selection means for selecting content that the user likes,
A providing apparatus comprising: providing means for providing content information, which is information related to the content selected by the selecting means, to the information processing apparatus. Storage means for storing ranking information, which is information related to content ranking,
The providing device according to claim 5, wherein the selection unit selects content that the user likes among the content ranked in the ranking based on the ranking information. From the information processing device that generates the preference information representing the user's preference, the preference information is received,
Based on non-numeric information, which is information represented by non-numeric values, numerical information represented by numeric values, and the preference information, the content preferred by the user is selected.
A providing method including a step of providing content information, which is information about the selected content, to the information processing apparatus. From the information processing device that generates the preference information representing the user's preference, the preference information is received,
Based on non-numeric information, which is information represented by non-numeric values, numerical information represented by numeric values, and the preference information, the content preferred by the user is selected.
A program for causing a computer to execute a process including a step of providing content information, which is information relating to the selected content, to the information processing apparatus.

Images