JP5158450B2 - Information processing apparatus and method, and program - Google Patents

Description

  The present invention relates to an information processing device and method, and a program, and more particularly, to an information processing device and method, and a program capable of presenting content.

  Conventionally, metadata indicating user evaluation is added to content such as music and moving images.

  When the user gives metadata indicating his / her preference to the content, the user can know whether the content to which the metadata is given is his / her preference.

  Some conventional recording / reproducing apparatuses provide information desired by a user in response to a user operation based on variable data generated by quantifying an impression received from information (for example, patents). Reference 1).

JP 2000-207415 A

  However, it has not been possible to know whether or not the content to which metadata indicating the user's preference is attached is his / her preference.

  In order to know whether or not all of my content is my preference, I had to give metadata indicating my preference to all of my content.

  When a user owns a large number of contents, it is very troublesome to add metadata indicating his / her preference to all contents.

  The present invention has been made in view of such a situation, and makes it possible to display the presentation of content tailored to the user more easily.

  An information processing apparatus according to an aspect of the present invention is an information processing apparatus that requests information to be presented to a user, and is a search axis arbitrarily designated by the user, and a reference for presenting the content. A search axis generating means for generating the search axis, an input means for inputting the degree to which each of the plurality of contents is applied to the search axis as a feedback value of the user, and a regression equation for obtaining a value indicated by the search axis Alternatively, learning data generation means for generating learning data for obtaining a discriminant based on the feedback value, and regression for obtaining a value indicated by a search axis using the feature amount of the content and the learning data Search axis learning means for learning a formula or discriminant and the value of the search axis for the content to be presented to the user An information processing apparatus including a calculation means for calculating, based on the regression equation or discriminants.

  The calculation means may calculate a value indicated by each of the plurality of search axes.

  It is possible to further include display control means for displaying an image indicating the content at a position corresponding to the value of the content indicated by the search axis in a space having the search axis as an axis of the dimension.

  A sorting means for sorting out the content in a predetermined range indicated by the search axis among the contents in which the value indicated by the search axis is displayed can be further provided.

  One aspect of the present invention is information relating to content, and in an information processing method for an information processing apparatus that requests information to be presented to a user, a search axis generation unit is a search axis arbitrarily designated by the user, A search axis serving as a reference for presenting content is generated, and the input unit inputs a degree to which each of the plurality of contents applies to the search axis as a feedback value of the user, and the learning data generation unit includes the search axis. Learning data for obtaining a regression equation or a discriminant for obtaining the value indicated by is generated based on the feedback value, and the search axis learning means performs a search using the feature amount of the content and the learning data. Learn the regression or discriminant for obtaining the value indicated by the axis, and the calculation means The value of the search axis is an information processing method comprising the step of calculating, based on the regression equation or discriminants.

  One aspect of the present invention is a search axis arbitrarily designated by the user as a reference for presenting the content in an information processing apparatus that requests information to be presented to the user from a computer. A search axis generating means for generating a search axis, an input means for inputting the degree to which each of the plurality of contents is applied to the search axis as a feedback value of the user, and a regression equation for obtaining a value indicated by the search axis or A learning data generation means for generating learning data for determining a discriminant based on the feedback value, and a regression equation for determining a value indicated by a search axis using the feature amount of the content and the learning data Alternatively, the search axis learning means for learning the discriminant and the search axis for the content to be presented to the user Which is a program to function as an information processing apparatus and a calculating means for calculating, based on the regression equation or discriminants.

  In one aspect of the present invention, a search axis arbitrarily designated by the user, a search axis serving as a reference for presenting the content is generated, and the degree to which each of the plurality of contents applies to the search axis Learning data for obtaining a regression equation or discriminant for obtaining a value indicated by the search axis, which is input as a user feedback value, is generated based on the feedback value, and the feature amount of the content and the learning data Is used to learn a regression equation or discriminant for obtaining the value indicated by the search axis, and the value of the search axis is calculated based on the regression equation or discriminant for the content to be presented to the user.

  As described above, according to one aspect of the present invention, content can be presented.

It is a block diagram which shows the example of a structure of a music search system. It is a block diagram which shows the structure of a music search device. It is a figure explaining the production | generation of the search axis | shaft of a search axis | shaft learning part. It is a figure which shows the example of a music feature-value and a feedback value. It is a figure which shows the example of a music feature-value and a feedback value. It is a figure explaining calculation of the value of a search axis. It is a figure explaining calculation of the value of a search axis. It is a block diagram which shows the structure which learns a search axis. It is a figure which shows the example of the log | history of the reproduction | regeneration of the music by a user. It is a figure which shows the example of a feedback value. It is a figure which shows the example of the data for learning. It is a figure explaining the production | generation of the data for learning from the log | history of the reproduction | regeneration of the music by a user. It is a figure explaining learning of the discriminant of a search axis. It is a flowchart explaining the learning process of a search axis. It is a figure explaining the user interface which mapped the music in the space using a search axis. It is a figure explaining the user interface which mapped the music on the space using the search axis produced | generated by the information with respect to another user. It is a block diagram which shows the structure which searches using a search axis. It is a figure explaining acquisition of a search axis. It is a figure explaining calculation of the value of a search axis. It is a figure explaining the selection of music. It is a flowchart explaining the process of the search using a search axis. It is a figure explaining calculation of the similarity of a search axis. It is a figure explaining calculation of the similarity of a search axis. It is a figure which shows the example of the similarity of a search axis. It is a figure which shows the example of the similarity of a search axis. It is a figure which shows the example of the similar search axis displayed superimposed. It is a figure which shows the example of operation of selection of a search axis. It is a block diagram which shows the structure which calculates the similarity of a search axis. It is a flowchart explaining the process of presentation of a similar search axis. It is a figure which shows the example of the position of the user's music displayed in an overlapping manner, and the position of the music of another user. It is a block diagram which shows the structure which overlaps and displays the position of a user's music, and the position of another user's music in the space more than two dimensions. It is a figure explaining acquisition of a search axis. It is a figure explaining calculation of the value of a search axis. It is a flowchart explaining the process which displays and displays the content of another user. It is a flowchart explaining the content selection process. It is a block diagram which shows the structural example of the hardware of a computer.

  FIG. 1 is a block diagram illustrating an example of a configuration of a music search system. The music search system includes a music search device 11-1, a music search device 11-2, a server 12, and a network 13.

  The music search device 11-1, the music search device 11-2, and the server 12 are connected to each other via the network 13.

  Each of the music search device 11-1 and the music search device 11-2 is an example of an information processing device, and obtains information related to music, which is an example of content, and presents the information to each user.

  Note that content means information useful to the user, such as music (sound), moving images, still images, sentences (text), data, or programs. Hereinafter, music will be described as an example of content.

  For example, the music search device 11-1 and the music search device 11-2 are axes serving as a reference for presenting music as an example of content (hereinafter referred to as a search axis). The search axis defined by the history of the music playback by or the evaluation of each user's music is generated by learning, and the display related to the presentation of the music based on the generated search axis is controlled.

  Here, the search axis is a standard for presenting content according to information related to the content. For example, the search axis is a coordinate axis of a space that defines a position (coordinate position) according to information related to the content.

  Here, the search axis can be a coordinate axis in an arbitrary coordinate system. Specifically, for example, a search axis is a straight line in a space such as an orthogonal coordinate system or an oblique coordinate system, and a search axis in a polar coordinate system such as a circular coordinate system, a cylindrical coordinate system, or a spherical coordinate system, or other coordinate system. Is a reference indicating a straight line or a curve or angle. Note that the search axis is not necessarily displayed.

  For example, the music search device 11-1 is generated based on a search axis generated based on information for the user of the music search device 11-1, which is a reference for presenting the music, and information on the user of the music search device 11-2. The degree of similarity with the search axis is calculated, and the music value indicated by the two similar search axes is displayed. For example, the music search device 11-2 is generated based on a search axis generated based on information for the user of the music search device 11-2, which is a reference for presenting the music, and information on the user of the music search device 11-1. The degree of similarity with the search axis is calculated, and the music value indicated by the two similar search axes is displayed.

  The music search device 11-1 acquires a search axis generated from information for the user of the music search device 11-2 from the music search device 11-2 via the network 13. The music search device 11-2 acquires a search axis generated from information for the user of the music search device 11-1 from the music search device 11-1 via the network 13.

  Although details will be described later, the value of the search axis for the content is obtained by a regression equation or discriminant, and acquiring the search axis means acquiring a regression equation or discriminant representing the search axis.

The server 12 acquires and records the search axis generated by the information for each user from the music search device 11-1 or the music search device 11-2 via the network 13. The server 12 provides the search axis generated by the information for each recorded user to the music search device 11-1 or the music search device 11-2 via the network 13.

  Hereinafter, when there is no need to distinguish between the music search device 11-1 and the music search device 11-2, they are simply referred to as the music search device 11.

  FIG. 2 is a block diagram illustrating a configuration of the music search apparatus 11 that is an example of the information processing apparatus according to the embodiment of this invention. The music search apparatus 11 includes a search axis learning unit 31, a feature amount database (DB (Database)) 32, a search axis database (DB) 33, a learning data generation unit 34, a reproduction history feedback database (DB) 35, and an operation unit 36. , Search axis acquisition unit 37, other user search axis database (DB) 38, search axis value calculation unit 39, other user feature quantity database (DB) 40, corresponding music selection unit 41, content playback unit 42, content database (DB) 43, a search axis comparison unit 44, a display unit 45, and a communication unit 46.

  The search axis learning unit 31 is a search axis that serves as a reference for presenting music, and generates a search axis determined by learning based on the feature amount of the music and the history of music playback by the user or the user's evaluation of the music. For example, the search axis learning unit 31 provides the feature amount of the music supplied from the feature amount database 32 and the history of music playback by the user supplied from the playback history feedback database 35 via the learning data generation unit 34. Alternatively, a search axis serving as a reference for presenting the music is generated by learning using the user's evaluation of the music.

  Here, the feature amount may be a value indicating the feature of the content. For example, the feature amount is a value calculated from the metadata of the content, or the metadata or the content itself.

  That is, for example, the search axis learning unit 31 serves as a reference for presenting a song by learning using the feature amount of the song from the feature amount database 32 and the history or evaluation of the reproduction of the song from the learning data generation unit 34. A regression equation or discriminant for calculating a value indicated by the search axis (hereinafter referred to as a search axis value) is obtained.

  The search axis learning unit 31 supplies the generated search axis to the search axis database 33. That is, the search axis learning unit 31 supplies the search axis database 33 with a regression equation or a discriminant for calculating the value of the search axis serving as a reference for presenting music.

  The feature quantity database 32 stores the feature quantities of music pieces owned by the user. For example, the feature database 32 stores music metadata, which is a music feature.

  The search axis database 33 stores search axes generated by the search axis learning unit 31 and serving as a reference for presenting music. In other words, the search axis database 33 stores a regression equation or a discriminant for calculating a search axis value serving as a reference for presenting music.

  The learning data generation unit 34 removes data unnecessary for learning of the search axis to be generated from the history of the music playback by the user or the evaluation of the user music supplied from the playback history feedback database 35, and performs the search. Learning data used for learning in the axis learning unit 31 is generated. The learning data generation unit 34 supplies the generated learning data to the search axis learning unit 31.

  The reproduction history feedback database 35 stores the reproduction history of music by the user and the evaluation of the user's music that is a feedback value from the user. The reproduction history feedback database 35 supplies the learning data generation unit 34 with the stored history of reproduction of music by the user or evaluation of the user's music.

  The operation unit 36 includes a button, a keyboard, a mouse, a remote controller, a touch pad, and the like and a control unit that controls them, and information indicating a search axis used for search according to a user operation is obtained as a search axis acquisition unit 37. To supply. Alternatively, the operation unit 36 supplies the reproduction history feedback database 35 with an evaluation of the user's music, which is a feedback value from the user, according to the user's operation.

  In addition, when the display unit 45 displays information related to the presentation of music based on the search axis, the operation unit 36 specifies information that specifies a predetermined range based on the search axis according to the user's operation. Is supplied to the corresponding music selection unit 41.

  The search axis acquisition unit 37 acquires a search axis used for search from the search axis database 33. In addition, the search axis acquisition unit 37 acquires a search axis used for search, which is a search axis generated based on information for other users from the other user search axis database 38.

  The search axis acquisition unit 37 supplies the acquired search axis to the search axis value calculation unit 39.

  The other user search axis database 38 is a search axis serving as a reference for presenting music, and stores a search axis generated by information for other users. In other words, the other-user search axis database 38 stores a regression equation or discriminant for calculating a value indicated by a search axis generated by information for other users, which is a reference for presenting music.

  For example, when the communication unit 46 receives a search axis transmitted from another music search device 11 or the server 12 via the network 13, the received search axis is supplied to the other user search axis database 38. The user search axis database 38 stores search axes generated from information for other users supplied from the communication unit 46. Further, for example, when the music search device 11 is used by a plurality of users, the search axis database 33 and the other user search axis database 38 store search axes generated by information for different users, respectively. Stores a search axis generated by information for a user who is currently using the music search device 11, and the other user search axis database 38 is based on information for a user different from the user who is currently using the music search device 11. Stores the generated search axis.

  The search axis value calculation unit 39 calculates the search axis value of the music from the music feature quantity of the music supplied from the feature quantity database 32 for the search axis supplied from the search axis acquisition unit 37. Further, the search axis value calculation unit 39 determines the search axis of the music from the music feature amount of the music of another user supplied from the other user feature database 40 for the search axis supplied from the search axis acquisition unit 37. Calculate the value shown.

  The search axis value calculation unit 39 supplies the calculated value obtained by the music search axis to the corresponding music selection unit 41 or the display unit 45.

  The other user feature quantity database 40 stores the feature quantities of music pieces owned by other users. For example, the other user feature database 40 stores the metadata of the music as the feature of the music owned by the other user.

  For example, the other user feature quantity database 40 stores the feature quantity of the music owned by the user of the other music search device 11. Alternatively, for example, when the music search device 11 is used by a plurality of users, the other user feature value database 40 stores the feature values of the music owned by a user different from the user who is currently using the music search device 11. .

  More specifically, for example, when the communication unit 46 receives the feature amount of the music of another user transmitted from the other music search device 11 or the server 12 via the network 13, the music of the other user is received. Are supplied to the other-user feature-value database 40, the other-user feature-value database 40 stores the feature-value of the music of other users supplied from the communication unit 46.

  The corresponding music selection unit 41 selects music in a predetermined range from the music that is the presented content. For example, the corresponding music selection unit 41 selects music in a predetermined range based on the two search axes from among the music displayed for presentation based on the two search axes.

  Alternatively, for example, the corresponding music selection unit 41 may display the music related to the user among the music displayed on the basis of the search axis generated by the information for the user and the search axis generated by the information for the other user. A predetermined range of music is selected based on a search axis generated by information and a search axis generated by information for other users.

  More specifically, the corresponding music selection unit 41 is a music whose position of the search axis belongs to a predetermined range based on the search axis, which is indicated by information supplied from the operation unit 36 according to the user's operation. Sort out.

  The content playback unit 42 reads out the music that is the content selected by the corresponding music selection unit 41 from the content database 43, plays it, and outputs it.

  The content database 43 stores music that is content owned by the user. Note that the music that is the content stored in the content database 43 and the metadata that is the feature amount stored in the feature amount database 32 are associated with each other by, for example, ID information that identifies the song.

  The search axis comparison unit 44 acquires a search axis generated by information for the user from the search axis database 33, and acquires a search axis generated by information for another user from the other user search axis database 38. The search axis generated by the information for the user is compared with the search axis generated by the information for the other user. For example, the search axis comparison unit 44 calculates a similarity between a search axis generated based on information for a user and a search axis generated based on information for another user.

  More specifically, the search axis comparison unit 44 is similar to the value indicated by the search axis generated based on the information for the user regarding the music and the value indicated by the search axis generated based on the information for the other user regarding the music. The degree of similarity between the search axis generated by the information for the user and the search axis generated by the information for the other user is calculated.

  In other words, the search axis comparison unit 44 reads the metadata of the music owned by the user from the feature amount database 32, and the search generated by the information about the user's music from the metadata of the music owned by the user. The value indicated by the axis and the value indicated by the search axis generated based on the information about the user's music for other users are calculated. The search axis comparison unit 44 indicates the value indicated by the search axis generated by the information about the user's owned music and the search axis generated by the information about the other user's music. The similarity between the search axis generated by the information for the user and the search axis generated by the information for the other user, which is the similarity to the value, is calculated.

  For example, the search axis comparison unit 44 includes the metadata that is the feature amount of music and the search axis determined by the user's history of music playback or the user's evaluation of the music, the music metadata, and music playback by other users. The degree of similarity with the search axis determined by the history or the evaluation of the music by other users is calculated.

  The display unit 45 includes an LCD (liquid crystal display) or an organic EL (electroluminescence) display and a display control unit that controls display of the LCD or organic EL display. Display images and text. For example, the display unit 45 uses a search axis generated by information for similar users and a search generated by information for other users based on the value indicated by the search axis of the music supplied from the search axis value calculation unit 39. The display related to the presentation of the music based on the axis is controlled, and the display related to the presentation of the music is displayed.

  Further, for example, when displaying the presentation of music, the display unit 45 is generated by an image showing a search axis generated by information for the user and information for other users similar to the search axis generated by information for the user. The displayed image indicating the search axis is displayed in an overlapping manner.

  That is, the display unit 45 controls display related to content presentation based on the search axis generated by the search axis learning unit 31. Or the display part 45 controls the display regarding presentation of the content on the basis of a similar search axis.

  The communication unit 46 receives the search axis (regression equation or discriminant equation) generated by the information for the other user transmitted from the other music search device 11 or the server 12 via the network 13, and receives the other user. The search axis generated based on the information for is supplied to the other user search axis database 38. In addition, the communication unit 46 receives the feature amount of the content of the other user transmitted from the other music search device 11 or the server 12 via the network 13 and transmits the feature amount of the content of the other user to the other user. This is supplied to the feature amount database 40.

  Further, the communication unit 46 reads the user search axis from the search axis database 33 and transmits the user search axis to the other music search device 11 or the server 12 via the network 13. Furthermore, the communication unit 46 reads the feature amount of the user's content from the feature amount database 32, and transmits the feature amount of the user's content to the other music search device 11 via the network 13.

  As described above, the music search device 11 generates a search axis that is used in the search and is personalized. And the music search device 11 shares a search axis between users.

  Next, generation of a search axis will be described.

  The music search device 11 generates a search axis used in the search from a feedback value that is a value indicating feedback input by the user or a history of playback of the user's music.

  For example, depending on the search axis, it is determined that the music is a music that the user likes or dislikes, or the degree of likes or dislikes of the music is required. Further, for example, the search axis determines whether the music is a music having a high or low frequency of reproduction, or the frequency of music reproduction.

  Furthermore, for example, according to the search axis, it is determined that the music is music that the user has recently heard or has not heard. For example, according to the search axis, it is determined that the music is music that the user listens on the weekend or music that is heard on weekdays. Further, for example, it is assumed that the music is music that the user listens to during the day or music that is heard at night by the search axis.

  As shown in FIG. 3, the search axis learning unit 31 uses a regression expression or discriminant for estimating a value indicated by the search axis from a music feature amount that represents the characteristics of the music, a history of music playback or a feedback value. A search axis is generated by learning from the music feature amount representing the characteristics of the music using statistical analysis or machine learning.

  By applying the generated regression equation or discriminant of the search axis to the music feature amount of the music, the value indicated by the search axis is calculated for the music. Therefore, the value indicated by the search axis can be calculated for music that has not been played back or music for which no feedback value has been input, and music can be presented according to the value indicated by the search axis.

  That is, even if the user plays only some of the songs he owns or inputs a feedback value for only some of the songs, From the history or feedback, the value indicated by the search axis can be obtained for all of the songs owned by the user.

  In this way, the value indicated by the search axis can be obtained from the metadata of the music by the regression equation or discriminant for the music that has never been played back or the music for which no feedback value has been input. Music can be searched using the search axis.

  FIG. 4 is a diagram illustrating examples of music feature amounts and feedback values when generating a search axis indicating the degree of likes / dislikes of music from feedback values indicating the degree of likes / dislikes of music in five stages. The music feature amount is, for example, metadata.

  For example, the user operates the operation unit 36 to input feedback values indicating likes and dislikes of music, that is, the degree of preference of music in five levels. The feedback value is an evaluation of the user's music. Note that the feedback value is stored in the reproduction history feedback database 35 as described above.

  In the example of FIG. 4, the feedback value of the preference (likes and dislikes) for the music 1 is 5 indicating that he likes very much, and the feedback value of the preference for the music 2 is 3 indicating that he does not like or dislike the music. The preference feedback value for 3 is 2 which is rather disliked.

  Further, in the example of FIG. 4, the values of the feature amount 1, the feature amount 2, and the feature amount 3 among the music feature amounts for the song 1 are 24, 74, and 68, respectively. , The feature value 2 and the feature value 3 are 17, 81, and 50, and the feature value 1, the feature value 2, and the feature value 3 for the music piece 3 are 33, 75, and 97.

  The search axis learning unit 31 learns a regression equation based on, for example, each preference feedback value and music feature value shown in FIG.

  For learning the regression equation, a method such as linear regression (multiple regression analysis) or SVR (support vector regression) can be used.

  By this learning, for example, F (music feature amount) = feature amount 1 × 0.5 + feature amount 2 × 0.2−feature amount 3 × 0.9 +. A regression equation for calculating a value indicated by a certain search axis is obtained. That is, by applying this regression formula to the music feature amount of the music, the value of the music indicated by the search axis that is the degree of likes and dislikes is obtained.

  FIG. 5 shows an example of the music feature amount and feedback value when generating a search axis indicating that the music is a user's favorite or disliked music from a binary feedback value indicating likes and dislikes for the music. FIG.

  For example, the user inputs a feedback value indicating the likes or dislikes of the music by operating the operation unit 36.

  In the example of FIG. 5, the preference feedback value is a binary value indicating whether the user likes or dislikes music.

  In the example of FIG. 5, the preference feedback value for the song 1 is liked, the preference feedback value for the song 2 is liked, and the preference feedback value for the song 3 is disliked.

  Further, in the example of FIG. 5, the values of the feature amount 1, the feature amount 2, and the feature amount 3 among the music feature amounts for the song 1 are 24, 74, and 68, respectively. , The feature value 2 and the feature value 3 are 17, 81, and 50, and the feature value 1, the feature value 2, and the feature value 3 for the music piece 3 are 33, 75, and 97.

  The search axis learning unit 31 learns a discriminant based on, for example, each preference feedback value and music feature value shown in FIG.

  For discriminant learning, methods such as linear discrimination, decision tree, support vector machine (SVM), or Naive Bayes can be used.

  For example, F (music feature value) calculated by F (music feature value) = feature value 1 × 0.5 + feature value 2 × 0.2−feature value 3 × 0.9 +. ), But if F (music feature) ≥ 0.0, I like the song, and if F (music feature) <0.0, I like or dislike the song that the user likes A discriminant for the search axis is determined to be music. That is, by applying this discriminant to the music feature amount of the music, the value of the music indicated by the search axis that the music is the music that the user likes or dislikes is obtained.

  For example, if the regression equation or discriminant is learned using the song metadata as the song feature quantity of the song, the search axis for each song is determined from the song metadata by the regression equation or discriminant obtained by learning. The indicated value can be calculated.

  In the music search device 11, the user can freely input a search axis by inputting a name of a new search axis and a feedback value such as a degree of whether or not the search axis is applied to some music. Can be generated.

  The regression equation or discriminant equation may be a linear equation or a non-linear equation.

  Next, music search using the search axis will be described.

  In order to search for music using the created search axes, first, the values indicated by the search axes are calculated for all the music owned by the user.

  For example, as shown in FIG. 6, the search feature indicating the degree of likes and dislikes in the feature amount of the song 1, that is, the feature amount 1 that is 24, the feature amount 2 that is 74, and the feature amount 3 that is 68. By applying the regression formula, 0.3 which is the value of the music 1 on the search axis is calculated. Further, by applying a regression equation of the search axis indicating the degree of listening well to the feature amount 1, feature amount 2, and feature amount 3 for the song 1, the value of the song 1 on the search axis is 0.2. Is applied, and a regression equation of the search axis indicating the degree of whether or not to listen in the morning is applied to calculate 0.1 which is the value of the music 1 on the search axis.

  Further, as shown in FIG. 6, a search axis indicating the degree of likes and dislikes for the music feature quantity for music piece 2, that is, feature quantity 1 that is 17, feature quantity 2 that is 81, and feature quantity 3 that is 50. By applying the regression equation, 0.8 which is the value of the music piece 2 on the search axis is calculated. In addition, by applying a regression equation of the search axis indicating the degree of listening well to the feature amount 1, feature amount 2, and feature amount 3 for the song 2, the value of the song 2 on the search axis is 0.9. Is applied, and the regression formula of the search axis indicating the degree of whether to listen in the morning is applied, so that 0.7, which is the value of the music piece 2 on the search axis, is calculated.

  Similarly, as shown in FIG. 6, a search axis indicating the degree of likes and dislikes for the music feature quantity for music piece 3, that is, feature quantity 1 that is 33, feature quantity 2 that is 75, and feature quantity 3 that is 97. Is applied, the value 0.4 of the music 3 on the search axis is calculated. Further, by applying a search axis regression equation indicating the degree of listening to the feature amount 1, feature amount 2, and feature amount 3 for the song 3, the value of the song 3 on the search axis is 0.3. Is applied, and the search axis regression formula indicating the degree of listening in the morning is applied to calculate 0.1, which is the value of the music piece 3 on the search axis.

  The desired music can be searched from the values indicated by the respective search axes for the respective music based on the search conditions described by the values indicated by the search axes. For example, the value of the search axis indicating the degree of likes and dislikes is 0.5 or more, the value of the search axis indicating the degree of listening well is 0.4 or less, and the value of the search axis indicating the degree of listening in the morning is 0. With the search condition being two or more, music that meets the search condition can be searched.

  Further, for example, as shown in FIG. 7, the music feature amount for the song 1, that is, the feature amount 1 that is 24, the feature amount 2 that is 74, and the feature amount 3 that is 68, are set to the user's favorite songs. By applying a search axis discriminant that is a song or a disliked song, the value of the song 1 on the search axis is “like”, that is, the song 1 indicates that the user likes it. "Like" is calculated. Further, by applying a discriminant of a search axis that is a song that is often listened to or not much listened to to the feature amount 1, feature amount 2, and feature amount 3 for the song 1, the song 1 on the search axis is applied. "Yes", that is, "Yes" indicating that the music 1 is often listened to by the user is calculated. Further, by applying a discriminant of a search axis that is a song that is heard in the morning or a song that is not heard in the morning to the feature value 1, feature value 2, and feature value 3 for the song 1, the song 1 on the search axis is applied. Is “No”, that is, “No” indicating that the music piece 1 is not heard by the user in the morning.

  When the user wants to search for his favorite music, the music search device 11 selects a music having a value of “like” on the search axis indicating that the music is a user's favorite music or a disliked music. Search and present the searched music to the user.

  Further, as shown in FIG. 7, the music feature amount for the music piece 2, that is, the feature amount 1 that is 17, the feature amount 2 that is 81, and the feature amount 3 that is 50, is the music that the user likes. By applying the search axis discriminant that the song is a certain or disliked song, the value of “song 2” on the search axis is calculated and “like” is calculated. By applying the search axis discriminant, “Yes”, which is the value of the music 2 of the search axis, is calculated. Further, by applying a discriminant of a search axis that is a song that is heard in the morning or a song that is not heard in the morning to the feature value 1, the feature value 2, and the feature value 3 for the song 2, the song 2 on the search axis is applied. Value “Yes”, that is, “Yes” indicating that the music 1 is heard by the user in the morning is calculated.

  Similarly, as shown in FIG. 7, the music feature amount for the music piece 3, that is, the feature amount 1 which is 33, the feature amount 2 which is 75, and the feature amount 3 which is 97, By applying the search axis discriminant that the music is a dislike or dislike, the value of the music 1 on the search axis is “dislike”, that is, the dislike of the music 1 is disliked by the user "Is calculated. Further, by applying a discriminant of a search axis that is a song that is often listened to or not much listened to to the feature amount 1, feature amount 2, and feature amount 3 for the song 3, the song 3 on the search axis is applied. Is “No”, that is, “No” indicating that the user does not listen to the music 3 very much. Further, by applying a search axis discriminant that is a song to be heard in the morning or a song to be heard in the morning to the feature value 1, feature value 2, and feature value 3 for the song 3, the song 3 on the search axis is applied. The value “No” is calculated.

  Also in this case, it is possible to search for a desired music piece based on the search condition described by the value indicated by the search axis from the value indicated by the search axis for each music piece. In other words, it is possible to search for music that applies to or does not apply to each search axis.

  For example, a search condition that the value indicated by the search axis indicating that the music is a user's favorite music or a disliked music is “like”, and the search condition is satisfied by the user ( (Predicted) music can be searched. In addition, for example, a search condition that the value indicated by the search axis indicating that it is a song that is often listened to or not listened to much is “No”, and a search condition that is not often heard by the user is searched. can do.

  FIG. 8 is a block diagram showing a configuration for learning a search axis among the configurations shown in FIG.

  When learning the search axis, as shown in FIG. 8, first, the learning data generation unit 34 searches the reproduction history feedback database 35 for a reproduction history of the music by the user or a feedback value that is an evaluation of the user's music. To get.

  The learning data generation unit 34 generates learning data by removing data unnecessary for learning of a search axis to be generated from a history of reproduction of music by a user or a feedback value that is an evaluation of the user's music. .

  The generation of learning data will be described with reference to FIGS.

  FIG. 9 is a diagram illustrating an example of a history of music playback by the user. The history of music playback by the user (data) indicates a history of the user playing back music or performing operations related to playback.

  For example, as shown in FIG. 9, the user's reproduction history of music is that music 1 is reproduced at 14:04 on September 7, 2006, and music 2 is reproduced at 14:00 on September 7, 2006. Played at 8 minutes, song 3 was played at 14:13 on September 7, 2006, song 3 was skipped at 14:15 on September 7, 2006, and song 4 was played , Played at 14:15 on September 7, 2006, indicating that playback of song 4 was stopped at 14:18 on September 7, 2006.

  FIG. 10 is a diagram illustrating an example of a feedback value that is an evaluation of a user's music. The feedback value stores explicit feedback given to the music by the user. That is, the feedback value indicates an explicit user evaluation of the music.

  The types of funny, parental or old feedback values shown in the example of FIG. 10 are set by the user himself / herself. It is a thing.

  In the example of FIG. 10, the feedback value for likes and dislikes for the song 1 is 5 indicating that he likes very much, the feedback value for likes and dislikes for the song 2 is 3 indicating that he does not like and dislikes, and likes and dislikes for the song 4. The feedback value of 1 is 1 indicating that he / she is very disliked, the feedback value of likes / dislikes for song 5 is 5 indicating that he / she likes very much, and the feedback value of likes / dislikes for song 6 is rather like It is 4 which shows that there is.

  In the example of FIG. 10, the feedback value for likes and dislikes for the music 3 is not input.

  In the example of FIG. 10, the interesting feedback value for song 1 is 3 indicating that it is so interesting, and the interesting feedback value for song 2 is 1 indicating that it is not very interesting. The feedback value is 5 indicating very interesting, and the feedback value indicating whether the music 6 is interesting is 4 indicating interesting.

  In the example of FIG. 10, interesting feedback values for music 4 and music 5 are not input.

  In the example of FIG. 10, the feedback value for the parent for the music 1 is 1 indicating that it is not for parents at all, and the feedback value for the parent for the music 2 is 2 indicating that it is not very for parents. Yes, the feedback value for the parent of the music 4 is 5 indicating that it is for the parent, and the feedback value for the parent of the music 5 is 1 indicating that it is not for the parent at all.

  In the example of FIG. 10, the feedback value for the parents for music 3 and music 6 is not input.

  In the example of FIG. 10, the old feedback value for music 2 is 3 indicating that it is fairly old, the old feedback value for music 3 is 1 indicating that it is not very old, and the old value for music 4 is old. The feedback value is 1 and the feedback value for the song 6 is 2 indicating a little old.

  In the example of FIG. 10, old feedback values for music 1 and music 5 are not input.

  For example, the learning data generation unit 34 generates learning data from the history of music playback by the user in FIG. 9 or the feedback value in FIG.

  FIG. 11 is a diagram illustrating an example of learning data generated by the learning data generation unit 34.

  The learning data is data for learning the regression equation or discriminant of the search axis. For example, the learning data is data of a list including a music ID for identifying music and correct answer data.

  In the example shown in FIG. 11, the song ID for identifying the song 1 is associated with the correct data that is 5, the song ID for identifying the song 3 is associated with the correct data that is 4, A music ID for identifying the music 6 and the correct answer data 2 are associated with each other, a music ID for identifying the music 12 and the correct data 3 are associated with each other, and the music for identifying the music 16 This is learning data for a list in which an ID and correct answer data of 1 are associated with each other.

  For example, in the example of the learning data shown in FIG. 11, each correct answer data indicates whether the user likes or dislikes the song specified by the associated song ID, whether it is funny, parental, or old. Indicate.

  More specifically, for example, the correct answer data which is 5 indicating the likes and dislikes of the user indicates that the user likes the music specified by the associated music ID very much and is 4 indicating the likes and dislikes of the user. Certain correct answer data indicates that the user likes the music specified by the associated music ID. In addition, the correct answer data that is 3 indicating the user's likes and dislikes indicates that the user does not like or dislike the music specified by the associated music ID, and the correct answer data that is 2 that indicates the likes or dislikes of the user is The correct answer data, which is 1 indicating the user's likes and dislikes, indicates that the user does not like the song identified by the associated song ID. The song identified by the associated song ID Indicates that the user is very disliked.

  For example, the learning data generation unit 34 extracts one type of feedback value and the corresponding music ID from the feedback value shown in FIG. Then, the learning data generation unit 34 deletes the music ID whose feedback value is empty from the extracted result. Further, the learning data generation unit 34 generates learning data by using the feedback value corresponding to the music ID as correct data.

  FIG. 12 is a diagram for explaining generation of learning data from a music reproduction history by a user.

  The learning data generation unit 34 is a list composed of a song ID for identifying a song and correct data indicating whether or not the song identified by the song ID has recently been played from the history of playing the song by the user. Generate learning data.

  For example, in the history as of 17:00 on September 7, 2006, the music 1 is reproduced at 14:04 on September 7, 2006, and the music 2 is played on September 7, 2006. Played at 14: 8, song 3 was played at 14:13 on September 7, 2006, song 3 was skipped at 14:15 on September 7, 2006, and song 3 4 is played at 14:15 on September 7, 2006, and the playback of the song 4 by the user indicating that the playback of song 4 is stopped at 14:18 on September 7, 2006 The learning data generation unit 34 associates the song ID for identifying the song 1 with the correct data (True) indicating that the song 1 has been recently played, and identifies the song 2 from the history of Is associated with the correct answer data (True) indicating that song 2 was recently played. Thus, the music ID for identifying the music 3 is associated with the correct answer data (True) indicating that the music 3 has recently been reproduced, and the music ID for identifying the music 4 and the music 4 have recently been reproduced. Is associated with correct data (True) indicating that the music has been performed, and is associated with music ID for identifying the music 5 and correct data (False) indicating that the music 5 has not been reproduced recently. The learning data, which is a list in which the song ID for identifying the song 6 and the correct answer data (False) indicating that the song 6 has not been reproduced recently, are associated with each other.

  The learning data shown in FIG. 12 is generated, for example, when generating a search axis indicating that the music is recently played back or is not played back recently.

  Returning to FIG. 8, the learning data generation unit 34 generates such learning data and supplies the generated learning data to the search axis learning unit 31.

  The search axis learning unit 31 acquires the music feature amount of each piece of music from the feature amount database 32 and learns the regression equation or discriminant of the search axis from the learning data and the music feature amount.

  For example, as shown in FIG. 13, the search axis learning unit 31 corresponds to at least a music ID for identifying the music 1 and correct data (True) indicating that the music 1 is favored by the user. The music ID for identifying the music 2 is associated with the music ID for identifying the music 2 and the correct data (True) indicating that the music 2 is favored by the user. Is associated with correct data (True) indicating that the music is liked by the user, and a music ID for identifying the music 4 and correct data (True) indicating that the music 4 is liked by the user. Is associated with the song ID for identifying the song 5 and the correct answer data (False) indicating that the song 5 is not preferred by the user, and the song ID for identifying the song 6; Correct answer date indicating that song 6 is not liked by the user (False) is a learning data that is a list associated with the learning data, and the feature amount 1 and 94 are the feature amount 1 and 94 for the song 1, and the feature amount 1 and 29 are 36 for the song 2. Feature amount 2, feature amount 1 which is 75 for song 3, feature amount 2 which is 85, feature amount 1 and 71 which is 96 for song 4, feature amount 2 which is 24 for song 5, feature amount 1 which is 24 for song 5 Are expressed in Equation (1), Equation (2), and Equation (3) from the feature amount 2 including feature amount 2 that is 53 and 53, and feature amount 1 that is 46 and feature amount 2 that is 0. The discriminant of the search axis is learned that the music is the music that the user likes or dislikes.

F (musical feature amount) = feature amount 1 × 0.5 + feature amount 2 × 0.2−feature amount 3 × 0.9 +... +0.3 (1)
F (music feature) ≧ 0.0 → Yes ・ ・ ・ (2)
F (music feature) <0.0 → No (3)

  It should be noted that the equation (2) determines that the user likes the song if F (music feature value) obtained by the equation (1) satisfies the condition of F (music feature value) ≧ 0.0. It shows that. Further, the expression (3) determines that the user dislikes the music if F (musical feature quantity) obtained by the expression (1) satisfies the condition of F (musical feature quantity) <0.0. It shows that.

  Thus, for example, the search axis learning unit 31 learns a regression equation for calculating the value indicated by the search axis from the music feature quantity from the music feature quantity and the history of the music reproduction by the user or the evaluation of the user's music. To obtain a search axis. Alternatively, for example, the search axis learning unit 31 obtains, by learning, a discriminant for calculating the value indicated by the search axis from the music feature value based on the music feature value and the history of the user's music reproduction or the evaluation of the user's music. Thus, a search axis is generated.

  The search axis learning unit 31 supplies the created search axis to the search axis database 33. The search axis database 33 stores the search axes supplied from the search axis learning unit 31. That is, the search axis learning unit 31 supplies the created regression equation or discriminant to the search axis database 33, and the search axis database 33 stores the regression equation or discriminant.

  FIG. 14 is a flowchart for explaining search axis learning processing. In step S11, the search axis learning unit 31 determines a search axis to be learned. For example, the search axis learning unit 31 determines a search axis indicated by a signal supplied from the operation unit 36 according to a user operation on the operation unit 36 as a search axis to be learned.

  In step S <b> 12, the learning data generation unit 34 generates learning data from a feedback history that is an evaluation of a user's music reproduction history or a user's music. For example, the learning data generation unit 34 acquires a reproduction history and feedback value of the music played by the user from the reproduction history feedback database 35. Then, the learning data generation unit 34 creates learning data composed of a music ID for identifying a music piece and correct data corresponding to the music piece from a history of reproducing the music piece by the user or a feedback value.

  In step S <b> 13, the search axis learning unit 31 acquires the music feature quantity of the music used for learning from the feature quantity database 32.

  In step S14, the search axis learning unit 31 learns the regression equation or discriminant of the search axis determined in step S11 from the learning data and the music feature amount. For example, the search axis learning unit 31 learns a discriminant by a method such as linear discrimination, decision tree, SVM, or Naive Bayes. Further, for example, the search axis learning unit 31 learns a regression equation by a method such as linear regression (multiple regression analysis) or SVR.

  In step S <b> 15, the search axis learning unit 31 stores the learned regression formula or discriminant of the search axis in the search axis database 33.

  In step S16, the search axis learning unit 31 determines whether or not all the search axes have been learned. If it is determined that all the search axes have not been learned, the search axis learning unit 31 returns to step S11 and repeats the above-described processing. .

  If it is determined in step S16 that all search axes have been learned, the search axis learning process ends.

  In this way, the music search device 11 can learn an axis serving as a reference for presenting content.

  In other words, the music search device 11 can learn a search axis that is a reference for presenting content and is generated based on information for the user based on information for the user. Further, the music search device 11 generates a search axis that serves as a reference for presenting the content and is determined by learning based on the feature amount of the content and the history of content playback by the user or the user's evaluation of the content. Can do.

  Next, music search using the search axis will be described.

  As shown in FIG. 15, a search interface based on a regression equation is used as each axis of a space of two or more dimensions, and music is mapped to this space, thereby providing a user interface capable of sensory search. Can do. In this case, mapping a song to a space means calculating a value indicated by each search axis for each song and specifying a position in the space indicated by the calculated value for the song. When displaying the space, an image indicating a music piece, for example, a mark such as a circle or cross, an icon, or the like is displayed at a specified position in the displayed space.

  By doing in this way, the user can know at a glance presentation of music by two or more search axes.

  For example, as shown in FIG. 15, a space of two or more dimensions expressed by a search axis based on a regression equation, and by specifying a range in the displayed space, the value of the search axis belongs to this range. Music is searched. More specifically, in the search axis that shows the degree of likes and dislikes, by specifying the range that you like and that you do not hear much on the search axis that shows the degree of listening well, There is a search for songs you like.

  Further, the music search device 11 can search for music using not only the user search axis generated by the user's playback history and feedback of the music search device 11 but also the search axes of other users.

  For example, the music search device 11 can search for a user's favorite music and a user A's favorite music. Further, for example, the music search device 11 can search for music that Mr. A who is another user listens in the morning. Furthermore, for example, the music search device 11 can search for music that the user likes and that is often heard on the weekend of B, another user.

  Thereby, the user can know a common point and difference with other desired users, such as a favorite common point and a difference, for example.

  In this case, for example, as shown in FIG. 16, a search axis based on a regression equation, which is generated based on information for other users, is used as a space axis, and music is mapped in this space.

  For example, a search axis indicating the degree of likes and dislikes of a user and a search axis indicating the degree of likes and dislikes of another user A are used as axes of respective dimensions of a two-dimensional space, and music is mapped in this space. Is done. In this case, for example, if a favorite range is specified in both the search axis indicating the degree of likes and dislikes of the user and the search axis indicating the degree of likes and dislikes of the other user A, Search for songs you like.

  FIG. 17 is a block diagram showing a configuration for performing a search using a search axis among the configurations shown in FIG.

  When a search is performed using a search axis, when the user operates the operation unit 36 to indicate a search axis used for the search, the operation unit 36 performs a search indicating the search axis used for the search in accordance with the user's operation. The axis information is supplied to the search axis acquisition unit 37. As shown in FIG. 17, the search axis acquisition unit 37 acquires the search axis indicated by the search axis information from the search axis database 33 or the other user search axis database 38.

  For example, as shown in FIG. 18, when search axis information indicating “a song that is often listened to” and “a song that Mr. A likes” is supplied from the operation unit 36, the search axis acquisition unit 37 reads “ The search axis of the name “Song”, the search axis indicating the degree of listening well (the song axis of FIG. 18) is acquired from the search axis database 33, and the name “Song A's favorite song” is also obtained. A search axis (a favorite music axis of Mr. A in FIG. 18) indicating the degree of likes and dislikes of A who is another user is acquired from the other user search axis database 38.

  That is, in this case, the search axis acquisition unit 37 acquires the search axis regression expression indicating the degree of whether or not to listen well, which is the search axis of the name “songs that are often listened to”, from the search axis database 33. A search axis regression formula indicating the degree of likes and dislikes of other user A, who is the search axis with the name “A's favorite song”, is acquired from the other user search axis database 38.

  The search axis acquisition unit 37 supplies the acquired search axis to the search axis value calculation unit 39.

  The search axis value calculation unit 39 acquires the music feature value of the user's music from the feature data database 32 and acquires the music feature value of the music of another user from the other user feature data database 40.

  Then, the search axis value calculation unit 39 calculates the value indicated by the search axis for the search axis supplied from the search axis acquisition unit 37 from the music feature amount.

  For example, as shown in FIG. 19, the search axis value calculation unit 39 is a search axis that indicates the degree to which the music feature quantity of the music piece 1 including the feature quantity 1 that is 15 and the feature quantity 2 that is 94 is heard well. And a search axis regression expression that indicates the degree of likes and dislikes of other user A, respectively, and a value that is 5 of the search axis that indicates the degree of listening well for song 1 Then, a value that is 3 of the search axis indicating the degree of likes and dislikes of other user A is calculated.

  In addition, the search axis value calculation unit 39 uses a search axis regression formula indicating the degree of listening to the music feature quantity of the music piece 2 including the feature quantity 1 which is 36 and the feature quantity 2 which is 29, and other users. By applying the search axis regression expression indicating the degree of likes and dislikes of Mr. A, the value that is 1 on the search axis indicating the degree of whether or not the song 2 is listened well and the other user A Calculate a value that is 2 on the search axis indicating the degree of likes and dislikes.

  Similarly, the search axis value calculation unit 39 calculates the regression equation of the search axis indicating the degree of listening well to the music feature quantities of the music pieces 3 to 6 and the degree of likes and dislikes of A who is another user. By applying each of the search axis regression equations shown, the values of the search axis indicating the degree of listening well for each of the music pieces 3 to 6 and the degree of likes and dislikes of the other user A are shown. Calculate the value of the search axis.

  The search axis value calculation unit 39 supplies the corresponding song selection unit 41 with the value for each search axis calculated for each song.

  The corresponding music selection unit 41 selects the music that matches the search condition from the value of each search axis for each music.

  For example, as shown in FIG. 20, the corresponding music selection unit 41 has a search axis value of 3 or more indicating the degree of listening well as the search range, and the degree of likes and dislikes of A who is another user When the search condition indicating that the value of the search axis indicating 3 or more is supplied from the operation unit 36, the values of the search axes indicating the degree of whether each of the music pieces 1 to 6 is listened well are respectively 5, 1, 3, 2, 4, or 4, and the values of the search axes indicating the degree of likes and dislikes of the other user A of each of the music 1 to 6 are 3 When it is 5, 4, 2, or 3, the value of the search axis indicating the degree of listening well is 3 or more, and the value of the search axis indicating the degree of likes / dislikes of other user A is 3 or more A certain song 1, song 3, song 5, and song 6 are selected (selected).

  The corresponding music selection unit 41 outputs a corresponding music list indicating the selected music to the content reproduction unit 42. For example, the corresponding music selection unit 41 outputs a corresponding music list that is a playlist indicating the music to be reproduced.

  The content playback unit 42 reads the music shown in the corresponding music list supplied from the corresponding music selection unit 41 from the content database 43, plays it, and outputs it.

  FIG. 21 is a flowchart for explaining search processing using a search axis. In step S <b> 31, the search axis acquisition unit 37 acquires a search axis used for the search from the search axis database 33 or the other user search axis database 38. That is, for example, the search axis acquisition unit 37 acquires the regression equation or discriminant of the search axis indicated by the search axis information from the operation unit 36 from the search axis database 33 or the other user search axis database 38.

  For example, in step S <b> 31, the search axis acquisition unit 37 acquires one of the two search axes used for the search from the search axis database 33 and acquires the other one from the other user search axis database 38. .

  In step S <b> 32, the search axis value calculation unit 39 acquires the music feature amount of the user's music from the feature amount database 32, acquires the music feature amount of the other user's music from the other user feature amount database 40, and The search axis value of all the music is calculated from the acquired music feature. For example, the search axis value calculation unit 39 calculates a value indicated by the search axis used for the search by applying a regression equation or a discriminant of the search axis used for the search to the acquired music feature amount.

  For example, in step S31, the search axis acquisition unit 37 acquires one of two search axes used for the search from the search axis database 33, and acquires the other one from the other user search axis database 38. In step S32, the search axis value calculation unit 39 obtains only the music feature quantity of the user's music from the feature quantity database 32, and calculates the values indicated by the search axes of all the music pieces from the obtained music feature quantity. Also good. By doing in this way, the user can know the value indicated by the user's search axis and the value indicated by the search axis of another user for the music owned by the user.

  On the other hand, for example, in step S31, the search axis acquisition unit 37 acquires one of the two search axes used for the search from the search axis database 33 and the other one as the other user search axis database 38. In step S32, the search axis value calculation unit 39 acquires only the music feature quantity of the music of another user from the other user feature quantity database 40, and searches the search axis of all the music from the acquired music feature quantity. The value indicated by may be calculated. By doing in this way, the user can know the value indicated by his / her own search axis and the value indicated by the other user's search axis for music owned by other users.

  In step S <b> 33, the corresponding music selection unit 41 determines whether the music satisfies the search condition from the value of the search axis used for the search and the search condition.

  If it is determined in step S33 that the music matches the search condition, the procedure proceeds to step S34. In step S34, the corresponding music selection unit 41 reproduces the music determined to satisfy the search condition on the content reproduction unit 42. To output.

  Then, the procedure proceeds to step S35. For example, in step S34, the content reproduction unit 42 reads out, reproduces, and outputs the music determined to satisfy the search condition from the content database 43.

  When the music for which it is determined whether or not the search condition is met is a music of another user, the reproduction of the music is skipped. In this case, for example, data for trial listening of the music of another user may be acquired from the server 12 via the network 13 and the music may be reproduced.

  If it is determined in step S33 that the music does not meet the search condition, step S34 is skipped and the procedure proceeds to step S35.

  In step S35, the corresponding music selection unit 41 determines whether or not all the music pieces have been discriminated. If it is determined that all the music pieces have not been discriminated, the process returns to step S33, and the next music piece has been described above. Repeat the process.

  If it is determined in step S35 that all the music pieces have been determined, the search process using the search axis ends.

  It is also possible to calculate the similarity between a search axis generated based on information for a user and a search axis generated based on information for other users, and display the content presentation based on a plurality of similar search axes. .

  The similarity of the search axis is obtained by calculating the value indicated by each search axis and calculating the similarity of the value indicated by the search axis for all songs owned by the user. The search axis comparison unit 44 calculates the value indicated by each search axis for all songs owned by the user, calculates the similarity of the value indicated by the search axis, and sets the result as the similarity of the search axis. .

  First, the case where the value indicated by the search axis is estimated from the regression equation will be described.

  When calculating the similarity of the search axes, as shown in FIG. 22, when the music pieces owned by the user are music 1 to music 8 in total, first, for each of music 1 to music 8, The search axis value indicating the degree of likes and dislikes, which is the first search axis generated by information, is obtained, and the second search axis generated by information for other users indicates the frequency of reproduction. The search axis value is determined.

  That is, as shown in the example of FIG. 22, the values of the search axes indicating the degree of likes and dislikes for music 1 to music 8 are 0.5, 0.4, 0.1, 0.2, 0.4, 0.3, 0.2, and 0.5, respectively. . These values are calculated by applying a regression equation for estimating the value of the search axis indicating the degree of likes and dislikes to the music feature quantities of music 1 to music 8.

  The search axis values indicating the frequency of reproduction for music 1 to music 8 are 0.2, 0.5, 0.4, 0.1, 0.2, 0.5, 0.2, and 0.4, respectively. These values are calculated by applying a regression equation for estimating the value of the search axis indicating the frequency of reproduction to the music feature quantities of the music pieces 1 to 8.

  Then, as the similarity of the search axis values, the correlation coefficient between the value indicated by the search axis indicating the degree of likes and dislikes and the search axis value indicating the frequency of reproduction for each of the music pieces 1 to 8 is calculated. The

  For example, a preference vector composed of eight values for music 1 to music 8 on the search axis indicating the degree of likes and dislikes, and a reproduction frequency vector composed of eight values for music 1 to music 8 on the search axis indicating the frequency of reproduction. The correlation coefficient is calculated as the similarity between the search axis indicating the degree of likes and dislikes and the search axis indicating the frequency of reproduction.

  When the correlation coefficient between the preference vector and the reproduction frequency vector is 0.1, the similarity between the search axis indicating the degree of likes and dislikes and the search axis indicating the reproduction frequency is set to 0.1.

  Next, a case where the value indicated by the search axis is estimated from the discriminant will be described.

  When calculating the similarity of the search axes, as shown in FIG. 23, when the music pieces owned by the user are music 1 to music 8 in total, first, for each of music 1 to music 8, The second search generated by the information for other users, which is the first search axis generated by the information, the value indicated by the search axis indicating that it is a favorite song or a disliked song is obtained. The value indicated by the search axis, which is an axis, a song that is often listened to, or a song that is not often listened to is obtained.

  That is, in the example shown in FIG. 23, the values of the search axes for music 1 to music 8 that are favorite music or dislike music are Yes, Yes, No, No, Yes, No, respectively. , No, Yes. These values are calculated by applying a discriminant for estimating the value of the search axis that is a favorite song or a dislike song to the song feature amounts of the songs 1 to 8.

  Further, the values of the search axes for the music 1 to the music 8 that are frequently listened music or music that is not often listened to are No, Yes, Yes, No, No, Yes, No, Yes. These values are calculated by applying a discriminant for estimating the value of the search axis that is a song that is often heard or a song that is not often heard to the song feature amounts of the songs 1 to 8.

  Then, as the similarity of the search axis value, for each of the songs 1 to 8, the search axis value that is a favorite song or a dislike song, and a song that is often listened to or not listened to much A match rate with the value of the search axis is calculated. For example, the coincidence rate is the result of dividing the number of songs whose search axis values match by the number of songs whose search axis values match or are compared.

  More specifically, the value of the search axis for the song 1 that is a favorite song or a dislike song is Yes, and a search axis that is a song that is often listened to or not much listened to. The value of No is No, and the values of the respective search axes for music 1 do not match. In addition, the value of the search axis for the music 2 that is a favorite music or a dislike music is Yes, and the value of the search axis that is a music that is often listened to or is not much listened to is Yes, and the values of the respective search axes for music 2 are the same.

  In the example shown in FIG. 23, the search axis values for the four pieces of the music piece 2, the music piece 4, the music piece 7, and the music piece 8 out of the eight pieces of the music pieces 1 to 8 are the same. From 4 songs / 8 songs, the match rate is 0.50.

  For example, it is assumed that the “favorite” vector consisting of eight values for the music 1 to music 8 on the search axis that is a favorite music or a dislike music and a music that is often listened to or not listened to much. The match rate with the “well listened” vector consisting of 8 values for the music 1 to 8 on the search axis is calculated, and the match rate is good with the search axis indicating that it is a favorite song or a dislike song. The similarity to the search axis indicating that the song is a song to be listened to or a song to be listened to rarely.

  If the match rate between the “Like” vector and the “Frequently Listen” vector is 0.50, it is assumed that the song is a favorite song or a dislike song and a song that is often listened to or not listened to much. The similarity with the search axis is 0.50.

  When the similarity between the search axis whose value is estimated by the regression equation and the search axis whose value is searched by the discriminant is obtained, the value indicated by the search axis estimated by the regression equation is a predetermined threshold value. Are compared with each other to obtain a binary value, and then the coincidence rate is calculated, and the coincidence rate is used as the similarity of the search axes.

  Thus, the search axis comparison unit 44 calculates the similarity of the search axes.

  The search axis comparison unit 44 calculates the similarity between all of the search axes generated based on the information for the user and all of the search axes generated based on the information for the other users to be compared.

  For example, as shown in FIG. 24, a user's search axis such as a search axis indicating the degree of likes and dislikes, a search axis indicating the frequency of playback, a search axis indicating good or bad sense, a search axis indicating intensity, etc. And other search targets such as a search axis that indicates the level of likes and dislikes, a search axis that indicates the frequency of playback, a search axis that is or is not a famous song, a search axis that is or is not healed, etc. The similarity with each of the search axes is calculated.

  For example, as shown in FIG. 24, the search axis comparison unit 44 is a search axis generated based on information for the user, and is generated based on a search axis indicating the degree of likes and dislikes and information for other target users. The similarity that is 0.5 with the search axis indicating the degree of likes and dislikes is calculated. In addition, the search axis comparison unit 44 is a search axis generated by information for a user, a search axis indicating the degree of likes and dislikes, and a search axis generated by information for another target user. The similarity that is 0.8 with the search axis indicating is calculated. Similarly, the search axis comparison unit 44 is a search axis generated by information for users, a search axis indicating the degree of likes and dislikes, and a search axis generated by information for other target users. Calculate a similarity that is 0.9 or 0.6 with the search axis that is or is not or the search axis that is or is not healed, respectively.

  Furthermore, as shown in FIG. 24, the search axis comparison unit 44 is generated by a search axis indicating the frequency of reproduction, which is a search axis generated by information for the user, and information for other target users. 0.4, 0.5, and 0.9, respectively, a search axis that indicates the degree of likes and dislikes, a search axis that indicates the frequency of playback, a search axis that is famous or not, or a search axis that is or is not healed , Or calculate a similarity that is 0.5.

  Further, as shown in FIG. 24, the search axis comparison unit 44 is based on a search axis generated based on information for a user, a search axis that has a good or no sense, and information for another target user. 0.7, each of the generated search axes, the search axis indicating the degree of likes and dislikes, the search axis indicating the frequency of playback, the search axis that is or is not a masterpiece, or the search axis that is or is not healed Calculate the similarity that is 0.6, 0.4, or 0.1.

  Furthermore, as shown in FIG. 24, the search axis comparison unit 44 is generated based on the search axis indicating the intensity, which is a search axis generated based on information for the user, and information for other target users. A search axis that indicates the degree of likes and dislikes, a search axis that indicates the frequency of playback, a search axis that is or is not a masterpiece, or a search axis that is or is not healed, respectively. Or calculate a similarity that is 0.8.

  When the search axis comparison unit 44 calculates the similarity of the search axes, the display unit 45 presents a search axis having a high similarity to the user.

  For example, in the case of indicating that the similarity of a larger value is similar, among the calculated similarities of the search axes, the highest similarity, the similarity greater than a predetermined threshold value, or in advance in descending order. A predetermined number of similarities are selected, and the display unit 45 presents the selected similarity search axis to the user as a similar search axis.

  More specifically, when the similarity of the search axes shown in FIG. 24 is calculated, as shown in FIG. 25, the highest similarity that is 0.9 is selected and a search generated based on information for the user. The search axis that indicates the degree of likes and dislikes that is the axis, and the search axis that is a search axis that is generated based on information for the other target users, or a search axis that is a famous song or not, is similar to the user. Presented.

  When the target other user is Mr. A, at this time, a message such as “Your favorite song is a famous song for Mr. A” may be displayed on the display unit 45.

  In addition, a search axis indicating the frequency of reproduction, which is a search axis generated by information for the user, and a search axis generated by information for other target users, a search axis which is a famous song or not Are presented to the user as similar search axes.

  Furthermore, a search axis indicating the intensity, which is a search axis generated based on information for the user, and a search axis indicating the frequency of reproduction, which is a search axis generated based on information for other target users, It is presented to the user as a similar search axis.

  When the target other user is Mr. A, at this time, a message such as “Mr. A frequently plays songs that you think is intense” may be displayed on the display unit 45. Good.

  This message is generated by storing a message template in advance and inserting a word extracted from the name of the search axis into this template.

  Moreover, you may make it show to a user the search axis which conflicts among the search axes produced | generated by the information with respect to a user, and the search axes produced | generated by the information with respect to another user. For example, when the correlation value of the positive or negative value is used as the similarity of the search axis, the similarity having the largest absolute value is selected from the calculated similarities of the negative value of the search axis, and the display unit 45 Thus, the search axis of the selected similarity may be presented to the user.

  Further, an unrelated search axis may be presented to the user among a search axis generated based on information for the user and a search axis generated based on information for another user. For example, when the correlation value is the similarity of the search axis, the similarity with the smallest absolute value is selected from the calculated similarities of the search axis, and the search axis of the selected similarity is displayed by the display unit 45. May be presented to the user.

  Furthermore, among search axes generated by information for users and search axes generated by information for other users, similar search axes, that is, similar search axes generated by information for different users, For example, a search axis indicating the degree of likes and dislikes of the user and other users, or the similarity between the search axis of the user and other users that is or is not a masterpiece, and the display unit 45 is used for different users. You may make it show the similarity of the similar search axis produced | generated by information.

  When a similar search axis is presented to the user, the display unit 45 displays the position of the music, which is a two-dimensional or higher space, and is the axis of the space in the space where each axis is the search axis. An image showing a search axis similar to this search axis is displayed in a thinly superimposed manner on the image showing the search axis used by the.

  As shown in FIG. 26, the display unit 45 is a search axis generated based on information for the user, an image showing a search axis indicating the degree of likes and dislikes, and information generated for other user A's information. Each of the images indicating the search axes indicating the degree of likes and dislikes is displayed darkly as an axis of a two-dimensional space, as indicated by a solid line in FIG.

  The display unit 45 is a search axis generated by information for each user for each piece of music, the value indicated by the search axis indicating the degree of likes and dislikes, and the likes and dislikes generated by information about other users A A symbol such as a circle indicating each piece of music is displayed at a position in a two-dimensional space determined from the value indicated by the search axis indicating the degree.

  Furthermore, the display unit 45 is a famous song that is a search axis generated by information for another user A, similar to a search axis indicating the degree of likes and dislikes, which is a search axis generated by information for the user. Alternatively, an image indicating a search axis that is not so is displayed lightly on a two-dimensional space as indicated by a dotted line in FIG.

  Further, the display unit 45 is a search axis generated by information for a user similar to a search axis indicating a degree of likes and dislikes, which is a search axis generated by information for another user A, and has a good sense. Alternatively, an image indicating a search axis that is not so is displayed lightly on a two-dimensional space as indicated by a dotted line in FIG.

  Note that the inclination of a search axis similar to the search axis that is the axis in a two-dimensional or larger space for presenting music is determined by the similarity. For example, when the similarity that is the correlation value is calculated, the inclination of the similar search axis with respect to the reference axis or the axis that is orthogonal thereto is taken as the similarity. Specifically, when the similarity that is the correlation value of 0.8 is calculated, the inclination of the similar search axis with respect to the axis orthogonal to the reference axis is set to 0.8. When the similarity that is the correlation value of −0.2 is calculated, the inclination of the similar search axis with respect to the reference axis is −0.2. In addition, for example, when the similarity is a value of 0 to 1 and a larger similarity indicates that it is more similar, the search axis that is the axis in a two-dimensional or higher space that presents the music The angle of the search axis similar to is obtained by tan-1 (1-similarity).

  In this way, the search axis image that is the axis in the two-dimensional or higher space presenting the music and the image that shows the search axis similar to these search axes are displayed in a thin and superimposed manner.

  In addition, when the user performs an operation of selecting a search axis with a high degree of similarity, the music search device 11 presents music that matches the selected search axis to the user.

  For example, when the value indicated by the selected search axis is estimated by the regression equation, the music search device 11 has a large value indicated by the search axis for music whose value indicated by the search axis is greater than a predetermined threshold value. Present to the user in order.

  For example, when the value indicated by the selected search axis is estimated by a discriminant, the music search device 11 presents the user with a song whose value indicated by the search axis is a predetermined value of No or Yes. .

  That is, as shown in FIG. 27, a desired search axis, for example, a search axis having a high similarity to a predetermined search axis displayed on the display unit 45 when the operation unit 36 is operated by the user, for example, When a button for selecting a search axis that is generated based on information for other user A and is or is not a masterpiece is clicked, the music search device 11 uses the selected search axis as a reference. Music 4, Music 8, Music 14, Music 7, etc., which are music that is considered to be a masterpiece, are presented to the user. In this case, for example, the music search device 11 presents to the user a playlist that includes music that applies to the selected search axis.

  FIG. 28 is a block diagram showing a configuration for calculating the similarity of search axes in the configuration shown in FIG.

  The search axis comparison unit 44 includes an axis value calculation unit 101 and a similarity calculation unit 102.

  The search axis comparison unit 44 acquires a search axis for calculating the similarity from the search axis database 33 or the other user search axis database 38. That is, the search axis comparison unit 44 acquires a regression equation or discriminant for estimating a value indicated by the search axis for calculating the similarity from the search axis database 33 or the other user search axis database 38.

  The search axis comparison unit 44 acquires the feature amount of the music owned by the user from the feature amount database 32. Then, the axis value calculation unit 101 of the search axis comparison unit 44 applies the regression equation or discriminant to the feature amount of each song owned by the user, thereby obtaining the value indicated by each search axis for the song. calculate.

  The similarity calculation unit 102 of the search axis comparison unit 44 calculates the similarity of the value indicated by the search axis for obtaining the similarity, and sets the similarity as the similarity of the search axis.

  Note that the search axis comparison unit 44 acquires the feature quantity of a song owned by another user from the other user feature quantity database 40, and calculates the similarity of the search axis from the feature quantity of the song owned by the other user. You may make it do.

  Further, the search axis comparison unit 44 may calculate the similarity of the search axes from the feature amount of the music owned by the user and the feature amount of the music owned by another user.

  Further, the search axis comparison unit 44 may calculate the similarity of the search axes from the feature quantities of some of the music pieces of the user or other users. For example, the search axis comparison unit 44 may calculate the similarity of the search axes from a predetermined number of music feature quantities. Further, for example, the search axis comparison unit 44 may calculate the similarity of the search axes from a predetermined number of feature quantities of randomly selected music pieces.

  The search axis comparison unit 44 obtains a similar search axis from the similarity of the search axes and outputs a similar search axis.

  FIG. 29 is a flowchart for explaining the process of presenting a similar search axis. In step S <b> 71, the search axis comparison unit 44 acquires each of the two search axes for comparing the similarity from the search axis database 33 or the other user search axis database 38.

  In step S72, the search axis comparison unit 44 calculates the values of the two search axes of all the music pieces from the feature quantities of all the music pieces owned by the user. That is, for example, the search axis comparison unit 44 acquires the feature values of all the songs owned by the user from the feature value database 32, and the axis value calculation unit 101 of the search axis comparison unit 44 sets the feature values of the respective songs. The regression formula or discriminant of each of the two search axes is applied to calculate the values indicated by the respective search axes for all the songs owned by the user.

  In step S73, the search axis comparison unit 44 calculates the similarity between the two search axes. That is, the similarity calculation unit 102 of the search axis comparison unit 44 calculates the similarity of the value indicated by the search axis calculated in step S72 as the similarity of the search axis.

  In step S74, the search axis comparison unit 44 determines whether the similarity is calculated for all combinations of search axes, and if it is determined that the similarity is not calculated for all combinations of search axes, The procedure returns to step S71 and repeats the above-described processing.

  If it is determined in step S74 that similarities have been calculated for all combinations of search axes, the procedure proceeds to step S75, and in step S75, the display unit 45 displays the highest number of combinations of search axes with high similarity. Is presented to the user, and the process of presenting the similar search axis ends. More specifically, in step S75, the search axis comparison unit 44 selects a predetermined number of similarities in descending order of the calculated similarity, and selects the search axes with the selected similarity as similar search axes. To the display unit 45. The display unit 45 displays similar search axes supplied from the search axis comparison unit 44.

  In this case, even if the display unit 45 displays the name of a similar search axis, as described with reference to FIG. 26, the display unit 45 is a two-dimensional space or more, and each axis is used as a search axis. In a space, an image showing a search axis similar to those search axes may be displayed thinly superimposed on an image showing a search axis used for displaying the position of the music, which is the axis of the space. .

  Next, a case will be described in which the position of the user's music and the position of the music of another user are displayed in a two-dimensional or more space with the search axis for presenting the music as the axis.

  For example, as shown in FIG. 30, a search axis indicating the degree of likes and dislikes generated by information for the user and a search axis indicating the degree of likes and dislikes generated by information for other user A The position of the user's music and the position of Mr. A's music are displayed in a two-dimensional space. In FIG. 30, a circle indicates the position of the user's music in a two-dimensional space with the search axis as the axis, and a cross indicates the position of Mr. A's music in that space.

  FIG. 31 is a block diagram illustrating a configuration in which the position of the user's music and the position of the music of another user are displayed in a two-dimensional or higher space in the configuration illustrated in FIG.

  When the search axis information indicating the search axis used for the search is supplied from the operation unit 36 to the search axis acquisition unit 37 according to the user's operation, the search axis acquisition unit 37 displays the search axis indicated by the search axis information. Is acquired from the search axis database 33 or the other user search axis database 38. That is, the search axis acquisition unit 37 acquires a regression equation or discriminant for estimating a value indicated by the search axis indicated by the search axis information from the search axis database 33 or the other user search axis database 38.

  For example, as illustrated in FIG. 32, when search axis information indicating “a song that is often listened to” and “a song that Mr. A likes” is supplied from the operation unit 36, the search axis acquisition unit 37 reads “ The search axis of the name “Song”, indicating the degree of listening well, is acquired from the search axis database 33, and the search axis of the name “Song A's favorite song” is another search axis. A search axis indicating the degree of likes and dislikes of the user A is acquired from the other user search axis database 38.

  That is, in this case, the search axis acquisition unit 37 acquires the search axis regression expression indicating the degree of whether or not to listen well, which is the search axis of the name “songs that are often listened to”, from the search axis database 33. A search axis regression formula indicating the degree of likes and dislikes of other user A, who is the search axis with the name “A's favorite song”, is acquired from the other user search axis database 38.

  The search axis acquisition unit 37 supplies the acquired search axis, that is, the regression equation or discriminant thereof to the search axis value calculation unit 39.

  The search axis value calculation unit 39 acquires the feature amount of the song owned by the user from the feature amount database 32, and acquires the feature amount of the song owned by another user from the other user feature amount database 40. Then, the search axis value calculation unit 39 uses the search axis regression formula or discriminant supplied from the search axis acquisition unit 37 to the feature quantity of each song owned by the user or the feature quantity of the song owned by another user. Is applied to calculate the value indicated by each search axis for the music.

  For example, the search axis value calculation unit 39 acquires the feature quantities of all the songs owned by the user from the feature quantity database 32, and all the songs owned by other target users from the other user feature quantity database 40. And the value indicated by each search axis for each song is calculated from the feature amount of the song.

  Specifically, as shown in FIG. 33, for example, the search axis value calculation unit 39 listens to the music feature amount of the song 1 including the feature amount 1 which is 15 and the feature amount 2 which is 94. A search axis indicating the degree of whether or not the song 1 is often listened to by applying a regression equation of the search axis indicating the degree and a regression expression of the search axis indicating the degree of likes and dislikes of other user A, respectively. And a value that is 3 of the search axis indicating the degree of likes and dislikes of other user A, is calculated.

  In addition, the search axis value calculation unit 39 uses a search axis regression formula indicating the degree of listening to the music feature quantity of the music piece 2 including the feature quantity 1 which is 36 and the feature quantity 2 which is 29, and other users. By applying the search axis regression expression indicating the degree of likes and dislikes of Mr. A, the value that is 1 on the search axis indicating the degree of whether or not the song 2 is listened well and the other user A Calculate a value that is 2 on the search axis indicating the degree of likes and dislikes.

  Similarly, the search axis value calculation unit 39 calculates the regression equation of the search axis indicating the degree of listening well to the music feature quantities of the music pieces 3 to 6 and the degree of likes and dislikes of A who is another user. By applying each of the search axis regression equations shown, the values of the search axis indicating the degree of listening well for each of the music pieces 3 to 6 and the degree of likes and dislikes of the other user A are shown. Calculate the value of the search axis.

  The search axis value calculation unit 39 supplies the value of each search axis for the music to the display unit 45.

  The display unit 45 determines the position of the song owned by the user and the position of the song owned by another user from a value indicated by each search axis for the song in a two-dimensional or more space around the search axis. Are displayed in layers.

  For example, as shown in FIG. 30, the display unit 45 has a search axis indicating the degree of likes and dislikes generated by the information for the user and a search indicating the degree of likes and dislikes generated by the information for the other user A. The position of the user's music and the position of Mr. A's music are displayed in a two-dimensional space with the axis as the axis.

  FIG. 34 is a flowchart for describing processing for displaying other users' content in an overlapping manner. In step S91, the search axis acquisition unit 37 acquires two or more search axes used for display from the search axis database 33 or the other user search axis database 38. That is, the search axis comparison unit 44 acquires from the search axis database 33 or the other user search axis database 38 the regression equation or discriminant for estimating the values of two or more search axes used for display.

  In step S <b> 92, the search axis value calculation unit 39 calculates the values of the search axes of all the music pieces of the user and the other user from the feature amount of the user's music and the feature amount of the other user's music. . That is, the search axis value calculation unit 39 acquires the feature quantities of all the songs owned by the user from the feature quantity database 32, and all the songs owned by other target users from the other user feature quantity database 40. Get the feature quantity. Then, the search axis value calculation unit 39 applies the regression formula or discriminant of each search axis to the feature quantity of each song owned by the user or the feature quantity of the song owned by another user, thereby Calculate the value of each search axis for.

  In step S93, the display unit 45 arranges and displays all the music pieces of the user and other users in a space having the search axis as a dimension, and the process of displaying the contents of other users in an overlapping manner ends. For example, in step S93, as shown in FIG. 30, the display unit 45 is a position determined by the value of each search axis for each piece of music in a two-dimensional or higher space having the search axis as the axis of that dimension. And the position of each music is displayed.

  FIG. 35 is a flowchart for explaining content selection processing. Steps S121 to S124 are the same as steps S71 to S74 in FIG.

  If it is determined in step S124 that similarities have been calculated for all combinations of search axes, the procedure proceeds to step S125. In step S125, the search axis comparison unit 44 refers to similarities and searches for similarities. Select an axis. The search axis comparison unit 44 outputs the selected search axis to the display unit 45.

  In step S126, the display unit 45 displays the selected similar search axes in an overlapping manner. For example, as described with reference to FIG. 26, the display unit 45 displays one image of the selected similar search axis images in a dark color, and displays the other image in a thin layer on the image.

  In step S127, the search axis value calculation unit 39 calculates the value of the search axis from the feature amounts of the user's music and other users' music in the same process as in step S92. For example, the search axis value calculation unit 39 calculates the value of the search axis that is darkly displayed on the display unit 45 from the feature amounts of the user's music and other users' music.

  In step S128, the display unit 45 displays the position of the search axis corresponding to the value of the search axis for the user's music and other users' music. More specifically, the display unit 45 displays an image indicating the position of the search axis corresponding to the value of the search axis for the user's music and other users' music.

  In step S129, the corresponding music selection unit 41 selects music at a position belonging to the range of the search axis designated by the user, and the content selection process ends. For example, in step S129, the corresponding music selection unit 41 sets the search axis within a predetermined range based on the search axis, which is specified by information supplied from the operation unit 36 according to the user's operation. The content to which the position belongs is selected.

  As described above, the personalized search axis in the search is used among users and shared among users.

  The music search device 11 generates a search axis that is personalized to the user automatically or in response to a user input. The generated search axis is shared among users.

  In the music search device 11, the user can use a general search axis prepared in advance (for example, a search axis for searching content from the metadata itself), his / her search axis generated in his / her music search device 11, The music can be searched using the search axis of the other user generated in the music search device 11 owned by the other user.

  The music search device 11 compares the similarities between the generated search axes and the search axes of other users. The music search device 11 presents a user search axis having a high degree of similarity to the search axes of other users to the user. Alternatively, the music search device 11 presents the search axis of another user who is highly similar to the user's search axis to the user.

  When the user selects the presented search axis, the music search device 11 searches for music using the selected search axis.

  The music search device 11 arranges an image indicating a music to be searched on a two-dimensional or higher map having two or more search axes as coordinates in searching for music, and the range on the map is Search for music by being specified. The music search device 11 can display an image indicating a music owned by another user on the map.

  The music search device 11 compares each of the search axes of the user with each of the search axes of the other users, obtains a similarity, and presents a combination of similar search axes to the user. , It is possible to know what search axis is used for other users.

  Since the music search device 11 displays an image indicating its own content and an image indicating the content of another user on a map created using a predetermined search axis, the user can select the content of the other user. Can be known.

  As described above, the user can search for music with a personalized search axis. For example, a user searches for a favorite song, a song that is often listened to, a fashionable song (a song that the user thinks is fashionable), a song that the user wants to listen to during a meal (a song that the user wants to hear during a meal), and the like. Can do.

  Thus, when the search axis is generated, the content can be presented. Further, a search axis serving as a reference for presenting content is generated by learning, and a search axis determined by the content feature amount and the history of content reproduction by the user or evaluation of the user content is generated by learning. When the display related to the presentation of the content as a reference is controlled, the presentation of the content tailored to the user can be displayed more easily.

  Note that the music search device 11 may be a stationary electronic device such as a hard disk recorder, a hard disk player, or a home game machine, or a portable electronic device such as a portable game machine, a portable player, Or it can be set as a mobile phone.

  The series of processes described above can be executed by hardware or can be executed by software. When a series of processing is executed by software, a program constituting the software executes various functions by installing a computer incorporated in dedicated hardware or various programs. For example, it is installed from a program recording medium in a general-purpose personal computer or the like.

  FIG. 36 is a block diagram illustrating a hardware configuration example of a computer that executes the above-described series of processing by a program.

  In a computer, a central processing unit (CPU) 201, a read only memory (ROM) 202, and a random access memory (RAM) 203 are connected to each other by a bus 204.

  An input / output interface 205 is further connected to the bus 204. The input / output interface 205 includes an input unit 206 composed of a keyboard, mouse, microphone, etc., an output unit 207 composed of a display, a speaker, etc., a storage unit 208 composed of a hard disk or nonvolatile memory, and a communication unit 209 composed of a network interface. A drive 210 for driving a removable medium 211 such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory is connected.

  In the computer configured as described above, the CPU 201 loads, for example, the program stored in the storage unit 208 to the RAM 203 via the input / output interface 205 and the bus 204 and executes the program. Is performed.

  The program executed by the computer (CPU 201) is, for example, a magnetic disk (including a flexible disk), an optical disk (CD-ROM (Compact Disc-Read Only Memory), DVD (Digital Versatile Disc), etc.), a magneto-optical disk, or a semiconductor. The program is recorded on a removable medium 211 that is a package medium composed of a memory or the like, or provided via a wired or wireless transmission medium such as a local area network, the Internet, or digital satellite broadcasting.

  The program can be installed in the computer by loading the removable medium 211 in the drive 210 and storing it in the storage unit 208 via the input / output interface 205. Further, the program can be installed in a computer by being received by the communication unit 209 via a wired or wireless transmission medium and stored in the storage unit 208. In addition, the program can be installed in the computer in advance by storing the program in the ROM 202 or the storage unit 208 in advance.

  The program executed by the computer may be a program that is processed in time series in the order described in this specification, or in parallel or at a necessary timing such as when a call is made. It may be a program for processing.

  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.

  11-1, 11-2, 11 Music search device, 12 servers, 13 networks, 31 search axis learning unit, 32 feature quantity database, 33 search axis database, 34 learning data generation unit, 35 playback history feedback database, 36 operation Unit, 37 search axis acquisition unit, 38 other user search axis database, 39 search axis value calculation unit, 40 other user feature quantity database, 41 corresponding music selection unit, 42 content reproduction unit, 43 content database, 44 search axis comparison unit, 45 display unit, 46 communication unit, 101 axis value calculation unit, 102 similarity calculation unit, 201 CPU, 202 ROM, 203 RAM, 208 storage unit, 209 communication unit, 211 removable media

Claims (6)

In an information processing apparatus that requests information to be presented to a user, which is information related to content,
Search axis generation means for generating a search axis which is a search axis arbitrarily designated by the user and serves as a reference for presenting the content;
An input means for inputting, as a feedback value of the user, a degree to which each of the plurality of contents applies for the search axis;
Learning data generation means for generating learning data for obtaining a regression equation or discriminant for obtaining a value indicated by the search axis based on the feedback value;
Search axis learning means for learning a regression equation or a discriminant for obtaining a value indicated by a search axis using the feature amount of the content and the learning data;
An information processing apparatus comprising: calculation means for calculating a value of the search axis for content to be presented to a user based on the regression equation or discriminant equation. The information processing apparatus according to claim 1, wherein the calculation unit calculates a value indicated by each of the plurality of search axes. The information processing according to claim 1, further comprising display control means for displaying an image indicating the content at a position corresponding to a value of the content indicated by the search axis in a space having the search axis as a dimension axis. apparatus. The information processing apparatus according to claim 1, further comprising: a selecting unit configured to select the content in a predetermined range indicated by the search axis among the contents in which the value indicated by the search axis is displayed. In an information processing method of an information processing apparatus that requests information to be presented to a user, which is information related to content,
The search axis generation means generates a search axis which is a search axis arbitrarily designated by the user and serves as a reference for presenting the content,
The input means inputs, as the feedback value of the user, the degree to which each of the plurality of contents applies for the search axis,
The learning data generating means generates learning data for obtaining a regression equation or a discriminant for obtaining a value indicated by the search axis based on the feedback value,
The search axis learning means learns a regression equation or a discriminant for obtaining a value indicated by the search axis using the feature amount of the content and the learning data,
An information processing method including a step of calculating a value of the search axis for content to be presented to a user based on the regression equation or discriminant equation. Computer
In an information processing apparatus that requests information to be presented to a user, which is information related to content,
Search axis generation means for generating a search axis which is a search axis arbitrarily designated by the user and serves as a reference for presenting the content;
An input means for inputting, as a feedback value of the user, a degree to which each of the plurality of contents applies for the search axis;
Learning data generation means for generating learning data for obtaining a regression equation or discriminant for obtaining a value indicated by the search axis based on the feedback value;
Search axis learning means for learning a regression equation or a discriminant for obtaining a value indicated by a search axis using the feature amount of the content and the learning data;
A program that functions as an information processing apparatus including a calculation unit that calculates a value of the search axis based on the regression equation or the discriminant for content to be presented to a user.

Images