WO2006134866A1 - 情報処理装置、方法、およびプログラム - Google Patents
情報処理装置、方法、およびプログラム Download PDFInfo
- Publication number
- WO2006134866A1 WO2006134866A1 PCT/JP2006/311742 JP2006311742W WO2006134866A1 WO 2006134866 A1 WO2006134866 A1 WO 2006134866A1 JP 2006311742 W JP2006311742 W JP 2006311742W WO 2006134866 A1 WO2006134866 A1 WO 2006134866A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- cluster
- music
- content
- identified
- metadata
- Prior art date
Links
- 230000010365 information processing Effects 0.000 title claims abstract description 26
- 238000000034 method Methods 0.000 title claims description 165
- 238000012552 review Methods 0.000 claims abstract description 12
- 238000003672 processing method Methods 0.000 claims abstract description 7
- 230000033764 rhythmic process Effects 0.000 claims abstract description 6
- 238000011156 evaluation Methods 0.000 claims description 22
- 238000004364 calculation method Methods 0.000 abstract description 58
- 101710114762 50S ribosomal protein L11, chloroplastic Proteins 0.000 abstract description 15
- 238000000605 extraction Methods 0.000 description 57
- 230000000875 corresponding effect Effects 0.000 description 37
- 238000010586 diagram Methods 0.000 description 34
- 239000013598 vector Substances 0.000 description 29
- 238000012545 processing Methods 0.000 description 21
- 238000013507 mapping Methods 0.000 description 16
- 101710156159 50S ribosomal protein L21, chloroplastic Proteins 0.000 description 12
- 101710082414 50S ribosomal protein L12, chloroplastic Proteins 0.000 description 11
- 101710164994 50S ribosomal protein L13, chloroplastic Proteins 0.000 description 10
- 101710087140 50S ribosomal protein L22, chloroplastic Proteins 0.000 description 10
- 101710118399 50S ribosomal protein L24, chloroplastic Proteins 0.000 description 10
- 101710115003 50S ribosomal protein L31, chloroplastic Proteins 0.000 description 10
- 101100328077 Bos taurus CL43 gene Proteins 0.000 description 8
- 238000007781 pre-processing Methods 0.000 description 6
- 239000000284 extract Substances 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000001020 rhythmical effect Effects 0.000 description 2
- 239000011435 rock Substances 0.000 description 2
- 101710142585 50S ribosomal protein 6, chloroplastic Proteins 0.000 description 1
- 101000725126 Spinacia oleracea 50S ribosomal protein L35, chloroplastic Proteins 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- PWPJGUXAGUPAHP-UHFFFAOYSA-N lufenuron Chemical compound C1=C(Cl)C(OC(F)(F)C(C(F)(F)F)F)=CC(Cl)=C1NC(=O)NC(=O)C1=C(F)C=CC=C1F PWPJGUXAGUPAHP-UHFFFAOYSA-N 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G11—INFORMATION STORAGE
- G11B—INFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
- G11B27/00—Editing; Indexing; Addressing; Timing or synchronising; Monitoring; Measuring tape travel
- G11B27/10—Indexing; Addressing; Timing or synchronising; Measuring tape travel
- G11B27/19—Indexing; Addressing; Timing or synchronising; Measuring tape travel by using information detectable on the record carrier
- G11B27/28—Indexing; Addressing; Timing or synchronising; Measuring tape travel by using information detectable on the record carrier by using information signals recorded by the same method as the main recording
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
- G06F16/43—Querying
- G06F16/435—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G11—INFORMATION STORAGE
- G11B—INFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
- G11B27/00—Editing; Indexing; Addressing; Timing or synchronising; Monitoring; Measuring tape travel
- G11B27/10—Indexing; Addressing; Timing or synchronising; Measuring tape travel
- G11B27/102—Programmed access in sequence to addressed parts of tracks of operating record carriers
- G11B27/105—Programmed access in sequence to addressed parts of tracks of operating record carriers of operating discs
Definitions
- the present invention relates to an information processing apparatus, an information processing method, and a program, and in particular, classifies content into clusters, manages content features using clusters into which content is classified, and searches and recommends content.
- CBF content-based filtering
- the preference information of the user is obtained by regarding the metadata of the music as a feature vector and adding the feature vectors of the music according to the user's operation (play, record, skip, delete, etc.) for the music. It was generated. For example, the feature vector of the reproduced music is multiplied by 1, the feature vector of the recorded music is multiplied by 2, the feature vector of the skipped music is multiplied by 1, and the feature vector of the erased music is multiplied by two. .
- Patent Document 1 Japanese Patent Application Laid-Open No. 2004-194107
- the present invention has been made in view of such a situation, and the user is able to search for the content similar to the content of the user's preference or the content similar to the designated content with a smaller amount of calculation, It is something that can be presented.
- An information processing apparatus is an information processing apparatus for selecting a content satisfying a predetermined condition from the power of a content group, wherein each content constituting the content group is divided into content meta of content.
- Content classification means for classifying into each of a plurality of first clusters in each layer according to data, and the first cluster in the layer into which each content and each content are classified
- Holding means for holding a database indicating a correspondence relationship with the above, and the first corresponding to the predetermined condition for each hierarchy.
- the present invention is characterized in that it comprises specifying means for specifying a cluster of and specifying content corresponding to the specified first cluster, and presenting means for presenting the content specified by the specifying means.
- a storage unit is further provided which stores a preference value indicating a degree of preference of the user in association with each first cluster in which the content is classified by the content classification unit, and the identification unit includes the storage
- the first cluster can be specified based on the preference value stored by the means, and the content corresponding to the specified first cluster can be specified.
- the content may be music, and the metadata may include at least one of tempo, beat, or rhythm of the music.
- the metadata may include review text for the corresponding content.
- Metadata classification means for classifying the metadata of the content into any one of a plurality of second clusters and assigning the hierarchy to the second cluster, the content classification means assigning each content In each of the layers, the plurality of first clusters can be classified.
- the identification means further indicates the content with a degree of similarity with the content serving as the similarity source. Can be identified.
- the content is identified by the similarity weighted by the weight for each hierarchy according to the weight of attribution of the content as the similarity source to the first cluster. Can.
- An information processing method is an information processing method of an information processing apparatus for selecting a content satisfying a predetermined condition from the power of a content group, and each content constituting the content group is A database that shows the correspondence between the content and the clusters in the hierarchy into which each content and each content were classified, and the classification step of classifying the content into a group of multiple clusters in each of the layers according to content metadata. And a step of specifying the content corresponding to the designated cluster, and specifying the content corresponding to the designated cluster. And providing a presentation step.
- a program is a program for selecting a content that satisfies a predetermined condition from the power of a content group, wherein each content constituting the content group is a content metadata And storing a database indicating a correspondence between each content and each cluster in the hierarchy into which each content is classified.
- a process including a step, a specifying step of specifying the cluster corresponding to the predetermined condition for each hierarchy, specifying a content corresponding to the specified cluster, and a presenting step of presenting the specified content Running on a computer.
- each content forming a content group is classified into! / ⁇ of a plurality of clusters in each of the layers according to the metadata of the content, and each content is And a database indicating the correspondence with the cluster in the hierarchy into which each content is classified is held, the cluster corresponding to the predetermined condition is designated for each hierarchy, and the cluster corresponds to the designated cluster.
- Content is identified and the identified content is presented.
- FIG. 1 is a block diagram showing a configuration example of a recommendation system to which the present invention is applied.
- FIG. 2 A diagram showing the concept of clusters for classifying metadata and a cluster layer.
- FIG. 3 is a diagram showing an example of cluster information.
- FIG. 4 is a diagram showing an example of HD information.
- FIG. 5 is a diagram showing an example of preference information.
- FIG. 6 is a diagram for explaining a method of selecting two types of methods from the first to fourth methods of clustering.
- FIG. 7 is a diagram for explaining a method of selecting two types of methods from the first to fourth methods of clustering.
- FIG. 8 is a diagram for explaining a method of selecting two types of clustering from the first to fourth methods.
- FIG. 9 is a diagram for explaining a method of selecting two types of clustering from the first to fourth methods.
- FIG. 10 is a diagram for explaining a method of selecting two types of methods from the first to fourth methods of clustering.
- FIG. 11 is a flowchart illustrating a first similar music search process.
- FIG. 12 is a flowchart illustrating a second similar music search process.
- FIG. 13 is a flowchart for explaining third similar music search processing.
- FIG. 14 is a flowchart illustrating a first music recommendation process.
- FIG. 15 is a flowchart illustrating a second music recommendation process.
- FIG. 16 is a block diagram showing a configuration example of a general-purpose personal computer.
- FIG. 17 is a block diagram showing an example of another configuration of a recommendation system according to an embodiment of the present invention.
- FIG. 18 is a flowchart illustrating an example of pre-processing when offline.
- FIG. 19 is a diagram showing an example of metadata of each music piece subjected to soft clustering.
- FIG. 20 is a diagram showing an example of metadata of each music.
- FIG. 21 is a diagram showing an example of cluster information.
- FIG. 22 is a flow chart for explaining a fourth similar music search process.
- FIG. 23 is a diagram showing an example of cluster information.
- FIG. 24 is a diagram showing an example of the degree of similarity.
- FIG. 25 is a flowchart illustrating a fifth similar music search process.
- FIG. 26 is a flowchart illustrating a third music recommendation process.
- FIG. 27 shows an example of preference values.
- FIG. 28 is a diagram showing an example of cluster information.
- FIG. 29 is a diagram showing an example of the degree of similarity.
- FIG. 30 is a diagram showing an example of weights.
- FIG. 31 is a diagram showing an example of the degree of similarity.
- FIG. 32 is a flowchart illustrating a fourth music recommendation process.
- FIG. 33 is a diagram showing an example of preference values.
- FIG. 34 is a diagram showing an example of the degree of similarity.
- FIG. 1 shows an example of the configuration of a recommendation system according to an embodiment of the present invention.
- the recommendation system 1 searches for music that matches the user's preference or music similar to a user-specified music, and presents the search to the user.
- the recommendation system 1 can also be applied to the case of recommending content other than music, such as television programs, movies, books, and the like.
- the recommendation system 1 is configured to easily record metadata of many songs to be searched.
- the music database (DB) 11 each music recorded in the music database 11 is clustered based on metadata of the music!
- the keyword setting part 13 which sets the keyword which shows the characteristic of each cluster, respectively, and cluster information database (DB) 14 holding the cluster information of each music are comprised.
- the recommendation system 1 includes a search music specification unit 21 for specifying a music that is a similar source of music to be searched (hereinafter referred to as an original music), a conventional cluster identification with metadata of the original music.
- a cluster mapping unit 22 that maps to an optimal cluster using a method (classification method), a music extraction unit 23 that extracts one or more musics to be presented to the user, and preference information indicating the preference of the user is recorded.
- Preference information database (DB) 24 a preference input unit 25 for inputting user's preferences, a random selection unit 26 for randomly selecting one song from the middle power of the extracted songs, the extracted song and the original song or user
- the similarity calculation unit 27 calculates the similarity with the preference of the user, selects the music with the highest similarity, generates a selection reason statement indicating the reason for selection in the random selection unit 26 or the similarity calculation unit 27 That election and-option reason generating unit 28, composed from the music Hisage radical 113 29 to present a selection statement of the reason the song was selected by the user.
- the music database (DB) 11 corresponds to CDDB (CD Data Base), Music Navi, or the like, which is a data server on the Internet that supplies metadata of music recorded on music CDs.
- CDDB CD Data Base
- Music Navi or the like, which is a data server on the Internet that supplies metadata of music recorded on music CDs.
- the clustering unit 12 shows each item (title, artist name, genre, review text, tempo, beat, rhythm, etc.) of the metadata of the music for all the music in the music database 11 as shown in FIG.
- the music is classified (clustered) into any one of a plurality of clusters provided in the cluster layer where the actual information of each item is classified into one of the cluster layers (layers 1 to n).
- one music piece may be classified into a plurality of clusters.
- the distance between clusters (indicating the degree of similarity) existing in the same cluster layer shall be known.
- the clustering method will be described later.
- Cluster information is generated and output to the cluster information database 14.
- each cluster is arbitrary and can encompass multiple songs.
- a cluster may be provided which can classify only a single piece of music.
- the ID (artist ID, album ID, title ID) of the real information of the music that can be classified only as the cluster ID of the cluster may be used.
- the cluster information database 14 holds cluster information of each music generated by the clustering unit 12. Also, the cluster information database 14 generates cluster ID song ID information indicating the song ID of the song whose metadata has been classified into each cluster based on the cluster information held and holds this. Further, the cluster information database 14 also holds keywords set for each cluster layer and each cluster set by the keyword setting unit 13.
- FIG. 3 shows an example of cluster information.
- FIG. 4 shows an example of cluster music ID information corresponding to the cluster information shown in FIG.
- the processes of the clustering unit 12, the keyword setting unit 13, and the cluster information database 14 need to be performed in advance before the similar music search process and the music recommendation process (described later) are performed.
- the search music specification unit 21 outputs the music ID and metadata of the original music specified by the user to the cluster mapping unit 22.
- the cluster mapping unit 22 selects the optimum cluster using the existing cluster identification method (classification method) for the metadata of the original music input from the search music specification unit 21.
- classification method k-Near est-Neighbor method etc. can be applied.
- cluster information of the original music is already If it exists in the star information database 14, it is read out and supplied to the music extraction unit 23.
- the music extraction unit 23 refers to the cluster information database 14 based on the cluster information of the original music supplied from the cluster mapping unit 22 and acquires the music IDs of the music classified into the same cluster as the original music. It is acquired and supplied to the random selection unit 26 or the similarity calculation unit 27. In addition, the music extraction unit 23 refers to the cluster information database 14 based on the preference information in the preference information database 24 to acquire a music ID of a song that matches the user's preference, and selects the random selection unit 26 or the similarity calculation unit. Supply to 27.
- the preference information database 24 stores preference information indicating the preference of the user.
- preference values indicating the degree of preference of the user for each cluster are recorded. This preference value is a normalized value, and is updated by the preference input unit 25. Further, the preference information database 24 calculates the variance of preference values in each cluster layer, and detects a cluster layer in which the variance of preference values is the smallest (that is, the user preference is concentrated on a specific cluster).
- FIG. 5 shows an example of preference information.
- the preference value for the cluster CL11 is 0.5.
- the preference value for cluster CL32 is 0.1.
- the preference input unit 25 updates the preference value corresponding to each cluster based on the history of the user's operation (play, record, skip, delete, etc.) on the music. Further, the preference input unit 25 notifies the cluster information database 14 of the cluster layer that the user emphasizes based on the setting from the user.
- the random selection unit 26 randomly selects one music ID as the medium power of the music extracted by the music extraction unit 23 and outputs the music ID to the selection reason generation unit 28.
- the similarity calculation unit 27 calculates the similarity between the music extracted by the music extraction unit 23 and the original music or the preference of the user, selects the music with the highest similarity, and outputs it to the selection reason generation unit 28. .
- the random selection unit 26 and the similarity calculation unit 27 need only operate either one or the other.
- the selection reason generation unit 28 also supports the cluster information database 14 in the cluster layer and the cluster.
- the corresponding keyword is acquired, and a selected reason sentence indicating the reason for selection is generated using the acquired keyword or the like, and is output to the music presentation unit 29 together with the music ID of the selected music.
- the selection reason statement is generated as follows. For example, when selecting similar songs or songs matching the preference, the keywords of the cluster layer or the keyword of the cluster set in the priority cluster layer are used. Specifically, when giving priority to the cluster layer corresponding to the review text, “Summer appearing in the review text, and“ seafront ”are not favorite? Generate selected reason sentences such as "". Alternatively, the review text of the selected song is referred to as the selection reason sentence as it is, or a selection reason sentence is generated using a word extracted from the review text of the selected song. Note that Tf / idlf-past can be applied to extract words used for selection reason sentences from review texts.
- the music presentation unit 29 includes, for example, a display, and presents the user with the music of the music ID input from the selection reason generation unit 28 and the selection reason sentence.
- the clustering method may be any method, but the clustering method and the distance measure that are most suitable for each cluster layer are selected. For example, if actual information in the metadata is a numerical value, in the case of a title, etc., make a numerical value by using a quantitative method such as principal component analysis, define a distance measure such as Euclidean distance, and cluster. become.
- Typical clustering methods include K-means method and hierarchical clustering method.
- clustering for example, constrained clustering reflecting preference distance.
- a partial survey of correct answers a set of near-preferred real information, a set of distant real information, etc.
- a numerical expression, distance, and clustering method will be used.
- it is desirable to select a clustering method ie, a clustering method with different characteristics in which the independence of each formed cluster layer is high.
- first to fourth methods a method of selecting two types of clustering methods having different characteristics with respect to the intermediate force of four types of clustering methods (hereinafter referred to as first to fourth methods) will be described.
- first to fourth methods two types of clustering methods having different characteristics with respect to the intermediate force of four types of clustering methods
- the artists A to C are clustered in the cluster CL 1
- the artists D to G are clustered in the cluster CL 2
- Is cluster CL1 artist C to F is cluster CL2, and artist 0 3 ⁇ 4 [is cluster to cluster CL3.
- artists A, D, G, J are cluster CL1, artist E, H Are clustered in cluster CL2, artists C, F, and I are clustered in cluster CL3, and artists D, I, J are clustered in cluster CL1, artists E to G are clustered in cluster CL2, artists A to C and H in the fourth method.
- Force S Cluster Suppose that it is clustered in CL3.
- the duplication rate (%) of the results according to the first to fourth methods is as shown in FIG. That is, the overlap rate of the first method and the second method is 0.8, the overlap rate of the first method and the third method is 0.3, and the overlap rate of the first method and the fourth method is 0.4, The overlap rate of the second and third methods is 0.3, the overlap rate of the second and fourth methods is 0.3, and the overlap rate of the third and fourth methods is 0.4.
- the characteristics of the two methods are considered to differ as the overlapping rate shown in FIG. 7 decreases, so that the combination of the first and third methods with the overlapping rate of 0.3 as the second method, the second method It is desirable to adopt a combination of the third method and the second method and the fourth method.
- the result shown in FIG. 8 is obtained when the user himself / herself determines whether or not two of the artists Yano 3 [2] should be classified into the same cluster.
- 1 means that it should be classified into the same cluster
- 0 means that it should be classified into a different cluster. That is, in the same figure, for example, it is indicated that it is judged that the taste A should be classified into the same cluster as the artist B, C, F, H, I.
- the artist B is the artist C , D, E, J are shown to be classified into the same cluster.
- the accuracy rates of the above-described first to fourth methods are as shown in FIG. That is, the correct rate of the first method is 62.2%, the correct rate of the second method is 55.6%, and the correct rate of the third method is 40.0%, the fourth The accuracy rate of the method is 66.7%.
- the overlap rates of the first to fourth methods are calculated as shown in FIG.
- the accuracy rate is extremely low from the results shown in Fig. 9!
- a specific method is identified, and the specified method is not included!
- the duplication rate of the accuracy rate among the combinations is the lowest.
- the third method is identified as a method with an extremely low accuracy rate, and the combination of the second method and the fourth method is regarded as the one with the lowest overlapping rate of correct answers among combinations not including the third method. It is selected.
- an absolute threshold may be specified for the above-mentioned duplication rate or accuracy rate, and a ⁇ method that can not satisfy the threshold may be excluded!
- a comprehensive index such as the following two examples is created based on two indices (the overlapping rate and the correct answer rate), and a clustering method of You may choose to select a combination.
- the clustering unit 12, the keyword setting unit 13, and the cluster information database 14 are operating as pre-processing for performing similar music search processing and music recommendation processing described below, and the cluster information database 14 has already been operated. Are set for each cluster layer and each cluster set by the cluster information set by the clustering unit 12, the cluster music ID information generated by the cluster information database 14, and the cluster music ID information set by the keyword setting unit 13. Keywords will be retained.
- FIG. 11 is a flow chart for explaining the first similar music search process.
- the cluster information database 14 is input from the preference input unit 25 as pre-processing of the first similar music search processing. It is assumed that the layer numbers of higher priority layers are sequentially renumbered to 1, 2, ⁇ , n according to the priority of the user to which the user is to be applied.
- step S 1 the search music designation unit 21 outputs the music ID and the metadata of the original music designated by the user to the cluster mapping unit 22.
- the cluster mapping unit 22 maps the metadata of the input original music to the optimum cluster using the conventional cluster identification method, and supplies the result (hereinafter referred to as optimum cluster information) to the music extraction unit 23. .
- step S 2 the music extracting unit 23 refers to the cluster information database 14 and assumes a set C having elements of the music IDs of all the music pieces for which the cluster information is held in the cluster information database 14.
- step S3 the music extraction unit 23 initializes the layer number i to one.
- step S4 the music extracting unit 23 determines whether the layer number W (n is the total number of cluster layers) or less. If it is determined that the layer number is equal to or less than three, the process proceeds to step S5. In step S5, based on the optimum cluster information of the original music input from the cluster mapping unit 22, the music extraction unit 23 identifies which cluster the original music belongs to in the i-th layer. The identified cluster is called CLix.
- step S 6 the music extraction unit 23 refers to the cluster 1 music I ⁇ f blue report of the cluster information database 14 and acquires the music ID of the music belonging to the specified cluster CLix.
- step S7 the music extraction unit 23 assumes a set A having the music ID acquired in the process of step S6 as an element.
- step S8 the music extraction unit 23 extracts an element (music ID) common to the set C and the set A, and in step S9, the power / power with which the common music ID exists (ie, in step S8). Then, determine whether it has been possible to extract a song ID common to Set C and Set A).
- step S10 If it is determined that set C and set A have a common song ID, the process proceeds to step S10, and the elements of set C are reduced to only the common song IDs extracted in step S8.
- step S11 the music extraction unit 23 increments the layer number i by one, returns to step S4, and repeats the subsequent processing.
- step S10 If it is determined in step S9 that there is no music ID common to the sets C and A, step S10 is skipped, and the process proceeds to step S11.
- the elements (music ID) of the set C are reduced by repeating the processes of steps S4 to Sl l.
- step S4 When it is determined in step S4 that the number is larger than the layer number and is not n or less, the process proceeds to step S12.
- step S 12 the music extraction unit 23 outputs the elements of the set C (music ID) to the random selection unit 26.
- the random selection unit 26 randomly selects one piece of music of the set C and outputs it to the selection reason generation unit 28. Note that the random selection unit 26 may output the elements (music ID) of the set C to the similarity calculation unit 27 which is not included, and the similarity calculation unit 27 may select one music.
- step S13 the selection reason generation unit 28 generates a selection reason sentence indicating the reason why the music selected by the random selection unit 26 (or the similarity calculation unit 27) is selected, and the selected music Output to the music presentation unit 29 together with the music ID of the
- step S14 the music presentation unit 29 presents the user with the music of the music ID input from the selection reason generation unit 28 and the selection reason sentence.
- FIG. 12 is a flowchart for explaining the second similar music search process.
- the preference information database 24 calculates the variance of the preference value in each cluster layer, and the variance of the preference value is the smallest (that is, the user preference is for a particular cluster). It is assumed that cluster layers are detected in a concentrated manner, and clusters are identified with concentrated preferences.
- the layer number of the cluster layer is P, and the cluster is CLpp.
- step S 31 preference information database 24 calculates the variance of preference values in each cluster layer, and the variance of preference values is the smallest (that is, the user preference is set to a specific cluster).
- Cluster layer is detected and designated as the P layer is an integer from 1 to n). Furthermore, the preference is concentrated in the layer P, and the cluster is identified as CLpp.
- step S 32 the search music designation unit 21 outputs the music ID and metadata of the original music designated by the user to the cluster mapping unit 22.
- the cluster mapping unit 22 optimizes the input metadata of the original music using the conventional cluster identification method. , And this optimal cluster information is generated and supplied to the music extraction unit 23.
- step S 33 the music extracting unit 23 refers to the cluster information database 14 and assumes a set C having elements of the music IDs of all the music pieces for which cluster information is held in the cluster information database 14.
- step S34 the music extraction unit 23 initializes the layer number i to one.
- step S35 the music extracting unit 23 determines whether the layer number i is less than or equal to n (n is the total number of cluster layers). If it is determined that the layer number is equal to or less than three, the process proceeds to step S36.
- step S36 the music extraction unit 23 determines whether or not P and the layer number i match in step S31, and if it is determined that they match, the process proceeds to step S37, and the processing target in the next step 39 is Identify cluster CLpp.
- step S36 determines whether P and the layer number i identified in step S31 do not match. If it is determined in step S36 that P and the layer number i identified in step S31 do not match, the process proceeds to step S38.
- step S 38 based on the optimum cluster information of the original music input from the cluster mapping unit 22, the music extraction unit 23 identifies to which cluster the original music belongs to the i-th layer. The identified cluster is called CLix.
- step S 39 the music extracting unit 23 refers to the cluster of the cluster information database 14-music HD information, and the cluster CLpp specified in the process of step S 37 or the music belonging to the cluster CLix specified in the process of step S 38 Get the song ID of
- step S40 the music extraction unit 23 assumes a set A having the music IDs acquired in the process of step S39 as elements.
- step S41 the music extraction unit 23 extracts an element (music ID) common to the set and the set A, and in step S42, whether or not there is a common music ID (ie, in the process of step S41) Determine whether a song ID common to Set C and Set A can be extracted. If it is determined that set C and set A have a common song ID, the process proceeds to step S43, and the elements of set C are reduced to the common song ID extracted in step S41.
- step S44 the music extraction unit 23 increments the layer number i by one, returns to step S35, and repeats the subsequent processing.
- step S43 is skipped, and the process proceeds to the step S44.
- the elements (music ID) of the set C are reduced.
- step S45 the process proceeds to step S45.
- step S 45 the music extraction unit 23 outputs the elements of the set C (music ID) to the random selection unit 26.
- the random selection unit 26 randomly selects one piece of music of the set C and outputs it to the selection reason generation unit 28. Note that the random selection unit 26 may output the elements (music ID) of the set C to the similarity calculation unit 27 which is not included, and the similarity calculation unit 27 may select one music.
- step S46 the selection reason generation unit 28 generates a selection reason sentence indicating the reason why the music selected by the random selection unit 26 (or the similarity calculation unit 27) is selected, and selects the selected music Output to the music presentation unit 29 together with the music ID of the
- step S47 the music presentation unit 29 presents the user with the music of the music ID input from the selection reason generation unit 28 and the selection reason sentence.
- the second similar music search processing described above it is not necessary to calculate the distance between the feature vector of the original music and the feature vector of the other music, and the cluster having a high preference value indicating the user's preference It is possible to present a song that belongs to and similar to the original song.
- FIG. 13 is a flowchart for explaining the third similar music search process.
- step S 61 the search music designation unit 21 outputs the music ID of the original music designated by the user and the metadata to the cluster mapping unit 22.
- the cluster mapping unit 22 maps the metadata of the input original music to the optimum cluster using the conventional cluster identification method, and supplies this optimum cluster information to the music extraction unit 23.
- step S62 the music piece extraction unit 23 initializes a group C having the music piece ID to which the evaluation value is added as an element. That is, at this point, the set C is an empty set.
- step S63 the music extraction unit 23 initializes the layer number i to one.
- step S 64 the music extracting unit 23 determines whether the layer number i is equal to or less than n (n is the total number of cluster layers). If it is determined that the layer number is equal to or less than three, the process proceeds to step S65.
- step S 65 the music extraction unit 23 receives the request from the cluster mapping unit 22. Based on the input optimum cluster information of the original music, it is specified to which cluster the original music belongs in the i-th layer. The identified cluster is called CLix.
- step S 66 the music piece extraction unit 23 refers to the preference information database 24 to acquire the preference value of the user for the cluster CLix identified in the process of step S 65, and based on the acquired preference value! / Decide the evaluation value to be given to the music belonging to cluster CLix.
- step S 67 the music extraction unit 23 refers to the cluster-music HD information of the cluster information database 14 and acquires the music ID of the music belonging to the specified cluster CLix.
- step S68 the music extraction unit 23 assigns the evaluation value determined in the process of step S66 to the music ID acquired in the process of step S67. Then, assume a set A having a music piece ID with an evaluation value as an element.
- step S69 the music extraction unit 23 adds the elements of the set A (song ID with evaluation value) to the set C.
- the music extracting unit 23 increments the layer number i by one, returns to step S64, and repeats the subsequent processing.
- step S64 By repeating the processes of steps S64 to S70, the elements of the set C (music IDs with evaluation values) increase.
- step S64 determines that the layer number is larger than n and it is determined that the number is not less than n.
- step S 71 the music extraction unit 23 selects the one with the highest evaluation value from the elements of the set C (music ID with evaluation value), and selects the random selection unit 26 (or the similarity calculation unit 2). 7) Output to the selection reason generation unit 28 via
- step S 72 the selection reason generation unit 28 generates a selection reason sentence indicating the reason why the music selected by the music extraction unit 23 is selected, and the music ID of the selected music and the selected music. Output to the presentation unit 29.
- step S 73 the music presentation unit 29 presents the user with the music of the music ID input from the selection reason generation unit 28 and the selection reason sentence.
- music is used as a search condition.
- the similar music search process can be performed by specifying an artist, an album, etc.
- the song ID in Figures 3 and 4 can be replaced with the artist ID or album ID.
- cluster layers corresponding to titles, albums, genres, etc. related to the artist in FIG. 2 are used.
- FIG. 14 is a flowchart illustrating the first music recommendation process.
- the cluster information database 14 sequentially receives layers from the higher priority cluster layer according to the user's priority for each cluster layer input from the preference input unit 25. It is assumed that the numbers have been rearranged to 1, 2, ⁇ , n.
- step S 91 the music extracting unit 23 refers to the cluster information database 14 and assumes a set C having the music IDs of all the music pieces whose cluster information is held in the cluster information database 14 as an element.
- step S92 the music extracting unit 23 initializes the layer number i to one.
- step S 93 the music extraction unit 23 determines whether the layer number is equal to or less than W3 (n is the total number of cluster layers). If it is determined that the layer number is equal to or less than three, the process proceeds to step S94.
- step S 94 the music piece extraction unit 23 refers to the preference information database 24 and specifies, among the clusters in the i-th layer, a cluster having the largest user preference value. The identified cluster is called CLix.
- step S 95 the music extraction unit 23 refers to the cluster-music HD information of the cluster information database 14, and acquires the music IDs of the music belonging to the specified cluster CLix.
- step S96 the music extraction unit 23 assumes a set A having the music IDs acquired in the process of step S95 as an element.
- step S97 the music extraction unit 23 extracts an element (music ID) common to the set C and the collection A, and in step S98, whether or not there is a common music ID (ie, in step S97) Then, determine whether it has been possible to extract a song ID common to the sets C and A).
- step S99 If it is determined that set C and set A have a common song ID, the process proceeds to step S99, and the elements of set C are reduced to only the common song IDs extracted in step S97.
- step S100 the music extraction unit 23 increments the layer number i by 1 and returns to step S93 to repeat the subsequent processing.
- step S99 is skipped, and the process proceeds to step S100.
- step S93 By repeating the processing of steps S93 to S100, the elements of the set C (music ID) are reduced. Then, if it is determined in step S93 that the number is larger than the layer number and not less than n, the process proceeds to step S101.
- step S101 the music extraction unit 23 outputs the elements of the set C (music ID) to the random selection unit 26.
- the random selection unit 26 randomly selects one piece of music of the set C and outputs it to the selection reason generation unit 28.
- the random selection unit 26 may output the elements (music ID) of the set C to the similarity calculation unit 27 which is similar to the similarity calculation unit 27 and one similarity may be selected by the similarity calculation unit 27.
- step S102 the selection reason generation unit 28 generates a selection reason sentence indicating the reason why the music selected by the random selection unit 26 (or the similarity calculation unit 27) is selected, and the selected music is selected.
- Output to the music presentation unit 29 together with the music ID of the the music presentation unit 29 presents the user with the music of the music ID input from the selection reason generation unit 28 and the selection reason.
- the first music recommendation process described above it is not necessary to calculate the distance between the feature vector corresponding to the user's preference and the music feature vector, and the priority of the user for each cluster layer is determined. It is possible to recommend to the user a song that matches the user's preferences while taking into consideration.
- FIG. 15 is a flowchart for explaining the second music recommendation process.
- step S121 the music piece extraction unit 23 initializes a set C assuming the music piece ID to which the evaluation value is attached as an element. That is, at this point, the set C is an empty set.
- step S122 the music extraction unit 23 initializes the layer number i to one.
- step S123 the music extracting unit 23 determines whether the layer number i is equal to or less than n (n is the total number of cluster layers). If it is determined that the layer number is 3 ⁇ 4 or less, the process proceeds to step S124.
- step S 124 the music extraction unit 23 refers to the preference information database 24 and identifies clusters in the i-th layer whose preference value corresponding to the preference of the user is equal to or greater than a predetermined value. The specified one is called a cluster group CLix.
- step S 125 the music piece extraction unit 23 applies the music piece belonging to each cluster in the cluster group CLix based on the preference value for each cluster in the cluster group CLix specified in the process in step S 124. Determine the evaluation value to be given.
- step S 126 the music extraction unit 23 refers to the cluster music HD information of the cluster information database 14 and acquires the music IDs of the music belonging to each cluster of the specified cluster group CLix.
- step S127 the music extraction unit 23 assigns the evaluation value determined in the process of step S125 to the music ID acquired in the process of step S126. Then, assume a set A having elements of music IDs with evaluation values.
- step S1208 the music extraction unit 23 adds the elements of the set A (song ID with evaluation value) to the set C. At this time, if the set C has the same music ID, the evaluation values are added.
- step S129 the music extraction unit 23 increments the layer number i by 1 and returns to step S123 to repeat the subsequent processing.
- step S123 By repeating the processes of steps S123 to S129, the elements of the set C (music IDs with evaluation values) increase.
- step S123 determines that the layer number i is greater than n and is not less than or equal to n.
- step S130 the music extraction unit 23 selects the one with the highest evaluation value from the elements of the set C (song ID with evaluation value), and selects the random selection unit 26 (or the similarity calculation unit 2). 7) Output to the selection reason generation unit 28 via
- step S131 the selection reason generation unit 28 generates a selection reason sentence indicating the reason why the music selected by the music extraction unit 23 is selected, and the music presentation unit 29 together with the music ID of the selected music.
- step S132 the music presentation unit 29 presents the user with the music of the music ID input from the selection reason generation unit 28 and the selection reason sentence to the user.
- calculation of the distance between the feature vector corresponding to the user's preference and the feature vector of the music is not essential, and it is given according to the user's preference. It is possible to recommend the user the music with the highest evaluation value.
- the feature vector corresponding to the original music or the preference of the user, and the search object It is possible to select a song to be presented before calculating the distance (such as cosine correlation) with the feature vector of the song.
- the preference of the user can be prioritized, so that the degree of satisfaction of the user with respect to the search or recommendation can be improved.
- cluster layers may be divided into groups and partially used. For example, ⁇ related artist layer, artist genre layer, artist review text layer ⁇ as a group for artist search recommendation, and ⁇ music feature amount layer (tempo, rhythm, etc), song genre layer, song review text layer ⁇ Let's define it as a search recommendation group.
- the above-described series of processes can also be executed by software that can be executed by hardware.
- a series of processes are executed by software, it is possible to execute various functions by installing a computer built in the hardware dedicated to the program power constituting the software or various programs. For example, it is installed from a recording medium on a general-purpose personal computer configured as shown in FIG.
- the personal computer 100 incorporates a CPU (Central Processing Unit) 101.
- An input / output interface 105 is connected to the CPU 101 via a bus 104.
- the ROM 104 is connected to a read only memory (ROM) 102 and a random access memory (RAM) 103.
- the input / output interface 105 includes an input unit 106 including an input device such as a keyboard and a mouse through which a user inputs an operation command, a CRT (Cathode Ray Tube) or an LCD (Liquid Crystal Display) for displaying a screen. Etc., and a storage unit 108 such as a hard disk drive for storing programs and various data, a modem, a LAN (Local Area Network) adapter, etc., and communication via a network represented by the Internet.
- a communication unit 109 that executes processing is connected.
- Optical disk including lexible disk
- optical disk including compact disc-read only memory (CD-ROM), digital versatile disc (DVD)
- magneto-optical disc including mini disk (MD)
- semiconductor memory etc.
- a drive 110 for reading and writing data to the recording medium 111 is connected.
- a program for causing the personal computer 100 to execute the series of processes described above is supplied to the personal computer 100 in a state of being stored in the recording medium 111, read by the drive 110, and incorporated in the storage unit 108. Installed on the drive.
- the program installed in the storage unit 108 is loaded from the storage unit 108 to the RAM 103 and executed according to a command from the CPU 101 corresponding to a command from the user input to the input unit 106.
- FIG. 17 is a block diagram showing an example of another configuration of the recommendation system 1 according to the embodiment of this invention.
- the same parts as those in FIG. 1 are denoted by the same reference numerals, and the description thereof will be omitted.
- the recommendation system 1 shown in FIG. 17 includes a music DB 11, a keyword setting unit 13, a cluster information DB 14, a search music specification unit 21, a cluster mapping unit 22, a music extraction unit 23, a preference information database 24, and a preference input unit.
- the metadata clustering unit 201 clusters metadata of each music recorded in the music database 11. That is, the metadata clustering unit 201 classifies the metadata of music as content into any one of a plurality of clusters, and assigns a hierarchy to the clusters.
- the metadata clustering unit 201 supplies the result of clustering of metadata of each music to the music clustering unit 202.
- the music clustering unit 202 clusters each music and generates cluster information of each music based on the clustering result of the metadata of each music by the metadata clustering unit 201. Do. In other words, the music clustering unit 202 generates cluster information corresponding to the result of clustering of each music and transmits the generated cluster information to the cluster information DB 14. Output.
- the metadata clustering unit 201 acquires metadata of a music from the music DB 11, and compresses the dimension of the acquired metadata.
- the metadata clustering unit 201 is the next to the metadata of the music obtained from the music DB 11, LbA ⁇ the latent semantic analysis), PLbA ⁇ the probabilistic latent semantic analysis), or quantification Compress by a method such as Class III.
- step S 201 the metadata clustering unit 201 may process the metadata of music piece as vector.
- step S202 the metadata clustering unit 201 clusters metadata of each music.
- the metadata clustering unit 201 performs soft clustering on metadata of each music.
- the metadata clustering unit 201 is configured such that, in each layer, the sum of the belonging weight to each cluster of the item is 1, Perform soft clustering of music metadata.
- the first cluster, the second cluster, the third cluster, and the fourth cluster in the first layer (layer number 1) of the metadata of the music identified by the music ID that is ABC123 The generic weights are 0. 0, 0. 8, 0. 0, and 0. 2 respectively.
- the belonging weights to the fifth cluster, the sixth cluster, the seventh cluster, and the eighth cluster in the second layer (layer number 2) of the metadata of the music identified by the music ID that is ABC123 are respectively , 0.4, 0.6, 0. 0, and 0. 0. 0.
- the belonging weights to the ninth cluster, the tenth cluster, and the eleventh cluster in the third layer (layer number 3) of the metadata of the music identified by the music ID that is ABC123 are each 0. 0, 0. 0, and 1. 0.
- the belonging weights to each of the four clusters in the nth layer (layer number n) of the metadata of the music identified by the music ID of ABC123 are 1.0, 0.0, and 0, respectively. , And 0. 0. 0.
- the belonging weights to the first cluster, the second cluster, the third cluster, and the fourth cluster in the first layer of the metadata of the music identified by the music ID that is CTH 863 are , 1. 0, 0. 0, 0. 0, and 0. 0, respectively.
- Attribution weights to the fifth cluster, the sixth cluster, the seventh cluster, and the eighth cluster in the second hierarchy of the metadata of the song identified by the song ID that is CTH 863 are respectively 0, 0, 0, 0.5, 0.5, and 0. 0. 0.
- the belonging weights to the ninth cluster, the tenth cluster, and the eleventh cluster in the third layer of the metadata of the music identified by the music ID, which is CTH 863 are 0.7, 0.3, and 0.3, respectively. 0. 0
- the belonging weights to each of the four clusters in the nth layer of the metadata of the music identified by the music ID which is CTH 863 are respectively 0. 0, 0. 8, 0. 2, and 0. 0. is there.
- the belonging weights to the first cluster, the second cluster, the third cluster, and the fourth cluster in the first layer of the metadata of the music identified by the music ID which is XYZ567 are respectively , 0, 0, 0, 4, 0, 6, and 0. 0.
- the attribution weight to the fifth cluster, the sixth cluster, the seventh cluster, and the eighth cluster in the second layer of the metadata of the music identified by the music ID which is XYZ567 is 0. 0, respectively. 0. 0, 0. 0, and 1.0.
- the belonging weights to the ninth cluster, the tenth cluster, and the eleventh cluster in the third layer of the metadata of the music identified by the music ID that is XYZ567 are 0.9, 0.0, and 0.9, respectively. It is 0.1.
- the belonging weights to each of the four clusters in the nth layer of the metadata of the music identified by the music ID that is XYZ567 are 0.3, 0.0, 0. 0, and 0.7, respectively. is there.
- the soft clustering of the metadata of each music is not limited to an item, that is, the sum of the belonging weights of the music to each cluster is 1, in each layer. Also, each item may not belong to any cluster in each layer.
- step S203 the metadata clustering unit 201 allocates layers of clusters.
- FIG. 20 is a diagram showing an example of metadata.
- the metadata shown in FIG. 20 is categorical data of either 0 or 1 for simplicity.
- Meta group 1 as a high-order classification includes Metadata 1, Metadata 2, and Metadata.
- the third group belongs, and the metagroup 2 as a high-order classification includes metadata 4, metadata 5, and metadata 6.
- meta group 1 includes metadata about an artist
- metadata 1 indicates the appearance of the artist
- metadata 2 indicates that it is a group.
- meta-group 2 has metadata relating to the genre
- metadata 4 indicates that it is a pop
- metadata 5 indicates that it is a lock.
- Metadata 1 to metadata 6 are 1, 0, 1, 1, 0, 0, respectively.
- metadata 1 regarding the music identified by the music ID identified by ABC123 or the music identified by the song ID identified by OPQ 385 is regarded as a vector.
- each of the metadata 2 to the metadata 6 is regarded as a vector for the music identified by the music ID that is ABC123 or the music identified by the music ID that is OPQ 385. That is, the value of one metadata for a plurality of songs is regarded as a vector.
- metadata 1 considered as a vector
- metadata 3 and metadata 4 represent clusters within the Manhattan distance 1
- metadata 2 represent clusters within the Manhattan distance 2
- metadata 5 and metadata 6 are clinging to other clusters within 1 Manhattan distance.
- these clusters are set as a new metadata hierarchy. That is, metadata is allocated closer to each layer in the hierarchy.
- FIG. 21 shows an example of metadata that is thus clustered and layer allocated.
- metadata 1, metadata 3 and metadata 4 belong to the first layer
- metadata 2 belong to the second layer
- each layer is formed of a collection of highly correlated metadata, and music clustering is performed in that layer, genres and artists are made hierarchical as they are. It is possible to reflect in the cluster the differences between subtle pieces of music that can not be expressed by ordinary hierarchy division.
- step S 204 the music clustering unit 202 clusters the music for each layer, and the process ends. That is, the music clustering unit 202 classifies each content into! /! Of a plurality of clusters in each of the assigned layers.
- music can be clustered so that subtle differences between music can be well expressed.
- step S221 the search music designation unit 21 sets an original music as a similarity source. That is, for example, in step S221, the search music specification unit 21 outputs the music ID of the original music to the music extraction unit 23 according to the user specification via the cluster mapping unit 22. Set the music.
- step S222 the similarity calculation unit 27 calculates the similarity between the original music and each of all the music other than the original music from the belonging weight of each cluster.
- the music extraction unit 23 also reads the cluster information database 14 power of the cluster information of the original music identified by the music ID and the cluster information of all the music other than the original music. Then, the music extraction unit 23 supplies the read cluster information to the similarity calculation unit 27.
- the similarity calculation unit 27 calculates the similarity between the original music and all the music other than the original music from the belonging weight of each cluster indicated by the cluster information of the original music and all the music other than the original music. calculate.
- the music clustering unit 202 soft-clusters each music within each layer, and stores it in the cluster information force S cluster information database 14 indicating the belonging weight of each cluster. .
- FIG. 23 is a diagram illustrating an example of cluster information indicating belonging weights of clusters.
- a cluster in the first hierarchy a cluster identified by a cluster ID of CL11, a cluster identified by a cluster ID of CL12, a cluster identified by a cluster ID of CL13, and CL14
- the belonging weights of the music identified by the music ID, which is ABC123, to the cluster identified by the cluster ID are 0. 0, 1.0, 0. 0, and 0. 2, respectively.
- a cluster in the second hierarchy a cluster identified by a cluster ID CL21, a cluster identified by a cluster ID CL22, a cluster identified by a cluster ID CL23, and a cluster ID CL24
- the belonging weights of the music identified by the music ID, which is ABC123, to the cluster to be cluster are 0.6, 0.8, 0.0, and 0. 0 respectively.
- the attribution weights of the songs identified by the song ID are 0, 0, 0, 0, and 1.0, respectively.
- a cluster in the fourth hierarchy a cluster identified by a cluster ID of CL 41, a cluster identified by a cluster ID of CL 42, a cluster identified by a cluster ID of CL 43, and a cluster ID of CL 44
- the belonging weights of the music identified by the music ID of ABC123 to the identified cluster are: 1. 0, 0. 0, 0. 0, and 0. 0, respectively.
- a cluster in the first hierarchy a cluster identified by a cluster ID that is CL11, a cluster identified by a cluster ID that is CL12, a cluster identified by a cluster ID that is CL13, and CL14
- the belonging weights of the song identified by the song ID, which is CTH 863, to the cluster identified by the cluster ID are 1. 0, 0. 0, 0. 0, and 0. 0, respectively.
- a cluster in the second hierarchy a cluster identified by a cluster ID CL21, a cluster identified by a cluster ID CL22, a cluster identified by a cluster ID CL23, and a cluster ID CL24
- the belonging weights of the music identified by the music ID, which is CTH 863, to the cluster to be processed are 0. 0, 0. 7, 0. 7, and 0. 0, respectively.
- a cluster in the third hierarchy a cluster identified by a cluster ID of CL31, a cluster identified by a cluster ID of CL32, and a cluster ID of CL33
- the belonging weights of the music identified by the music ID, which is CTH 863, to the cluster being processed are 0.9, 0.4, and 0. 0, respectively.
- a cluster identified by a cluster ID of CL41 which is a cluster in the fourth hierarchy, a cluster identified by a cluster ID of CL42, a cluster identified by a cluster ID of CL43, and a CL44
- the belonging weights of the song identified by the song ID CTH 863 to the cluster identified by a cluster ID are 0. 0, 1. 0, 0. 0.3, and 0. 0, respectively.
- a cluster identified by a cluster ID CL11 which is a cluster in the first hierarchy
- a cluster identified by a cluster ID CL12 a cluster identified by a cluster ID CL13
- CL14 The belonging weights of the music identified by the music ID, which is XYZ567, to the cluster identified by the cluster ID are 0. 0, 0. 6, 0. 8, and 0. 0, respectively.
- a cluster in the second hierarchy a cluster identified by a cluster ID CL21, a cluster identified by a cluster ID CL22, a cluster identified by a cluster ID CL23, and a cluster ID CL24
- the attribute weights of the music identified by the music ID, which is XYZ 567, to the cluster being processed are 0. 0, 0. 0, 0. 0, and 1.0, respectively.
- a cluster in the third hierarchy that is, a cluster identified by a cluster ID that is CL31, a cluster that is identified by a cluster ID that is CL32, and a cluster that is identified by a cluster ID that is CL33
- the attribution weights of the songs identified by the song ID are: 1. 0, 0. 0, and 0. 1, respectively.
- a cluster identified by a cluster ID of CL41 which is a cluster in the fourth hierarchy, a cluster identified by a cluster ID of CL42, a cluster identified by a cluster ID of CL43, and a CL44
- the belonging weights of the music identified by music ID, which is XYZ567, to the cluster identified by a cluster ID are 0.5, 0, 0, 0, 0, and 0, 9, respectively.
- a cluster identified by a cluster ID CL11 which is a cluster in the first hierarchy
- a cluster identified by a cluster ID CL12 a cluster identified by a cluster ID CL13
- a CL14 The attribution weights of the music identified by the music ID, which is the EKF 534, to the cluster identified by the cluster ID are 0.9, 0.0, 0, 0, and 0.5, respectively.
- a class identified by a cluster ID which is a cluster in the second hierarchy, that is CL21 To the cluster identified by the cluster ID that is CL22, the cluster identified by the cluster ID that is CL23, and the cluster identified by the cluster ID that is CL24 attributable to the song that is identified by the song ID that is EKF 534
- the weights are 0. 0, 0. 6, 0. 0, and 0. 8 respectively.
- a cluster identified by a cluster ID of CL31, a cluster identified by a cluster ID of CL32, and a cluster identified by a cluster ID of CL33 The attribution weights of the songs identified by the song ID are: 0.7, 0. 0, and 0. 7, respectively.
- a cluster identified by a cluster ID of CL41 which is a cluster in the fourth hierarchy, a cluster identified by a cluster ID of CL42, a cluster identified by a cluster ID of CL43, and a CL44
- the attribution weights of the songs identified by the song ID EKF 534 to the cluster identified by a cluster ID are 0.0, 0.9, 0.4, and 0.3, respectively.
- a cluster identified by a cluster ID CL11 which is a cluster in the first hierarchy
- a cluster identified by a cluster ID CL12 a cluster identified by a cluster ID CL13
- a CL14 The attribution weights of the song identified by the song ID that is OPQ 385 to the cluster identified by the cluster ID are 0.7, 0.2, 0.6, and 0.0, respectively.
- a cluster in the second hierarchy a cluster identified by a cluster ID CL21, a cluster identified by a cluster ID CL22, a cluster identified by a cluster ID CL23, and a cluster ID CL24
- the attribute weights of the song identified by the song ID, which is OPQ 385, to the cluster being clustered are 1. 0, 0. 0, 0. 0, and 0. 0, respectively.
- OPQ 385 to clusters in the third hierarchy a cluster identified by a cluster ID of CL31, a cluster identified by a cluster ID of CL32, and a cluster identified by a cluster ID of CL33.
- the attribution weights of the songs identified by the song ID are 0, 0, 1.0, and 0. 0 respectively.
- a cluster identified by a cluster ID of CL41 which is a cluster in the fourth hierarchy, a cluster identified by a cluster ID of CL42, a cluster identified by a cluster ID of CL43, and a CL44
- the belonging weights of the songs identified by the song ID of OPQ 385 to the cluster identified by a cluster ID are , 0.4, 0.9, 0.0, and 0. 0. 0.
- the similarity calculation unit 27 determines the similarity sim (i, j) between the original music identified by the music ID of i and the music identified by the music ID of j It is calculated by the operation shown in equation (1) from the belonging weight to the cluster.
- sim (i, j) ⁇ w ilc w ilc ... (1)
- L is a value indicating the number of layers, and 1 is a value identifying the layers.
- C (l) indicates the entire cluster, and c is a value that identifies the cluster.
- w is the song ID which is i
- w indicates the belonging weight of the first cluster c in the first layer of the song identified by the song ID that is j.
- FIG. 24 is a diagram showing an example of the degree of similarity calculated by the calculation shown in equation (1) from the cluster information of FIG. 23 showing the belonging weights of clusters.
- the similarity of each of the songs identified by the song IDs CTH 863 to OPQ 385 to the original song identified by the song ID ABC123 is shown.
- the music identified by the music IDs of CTH 863 to OPQ 385 is The similarity of each of the songs identified by the song ID of each of CTH 863 to OPQ 385 as follows: 0.57, 1.18, 1.27, 1 It will be 20.
- step S222 similarity calculation section 27 calculates 0.535, 1.18, 1.27, and 1.20 according to the calculation shown in equation (1), respectively, using ABC123.
- the similarity of each of the songs identified by the song ID of CTH 863 to OPQ 385 to the original song identified by the song ID is calculated.
- step S223 the similarity calculation unit 27 sorts all the songs other than the original music in the order of similarity to the original music based on the similarity.
- the similarity calculation unit 27 associates the music similarity obtained as a result of the calculation with the music ID of the music, and based on the similarity, it is similar to the original music. Songs in order of music By rearranging D, all songs other than the original music are sorted in order of similarity to the original music.
- step S224 the similarity calculating unit 27 selects an arbitrary number of higher rank songs from the sorted songs.
- the similarity calculation unit 27 supplies the music selection ID of the selected music to the selection reason generation unit 28.
- step S224 the similarity calculating unit 27 selects the top music and supplies the music ID of the top music to the selection reason generating unit 28.
- the similarity calculation unit 27 selects the top 10 songs, and supplies the selection ID of the top 10 songs to the selection reason generation unit 28.
- step S 225 selection reason generation unit 28 generates a selection reason sentence indicating the reason why the music piece selected by similarity calculation unit 27 is selected, and presents the music together with the music ID of the selected music. Output to section 29.
- the music presentation unit 29 presents the music of the music ID input from the selection reason generation unit 28 and the selection reason sentence to the user, and the process ends.
- Steps S241 to S 251 are the same as steps S 1 to S 11 in FIG. 11, and thus the description thereof will be omitted.
- step S 252 similarity calculation unit 27 determines, from the attribution weight of each cluster, the original music and set C, based on the element of collection C (music ID) supplied from music extraction unit 23. Calculate the degree of similarity with each of the songs. For example, in step S252, the similarity calculation unit 27 calculates the similarity between each of the original music and the music of the set C by the calculation shown in equation (1).
- step S 253 the similarity calculating unit 27 sorts the music of the set C in the order of similarity to the original music based on the similarity.
- the similarity calculating unit 27 associates the degree of similarity obtained as a result of the calculation with the music ID of the music of the set C, and based on the similarity, the music of the set C. Sort the songs in Set C by sorting the song IDs in order of similarity to the original song.
- step S224 similarity calculating section 27 determines any number of the sorted music pieces. Select the top songs of the. The similarity calculation unit 27 supplies the music selection ID of the selected music to the selection reason generation unit 28.
- step S224 the similarity calculation unit 27 selects the top music and supplies the music ID of the top music to the selection reason generation unit 28.
- the similarity calculation unit 27 selects the top 10 songs, and supplies the selection ID of the top 10 songs to the selection reason generation unit 28.
- step S 225 selection reason generation unit 28 generates a selection reason sentence indicating the reason why the music selected by similarity calculation unit 27 is selected, and presents the music together with the music ID of the selected music. Output to section 29.
- the music presentation unit 29 presents the music of the music ID input from the selection reason generation unit 28 and the selection reason sentence to the user, and the process ends.
- step S 271 the similarity calculating unit 27 sets the preference value of the user indicating the belonging weight of each cluster based on the element of the group C (music ID) supplied from the music extracting unit 23, and sets The degree of similarity with the cluster information indicating the belonging weight of each cluster, which is the cluster information of each song of C, is calculated.
- the preference information database 24 is soft-clustered, and a preference value indicating attribution weight of each cluster is recorded in each layer.
- FIG. 27 is a diagram showing an example of preference values indicating attribution weights of clusters.
- a cluster identified by a cluster ID CL11 which is a cluster in the first hierarchy
- a cluster identified by a cluster ID CL12 a cluster identified by a cluster ID CL13
- a CL14 The belonging weights of the preference value of the user identified by the user ID, which is U001, to the cluster identified by the cluster ID are respectively: 0. 0, 0. 8, 0. 0, and 0. 6 It is.
- the cluster in the second hierarchy identified by the cluster ID, which is CL21 To the cluster identified by the cluster ID that is CL22, the cluster identified by the cluster ID that is CL23, and the cluster identified by the cluster ID that is CL24 by the user identified by the user ID that is U001.
- the membership weights of preference values are 0.4, 0.6, 0.7, and 0.0, respectively.
- a cluster identified by the cluster ID CL31, a cluster identified by the cluster ID CL32, and a cluster identified by the cluster ID CL33, U001 The belonging weights of the preference value of the user specified by the user ID that is are 0.7, 0.5, and 0.5, respectively.
- the attribution weights of the preference value of the user identified by the user ID, which is U001, to the cluster identified by are respectively: 0. 0, 0. 5, 0. 4 and 0. 0. 0.
- FIG. 28 is a diagram showing an example of cluster information indicating attribution weight of each cluster.
- a cluster identified by a cluster ID CL11 which is a cluster in the first hierarchy
- a cluster identified by a cluster ID CL12 a cluster identified by a cluster ID CL13, and CL14
- the belonging weights of the music identified by the music ID, which is ABC123, to the cluster identified by the cluster ID are 0. 0, 1.0, 0. 0, and 0. 2, respectively.
- a cluster in the second hierarchy a cluster identified by a cluster ID CL21, a cluster identified by a cluster ID CL22, a cluster identified by a cluster ID CL23, and a cluster ID CL24
- the belonging weights of the music identified by the music ID, which is ABC123, to the cluster to be cluster are 0.6, 0.8, 0.0, and 0. 0 respectively.
- the attribution weights of the songs identified by the song ID are 0, 0, 0, 0, and 1.0, respectively.
- a cluster in the fourth hierarchy a cluster identified by a cluster ID of CL 41, a cluster identified by a cluster ID of CL 42, a cluster identified by a cluster ID of CL 43, and a cluster ID of CL 44 specific
- the attribution weights of the music identified by the music ID, which is ABC123, to the cluster to be cluster are 1. 0, 0. 0, 0. 0, and 0. 0, respectively.
- a cluster identified by a cluster ID CL11 which is a cluster in the first hierarchy
- a cluster identified by a cluster ID CL12 a cluster identified by a cluster ID CL13
- a CL14 The belonging weights of the song identified by the song ID, which is CTH 863, to the cluster identified by the cluster ID are 1. 0, 0. 0, 0. 0, and 0. 0, respectively.
- a cluster in the second hierarchy a cluster identified by a cluster ID CL21, a cluster identified by a cluster ID CL22, a cluster identified by a cluster ID CL23, and a cluster ID CL24
- the belonging weights of the music identified by the music ID, which is CTH 863, to the cluster to be processed are 0. 0, 0. 7, 0. 7, and 0. 0, respectively.
- CTH 863 to a cluster in the third hierarchy a cluster identified by a cluster ID of CL 31, a cluster identified by a cluster ID of CL 32, and a cluster identified by a cluster ID of CL 33.
- the attribution weights of the songs identified by the song ID are 0.9, 0.4, and 0. 0 respectively.
- a cluster identified by a cluster ID of CL41 which is a cluster in the fourth hierarchy, a cluster identified by a cluster ID of CL42, a cluster identified by a cluster ID of CL43, and a CL44
- the belonging weights of the song identified by the song ID CTH 863 to the cluster identified by a cluster ID are 0. 0, 1. 1, 0. 3, and 0. 0, respectively.
- a cluster in the first hierarchy a cluster identified by a cluster ID that is CL11, a cluster identified by a cluster ID that is CL12, a cluster identified by a cluster ID that is CL13, and CL14
- the belonging weights of the music identified by the music ID, which is XYZ567, to the cluster identified by the cluster ID are 0. 0, 0. 6, 0. 8, and 0. 0, respectively.
- a cluster in the second hierarchy a cluster identified by a cluster ID CL21, a cluster identified by a cluster ID CL22, a cluster identified by a cluster ID CL23, and a cluster ID CL24
- the attribute weights of the music identified by the music ID, which is XYZ 567, to the cluster being processed are 0. 0, 0. 0, 0. 0, and 1.0, respectively.
- the cluster identified by the cluster ID CL31, the cluster identified by the cluster ID CL32, and the cluster identified by the cluster ID CL33, XYZ567 The attribution weights of the songs identified by the song ID are: 1. 0, 0. 0, and 0. 1, respectively.
- a cluster identified by a cluster ID of CL41, which is a cluster in the fourth hierarchy a cluster identified by a cluster ID of CL42, a cluster identified by a cluster ID of CL43, and a CL44
- the belonging weights of the music identified by the music ID, which is XYZ567, to the cluster identified by a cluster ID are 0.4, 0, 0, 0, 0, and 0, 7, respectively.
- the similarity calculation unit 27 uses the belonging weight to the cluster in the preference value of the user, and the belonging weight to the cluster in the cluster information of the music identified by the music ID that is i, Calculate the similarity sim (u, i) by the operation shown in.
- L is a value indicating the number of layers, and 1 is a value identifying the layers.
- C (l) indicates the entire cluster, and c is a value that identifies the cluster.
- w is the song ID which is i
- h denotes the belonging weight of the 1st cluster of the 1st layer of the preference value of the user u.
- FIG. 29 is an example of the similarity calculated by the operation shown in equation (2) from the preference value indicating the belonging weight of the cluster in FIG. 27 and the cluster information indicating the belonging weight of the cluster in FIG. FIG.
- belonging weights of the preference value of the user specified by the user ID which is U001 among the belonging weights of the first layer and the belonging weight of the cluster information of the music specified by the music ID which is ABC123.
- belonging weight of the first layer are multiplied by the corresponding belonging weights, and when the multiplied results are integrated, the value arranged in the first layer for the music ID, which is ABC123 in FIG. Is required.
- the belonging weight of the preference value of the user specified by the user ID U001 and the cluster information of the music specified by the music ID ABC123 Attribution weight is multiplied by corresponding attribution weight and multiplied
- the results are accumulated, and the values assigned to the second, third, and fourth layers, respectively, for the song ID of ABC123 in FIG. 29 are 0.67, 0.53 and 0.00, respectively. Is required.
- the similarity between the preference value of the user identified by the user ID U001 and the cluster information of the song identified by the song ID ABC123 is 1st layer, 2nd layer, 2nd layer
- the third and fourth layers are considered to be the value 2.11 which is the value obtained by calculating 0.9.1, 0.67, 0.53 and 0.00, respectively, which were calculated by the following procedure.
- the value 0100 is a value arranged in the first layer for the music ID which is CTH 863 in FIG. Is required.
- each of the second, third, and fourth layers for the music ID which is CTH 863 in FIG.
- the assigned values of 0.92, 0.82, and 0.63 are obtained.
- the similarity between the preference value of the user specified by the user ID U001 and the cluster information of the music specified by the music ID CTH 863 is 1st layer, 2nd layer, 2nd layer
- the third and fourth layers are calculated as the value 2.37, which is the value obtained by calculating the values of 0.00, 0.92, 0.82, and 0.63 obtained by the following procedure.
- each of the second, third, and fourth layers for the music ID which is XYZ567 in FIG.
- the assigned values of 0.00, 0.72, and 0.00 are obtained.
- the similarity between the preference value of the user identified by the user ID U001 and the cluster information of the song identified by the music ID XYZ567 is 1st layer, 2nd layer, 2nd layer It is considered 1.15 that is the value obtained by calculating the values of 0.44, 0. 00, 0. 72, and 0. 00, which have been calculated by using the 3rd and 4th layers, respectively.
- the similarity may be calculated using the weight based on the distribution of the belonging weight of the user's preference value in each layer.
- the similarity calculation unit 27 uses the belonging weight to the cluster in the preference value of the user, and the belonging weight to the cluster in the cluster information of the music identified by the music ID that is i, Calculate the similarity sim (u, i) by the operation shown in.
- L is a value indicating the number of layers, and 1 is a value identifying the layers.
- C (l) indicates the entire cluster, and c is a value that identifies the cluster.
- w is the song ID which is i
- the weight of the first layer is shown.
- FIG. 30 is a diagram illustrating an example of weights for each hierarchy, which is the distribution of attribution weight of each hierarchy of the preference value of the user.
- the weight of the first layer, the weight of the second layer, the weight of the third layer, and the weight of the fourth layer for the user identified by the user ID which is U001 are respectively 0. 0. 17, 0. 0, 0. 01, and 0. 06.
- FIG. 31 is represented by Expression (3) from the preference value indicating the belonging weight of the cluster in FIG. 27, the cluster information indicating the belonging weight of the cluster in FIG. 28, and the weight for each hierarchy in FIG. It is a figure which shows the example of the similarity calculated by calculation.
- the degree of similarity shown in FIG. 30 is 10 times the result calculated by the calculation shown in equation (3).
- the cluster information of the music identified by the music ID which is ABC123, corresponding to the belonging weight of the first layer of the preference value of the user specified by the user ID that is U001 and the belonging weight of the user preference value.
- the attribution weight of the first layer is multiplied by the weight of the first layer, and the multiplied result is integrated. Then, a value of 1.27, which is the value allocated to the first layer for the music ID that is ABC123 in FIG. 31, is obtained.
- it is ABC123 corresponding to the belonging weight of the preference value of the user specified by the user ID of U001 and the belonging weight of the user preference value.
- the diagram is obtained.
- the value obtained by placing each of the second, third, and fourth layers with respect to the song ID, which is ABC123 of 31, is calculated as 0.49, 0. 03, 0, 00, respectively.
- the similarity between the preference value of the user identified by the user ID U001 and the cluster information of the song identified by the song ID ABC123 is 1st layer, 2nd layer, 2nd layer It is considered 1.79, which is the value obtained by calculating 1.27, 0. 49, 0. 03, and 0. 00, which were calculated by the following procedure: 3rd layer and 4th layer.
- Cluster information of the music identified by the music ID, CTH 863 corresponding to the belonging weight of the first layer of the preference value of the user specified by the user ID being U001 and the belonging weight of the preference value of the user
- the belonging weight of the first layer and the weight of the first layer are multiplied and the multiplied results are integrated, it is a value placed in the first layer for the music ID which is CTH 863 in FIG. 00 is required.
- CTH 863 corresponds to the belonging weight of the preference value of the user specified by the user ID of U001 and the belonging weight of the user preference value.
- the diagram is obtained.
- the result is 0.65, 0. 04, 0.27 power S, which is a value placed in each of the second, third, and fourth layers for the music ID which is CTH 863 of 31.
- the similarity between the preference value of the user specified by the user ID U001 and the cluster information of the music specified by the music ID CTH 863 is 1st layer, 2nd layer, 2nd layer
- the third and fourth layers, respectively, are calculated as the value 0.96, which is the value obtained by calculating the values of 00, 0.65, 0.04 and 0.27.
- the cluster information of the music identified by the music ID which is XYZ 567, corresponding to the belonging weight of the first layer of the preference value of the user specified by the user ID being U001 and the belonging weight of the user preference value.
- the attribution weight of the first layer is multiplied by the weight of the first layer, and the multiplied result is integrated. Then, the value 0.55 assigned to the first layer for the music ID, which is XYZ567 in FIG. 31, is obtained.
- it is XYZ567 corresponding to the belonging weight of the preference value of the user specified by the user ID of U001 and the belonging weight of the user preference value.
- the diagram is obtained. It is calculated by the values 0.000, 0. 04, 0, 00 which are values respectively assigned to the second layer, the third layer, and the fourth layer for the music ID which is XYZ567 of 31.
- the similarity between the preference value of the user specified by the user ID U001 and the cluster information of the music specified by the music ID XYZ567 is 1st layer, 2nd layer, 2nd layer
- the third and fourth layers, respectively, are calculated as the result of calculating 0.50, 0.50, 0. 04, and 0. 00, which are calculated by the following procedure.
- the value of the belonging weight of the preference value of the user identified by the user ID U001 is lower in the first layer than in the second to fourth layers. Since there is a large change, the value of each element of the first layer is predicted to be related to the preference of the user specified by the user ID of U001, as compared with the second to fourth layers.
- the similarity is more likely to change depending on the value that is predicted to be more related to the user's preference than to the value that is predicted to be less related to the user's preference. Can be detected, so that the user's favorite music can be detected more accurately.
- step S 272 the similarity calculating unit 27 sorts the music of the set C in the order of similarity to the preference of the user based on the similarity.
- the similarity calculation unit 27 associates the degree of similarity obtained as a result of the calculation with the music ID of the music of the set C, and based on the similarity, the music of the set C. By rearranging the music IDs of, the music of set C is sorted in order of similarity to the user's preference.
- step S273 the similarity calculating unit 27 selects an arbitrary number of higher rank songs from among the sorted songs.
- the similarity calculation unit 27 supplies the music selection ID of the selected music to the selection reason generation unit 28.
- the similarity is calculated by the operation shown in equation (2), and the music ID is ABC123
- the similarity for the identified song is 2.11
- the similarity for the song identified by the song ID CTH 863 is 2.37
- the similarity for the song identified by the song ID XYZ567 When the degree is set to 1.15, when one music is selected, the music specified by the music ID of CTH 863 having the highest similarity will be selected.
- the similarity is calculated using the weight based on the distribution of the a posteriori weight of the user's preference value in each layer by the calculation shown in equation (3), and the music ID is ABC123
- the similarity for the identified song is 1.79
- the similarity for the song identified by the song ID CTH 863 is 0. 96
- the similarity for the song identified by the song ID is XYZ567 If is set to 0.57, when one music is selected, the music specified by the music ID of ABC123 with the highest similarity will be selected.
- step S274 the selection reason generation unit 28 generates a selection reason sentence indicating the reason why the music piece selected by the similarity calculation unit 27 is selected, and presents the music together with the music ID of the selected music. Output to section 29.
- step S 275 the music presentation unit 29 presents the music of the music ID input from the selection reason generation unit 28 and the selection reason sentence to the user, and the process ends.
- step S281 to step S284 is similar to each of step S121 to step S124 of FIG. 15, the description thereof will be omitted.
- step S285 the music extraction unit 23 determines an evaluation value based on the preference value corresponding to each of the identified clusters and the weight of the i-th layer.
- FIG. 33 is a diagram showing an example of preference values that are also belonging weight strengths equal to or greater than a threshold value of 0.6 among belonging weights of preference values shown in FIG.
- the belonging weight which is less than 0.6 is replaced with 0. 0, whereby the preference value shown in FIG. 33 is obtained.
- a cluster identified by a cluster ID CL11 which is a cluster in the first hierarchy
- a cluster identified by a cluster ID CL12 a cluster identified by a cluster ID CL13
- CL14 The belonging weights of the preference value of the user specified by the user ID, which is U001, to the cluster specified by the cluster ID are respectively: 0, 0, 0, 8, 0, 0, And 0.6.
- a cluster in the second hierarchy a cluster identified by a cluster ID CL21, a cluster identified by a cluster ID CL22, a cluster identified by a cluster ID CL23, and a cluster ID CL24
- the belonging weights of the preference value of the user specified by the user ID, which is U001, to the cluster are 0. 0, 0. 6, 0. 7, and 0. 0, respectively.
- a cluster identified by the cluster ID CL31, a cluster identified by the cluster ID CL32, and a cluster identified by the cluster ID CL33 U001
- the belonging weights of the preference value of the user specified by the user ID that is are 0.7, 0.0, and 0. 0 respectively.
- the cluster identified in the fourth hierarchy the cluster identified by the cluster ID CL41, the cluster identified by the cluster ID CL42, the cluster identified by the cluster ID CL43, and the cluster ID CL44
- the attribution weights of the preference value of the user identified by the user ID, which is U001, to the cluster identified by are 0. 0, 0. 0, 0. 0, and 0. 0, respectively.
- step S285 the music extracting unit 23 determines that the music belonging to the cluster has a weight attributable to the cluster at the preference value which also has an attribute weighting power equal to or higher than the threshold and From the weight, the similarity is calculated by the operation shown in equation (3). That is, among the belonging weights of the original preference value, a value obtained by multiplying the belonging weight less than the threshold, for example, 0.6, is not added to the similarity, and of the belonging weights of the original preference value. The value obtained by multiplying with the membership weight which is equal to or higher than the threshold value is added to the similarity.
- FIG. 34 shows Equation (3) based on preference values including belonging weights equal to or higher than the threshold in FIG. 33, cluster information indicating belonging weights of clusters in FIG. 28, and weights for each hierarchy in FIG. It is a figure which shows the example of the similarity calculated by the shown calculation.
- the preference value of the user identified by the user ID that is U001 which corresponds to the attribution weight of the first layer of the preference value whose belonging weight is equal to or greater than the threshold, and the attribution weight of the user preference value
- the belonging weight of the first layer of the cluster information of the music identified by the music ID which is ABC123
- the weight of the first layer are multiplied and the multiplied result is integrated, it is easy to be ABC123 of FIG.
- a value of 0.15 allocated to the first layer for the song ID is obtained.
- the second and third layers And the fourth layer is the preference value of the user specified by the user ID of U001, corresponding to the belonging weight of the preference value consisting of the belonging weight or more of the threshold and the belonging weight of the user preference value,
- the belonging weight of the cluster information of the music identified by the music ID which is ABC123 is multiplied by the weight of that layer among the second, third, and fourth layers, and the multiplication result is integrated.
- the similarity between the preference value of the user identified by the user ID U001 and the cluster information of the song identified by the song ID ABC123 is 1st layer, 2nd layer, 2nd layer
- the third and fourth layers, respectively, are calculated as the caluculated values of 0. 15, 0. 05, 0. 00, and 0. 00, which have been determined by the following procedure.
- the preference value of the user identified by the user ID that is U001 which corresponds to the attribution weight of the first layer of the preference value whose belonging weight is equal to or greater than the threshold and the attribution weight of the user preference value,
- the belonging weight of the first layer of the cluster information of the music identified by the music ID which is CTH 863 is multiplied by the weight of the first layer and the multiplied result is integrated, it is easy to be CTH 863 of FIG.
- the value 0. 00 assigned to the first layer for the song ID is obtained.
- the preference value of the user identified by the user ID of U001 is the preference value of the user identified by the user ID of U001, and the belonging weight of the preference value consisting of the belonging weight or more of the threshold and the user's
- the belonging weight of the cluster information of the music identified by the music ID which is CTH 863 corresponding to the belonging weight of the preference value is multiplied by the weight of that layer among the second, third or fourth layer. And the multiplied results are integrated, the values placed in the second, third, and fourth layers for the song ID of CTH 863 in FIG. 34 are 0.10, 0. 00, 0, respectively. 00 is required.
- the similarity between the preference value of the user specified by the user ID U001 and the cluster information of the music specified by the music ID CTH 863 is 1st layer, 2nd layer, 2nd layer
- the third layer and the fourth layer are calculated as the value of 0.10, which is the value obtained by calculating 0000, 0.10. 0, 00, and 0. 00, which were calculated by the following procedure.
- the preference value of the user identified by the user ID that is U001 which corresponds to the attribution weight of the first layer of the preference value that is equal to or greater than the threshold and the attribution weight of the user preference value,
- the belonging weight of the first layer of the cluster information of the music identified by the music ID which is XYZ567, and the first The weight of the layer is multiplied, and the multiplied results are integrated to obtain 0. 07 which is the value arranged in the first layer for the music ID, which is XYZ567 in FIG.
- the preference value of the user identified by the user ID of U001 is the preference value of the user identified by the user ID of U001, and the belonging weight of the preference value consisting of the belonging weight or more of the threshold and the user's
- the belonging weight of the cluster information of the music identified by the music ID which is XYZ567, corresponding to the belonging weight of the preference value is multiplied by the weight of that layer among the second, third, or fourth layer. And the multiplied results are integrated, the values stored in layers 2, 3, and 4 for the music ID, which are XYZ 567 in FIG. , 0,00 force S required.
- the similarity between the preference value of the user specified by the user ID U001 and the cluster information of the music specified by the music ID XYZ567 is 1st layer, 2nd layer, 2nd layer
- the third layer and the fourth layer are calculated as the value 08.08, which is the caluculated value of 0. 07, 0. 00, 0. 00, and 0. 00, respectively, which were obtained by calculation.
- steps S286 to S292 is the same as each of steps S126 to S132 in FIG. 15, and thus the description thereof will be omitted.
- the present invention is not limited to this, and it is possible to calculate the weight with a larger value when the variance of the belonging weight in the layer is large.
- the entropy H may be calculated by equation (4), and a weight may be calculated which is a value obtained as a result of bowing I entropy from 1!
- the amount of calculation for selecting appropriate content can be reduced while minimizing the loss of information.
- the program may be processed by one computer, or a plurality of programs may be processed. Processing may be distributed by a computer. Furthermore, the program may be transferred to a remote computer for execution.
- a system refers to an entire apparatus configured by a plurality of apparatuses.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Reverberation, Karaoke And Other Acoustics (AREA)
Abstract
Description
Claims
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2006800008473A CN101044484B (zh) | 2005-06-16 | 2006-06-12 | 信息处理装置和信息处理方法 |
EP06766604A EP1804182A4 (en) | 2005-06-16 | 2006-06-12 | APPARATUS, METHOD AND PROGRAM FOR INFORMATION PROCESSING |
US11/660,313 US7953735B2 (en) | 2005-06-16 | 2006-06-12 | Information processing apparatus, method and program |
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2005176519 | 2005-06-16 | ||
JP2005-176519 | 2005-06-16 | ||
JP2006151011A JP4752623B2 (ja) | 2005-06-16 | 2006-05-31 | 情報処理装置、情報処理方法、およびプログラム |
JP2006-151011 | 2006-05-31 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2006134866A1 true WO2006134866A1 (ja) | 2006-12-21 |
Family
ID=37532228
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2006/311742 WO2006134866A1 (ja) | 2005-06-16 | 2006-06-12 | 情報処理装置、方法、およびプログラム |
Country Status (6)
Country | Link |
---|---|
US (1) | US7953735B2 (ja) |
EP (1) | EP1804182A4 (ja) |
JP (1) | JP4752623B2 (ja) |
KR (1) | KR20080011643A (ja) |
CN (1) | CN101044484B (ja) |
WO (1) | WO2006134866A1 (ja) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2008234819A (ja) * | 2006-12-29 | 2008-10-02 | Harman Internatl Industries Inc | 個人化されたコンテンツを備える車両インフォテインメントシステム |
JP2015084240A (ja) * | 2009-06-18 | 2015-04-30 | コア ワイヤレス ライセンシング エス.アー.エール.エル. | コンテンツを分類するための方法および装置 |
Families Citing this family (29)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100850774B1 (ko) * | 2006-11-13 | 2008-08-06 | 삼성전자주식회사 | 컨텐츠 분류 방법 및 그 방법을 수행할 수 있는 컨텐츠재생 장치 |
JP4979000B2 (ja) * | 2007-01-05 | 2012-07-18 | Kddi株式会社 | 情報検索方法、装置およびプログラム |
JP2008257414A (ja) * | 2007-04-04 | 2008-10-23 | Toyota Central R&D Labs Inc | 情報選択支援システム、端末装置、情報選択支援装置、情報選択支援方法、プログラム |
JP4389973B2 (ja) | 2007-06-26 | 2009-12-24 | ソニー株式会社 | 情報処理装置および方法、並びにプログラム |
JP5135931B2 (ja) * | 2007-07-17 | 2013-02-06 | ヤマハ株式会社 | 楽曲加工装置およびプログラム |
JP4433326B2 (ja) | 2007-12-04 | 2010-03-17 | ソニー株式会社 | 情報処理装置および方法、並びにプログラム |
US20110060738A1 (en) | 2009-09-08 | 2011-03-10 | Apple Inc. | Media item clustering based on similarity data |
CN102088750B (zh) * | 2009-12-08 | 2014-08-06 | 中国移动通信集团公司 | 一种多入多出技术中的传播径分簇方法和装置 |
JP5516720B2 (ja) | 2010-03-26 | 2014-06-11 | 富士通株式会社 | カテゴリ生成プログラム、カテゴリ生成装置およびカテゴリ生成方法 |
JP2012027845A (ja) * | 2010-07-27 | 2012-02-09 | Sony Corp | 情報処理装置、関連文提供方法、及びプログラム |
US9204200B2 (en) * | 2010-12-23 | 2015-12-01 | Rovi Technologies Corporation | Electronic programming guide (EPG) affinity clusters |
US8577876B2 (en) * | 2011-06-06 | 2013-11-05 | Met Element, Inc. | System and method for determining art preferences of people |
JP5551761B2 (ja) * | 2012-02-02 | 2014-07-16 | 株式会社コナミデジタルエンタテインメント | 情報提供システム、サーバ装置、コンピュータプログラムおよび制御方法 |
WO2013114753A1 (ja) * | 2012-02-02 | 2013-08-08 | 株式会社コナミデジタルエンタテインメント | 情報提供システム、サーバ装置、記録媒体および制御方法 |
WO2013114754A1 (ja) * | 2012-02-02 | 2013-08-08 | 株式会社コナミデジタルエンタテインメント | 情報提供システム、サーバ装置、記録媒体および制御方法 |
AU2014219089B2 (en) | 2013-02-25 | 2019-02-14 | Nant Holdings Ip, Llc | Link association analysis systems and methods |
JP2015036892A (ja) | 2013-08-13 | 2015-02-23 | ソニー株式会社 | 情報処理装置、情報処理方法、及び、プログラム |
JP2015056139A (ja) * | 2013-09-13 | 2015-03-23 | 株式会社東芝 | 電子機器、番組推奨システム、番組推奨方法及び番組推奨プログラム |
KR20160051983A (ko) * | 2014-10-30 | 2016-05-12 | 현대자동차주식회사 | 차량의 음악 추천 시스템 및 그 방법 |
WO2016098589A1 (ja) * | 2014-12-15 | 2016-06-23 | ソニー株式会社 | 情報処理装置、情報処理方法、プログラム、および情報処理システム |
US10776333B2 (en) * | 2016-07-22 | 2020-09-15 | International Business Machines Corporation | Building of object index for combinatorial object search |
US10936653B2 (en) | 2017-06-02 | 2021-03-02 | Apple Inc. | Automatically predicting relevant contexts for media items |
CN109448684B (zh) * | 2018-11-12 | 2023-11-17 | 合肥科拉斯特网络科技有限公司 | 一种智能编曲方法和系统 |
US10922337B2 (en) * | 2019-04-30 | 2021-02-16 | Amperity, Inc. | Clustering of data records with hierarchical cluster IDs |
US11003643B2 (en) * | 2019-04-30 | 2021-05-11 | Amperity, Inc. | Multi-level conflict-free entity clusterings |
GB2585890B (en) * | 2019-07-19 | 2022-02-16 | Centrica Plc | System for distributed data processing using clustering |
WO2021059473A1 (ja) * | 2019-09-27 | 2021-04-01 | ヤマハ株式会社 | 音響解析方法、音響解析装置およびプログラム |
CN113704597A (zh) * | 2020-05-21 | 2021-11-26 | 阿波罗智联(北京)科技有限公司 | 内容推荐方法、装置和设备 |
JP2022126099A (ja) * | 2021-02-18 | 2022-08-30 | 富士通株式会社 | 情報処理プログラム、情報処理方法及び情報処理装置 |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004035340A (ja) * | 2002-07-04 | 2004-02-05 | Sharp Corp | 基板、その基板を用いた板状体の製造方法、板状体およびその板状体から作製した太陽電池 |
JP2004117587A (ja) * | 2002-09-24 | 2004-04-15 | Xing Inc | 選曲支援装置 |
JP2004206679A (ja) * | 2002-12-12 | 2004-07-22 | Sony Corp | 情報処理装置および方法、記録媒体、並びにプログラム |
Family Cites Families (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2809341B2 (ja) * | 1994-11-18 | 1998-10-08 | 松下電器産業株式会社 | 情報要約方法、情報要約装置、重み付け方法、および文字放送受信装置。 |
US20050038819A1 (en) | 2000-04-21 | 2005-02-17 | Hicken Wendell T. | Music Recommendation system and method |
US6411724B1 (en) * | 1999-07-02 | 2002-06-25 | Koninklijke Philips Electronics N.V. | Using meta-descriptors to represent multimedia information |
US7162482B1 (en) * | 2000-05-03 | 2007-01-09 | Musicmatch, Inc. | Information retrieval engine |
EP1156430A2 (en) * | 2000-05-17 | 2001-11-21 | Matsushita Electric Industrial Co., Ltd. | Information retrieval system |
JP3707361B2 (ja) * | 2000-06-28 | 2005-10-19 | 日本ビクター株式会社 | 情報提供サーバ及び情報提供方法 |
US7073193B2 (en) * | 2002-04-16 | 2006-07-04 | Microsoft Corporation | Media content descriptions |
US7194527B2 (en) | 2002-06-18 | 2007-03-20 | Microsoft Corporation | Media variations browser |
US7657907B2 (en) * | 2002-09-30 | 2010-02-02 | Sharp Laboratories Of America, Inc. | Automatic user profiling |
JP4003127B2 (ja) * | 2002-12-12 | 2007-11-07 | ソニー株式会社 | 情報処理装置および情報処理方法、情報処理システム、記録媒体、並びにプログラム |
JP2004199544A (ja) * | 2002-12-20 | 2004-07-15 | Mitsubishi Electric Corp | 商品提案支援装置及び商品提案支援方法及びプログラム及びプログラムを記録したコンピュータ読み取り可能な記録媒体 |
US7120619B2 (en) | 2003-04-22 | 2006-10-10 | Microsoft Corporation | Relationship view |
JP2004355340A (ja) * | 2003-05-29 | 2004-12-16 | Sony Corp | 情報処理装置および方法、プログラム、並びに記録媒体 |
JP4305836B2 (ja) * | 2003-08-29 | 2009-07-29 | 日本ビクター株式会社 | コンテンツ検索表示装置およびコンテンツ検索表示方法 |
US7693827B2 (en) * | 2003-09-30 | 2010-04-06 | Google Inc. | Personalization of placed content ordering in search results |
JP2006048286A (ja) * | 2004-08-03 | 2006-02-16 | Sony Corp | 情報処理装置および方法、並びにプログラム |
US7558769B2 (en) * | 2005-09-30 | 2009-07-07 | Google Inc. | Identifying clusters of similar reviews and displaying representative reviews from multiple clusters |
-
2006
- 2006-05-31 JP JP2006151011A patent/JP4752623B2/ja not_active Expired - Fee Related
- 2006-06-12 EP EP06766604A patent/EP1804182A4/en not_active Withdrawn
- 2006-06-12 CN CN2006800008473A patent/CN101044484B/zh not_active Expired - Fee Related
- 2006-06-12 US US11/660,313 patent/US7953735B2/en not_active Expired - Fee Related
- 2006-06-12 WO PCT/JP2006/311742 patent/WO2006134866A1/ja active Application Filing
- 2006-06-12 KR KR1020077003694A patent/KR20080011643A/ko not_active Application Discontinuation
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004035340A (ja) * | 2002-07-04 | 2004-02-05 | Sharp Corp | 基板、その基板を用いた板状体の製造方法、板状体およびその板状体から作製した太陽電池 |
JP2004117587A (ja) * | 2002-09-24 | 2004-04-15 | Xing Inc | 選曲支援装置 |
JP2004206679A (ja) * | 2002-12-12 | 2004-07-22 | Sony Corp | 情報処理装置および方法、記録媒体、並びにプログラム |
Non-Patent Citations (1)
Title |
---|
See also references of EP1804182A4 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2008234819A (ja) * | 2006-12-29 | 2008-10-02 | Harman Internatl Industries Inc | 個人化されたコンテンツを備える車両インフォテインメントシステム |
JP2010049270A (ja) * | 2006-12-29 | 2010-03-04 | Harman Internatl Industries Inc | 個人化されたコンテンツを備える車両インフォテインメントシステム |
JP2011076709A (ja) * | 2006-12-29 | 2011-04-14 | Harman Internatl Industries Inc | 個人化されたコンテンツを備える車両インフォテインメントシステム |
JP2015084240A (ja) * | 2009-06-18 | 2015-04-30 | コア ワイヤレス ライセンシング エス.アー.エール.エル. | コンテンツを分類するための方法および装置 |
US9514472B2 (en) | 2009-06-18 | 2016-12-06 | Core Wireless Licensing S.A.R.L. | Method and apparatus for classifying content |
Also Published As
Publication number | Publication date |
---|---|
EP1804182A4 (en) | 2007-12-12 |
US7953735B2 (en) | 2011-05-31 |
CN101044484B (zh) | 2010-05-26 |
CN101044484A (zh) | 2007-09-26 |
US20090043811A1 (en) | 2009-02-12 |
JP2007026425A (ja) | 2007-02-01 |
JP4752623B2 (ja) | 2011-08-17 |
EP1804182A1 (en) | 2007-07-04 |
KR20080011643A (ko) | 2008-02-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2006134866A1 (ja) | 情報処理装置、方法、およびプログラム | |
JP4378646B2 (ja) | 情報処理装置、情報処理方法、およびプログラム | |
US10275499B2 (en) | Tag selection, clustering, and recommendation for content hosting services | |
JP4622808B2 (ja) | 楽曲分類装置、楽曲分類方法、楽曲分類プログラム | |
US7912565B2 (en) | Method for creating and accessing a menu for audio content without using a display | |
CN101821735B (zh) | 生成与内容项的集合相关联的元数据 | |
KR101194705B1 (ko) | 연관된 컨텐츠의 검색에 의한 위치 유사성 문턱값의 적응화 | |
US20090158155A1 (en) | Playlist generation, delivery and navigation | |
KR101540429B1 (ko) | 콘텐츠 플레이 리스트 추천 방법 및 장치 | |
US20100217755A1 (en) | Classifying a set of content items | |
US20090063568A1 (en) | Method and apparatus for constructing user profile using content tag, and method for content recommendation using the constructed user profile | |
WO2001020483A2 (en) | Finding perceptually similar music with precomputation | |
JP2000035974A (ja) | 主観的画像内容類似性に基づく検索用の方法及びコンピュ―タプログラム製品 | |
WO2001003002A2 (en) | Meta-descriptor for multimedia information | |
WO2007013390A1 (ja) | 情報処理装置、特徴抽出方法、記録媒体、および、プログラム | |
JP2009500670A (ja) | エレメントのシーケンスの発生の方法及び装置 | |
JP2008522311A (ja) | コンテンツアイテムの関連付けに基づく自動的なコンテンツ整理 | |
Stenzel et al. | Improving Content-Based Similarity Measures by Training a Collaborative Model. | |
JP2002342360A (ja) | 情報提供装置、その方法、そのコンピュータ・プログラムおよびそのプログラムを記録した記録媒体 | |
CN112800270A (zh) | 基于音乐标签和时间信息的音乐推荐方法及系统 | |
Goh et al. | User song preferences using artificial intelligence | |
JP2690861B2 (ja) | 情報再生装置 | |
Yu | Case Study: Artificial Intelligence in Spotify |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 2006766604 Country of ref document: EP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1127/DELNP/2007 Country of ref document: IN |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1020077003694 Country of ref document: KR |
|
WWE | Wipo information: entry into national phase |
Ref document number: 200680000847.3 Country of ref document: CN |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
WWP | Wipo information: published in national office |
Ref document number: 2006766604 Country of ref document: EP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
WWE | Wipo information: entry into national phase |
Ref document number: 11660313 Country of ref document: US |