Program, information processing method, and information processing device

Abstract

To provide a program, etc. for acquiring preferences of each customer.SOLUTION: A computer runs a program to acquire content data related to content selected by a user. The computer then outputs preference information regarding product- or service-related preferences of the user on the basis of the acquired content data.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present disclosure relates to a program, an information processing method, and an information processing apparatus.

Background Art

[0002] In Patent Document 1, a system is disclosed that determines a product or category to be recommended to a customer based on points obtained by quantifying the customer's preference for a product or service. In the technology disclosed in Patent Document 1, when a customer purchases or selects a product or service, points are added not only to the product or service but also to products or services belonging to the same category as the product or service. This makes it possible to recommend products or services that reflect not only the products or services purchased or selected by the customer but also the customer's preference for the category of the product or service.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In Patent Document 1, a product or category to be recommended to a customer is determined based on points for each product accumulated according to products the customer has purchased in the past. Therefore, there is a problem that it is impossible to determine a product or category to be recommended for a customer for whom points for each product have not been accumulated.

[0005] In one aspect, an object is to provide a program or the like that can acquire the preference of each customer.

Means for Solving the Problems

[0006] A program according to one aspect causes a computer to execute a process of obtaining content data related to content selected by a user and outputting preference information related to the preference of the user for products or services based on the obtained content data.

Advantages of the Invention

[0007] In one aspect, it is possible to obtain the preferences of each customer.

Brief Description of the Drawings

[0008]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Embodiment for Carrying Out the Invention

[0009] Hereinafter, the program, information processing method, and information processing apparatus of the present disclosure will be described in detail based on the drawings showing their embodiments.

[0010] (Embodiment 1) An information processing apparatus for estimating a user's preference (preferability) for a product or service based on content data related to content selected (designated) by the user will be described. When selecting desired content (preferred content) from various types of content such as articles, books, music (pieces, etc.), videos (movies, TV programs, plays, YouTube (registered trademark), etc.), photos, comics, animations, illustrations, computer graphics, games, etc., the user often makes a selection according to their own sensibilities. Also, when purchasing various products and services, the user selects the products and services to be purchased according to their own sensibilities. Thus, the user selects desired content and products and services to be purchased according to their own sensibilities, and it is considered that there is a correlation between the user's preference for various types of content and the user's preference for various products and services. Therefore, by estimating the user's sensibilities based on the content selected by the user, for example, data related to the content selected or purchased by the user, the user's preference (preferability) for various products or services can be estimated from the estimated sensibilities. Note that it is desirable to use a combination in which there is a strong correlation between the type of content used for estimating sensibilities and the type of product or service for which the preference is estimated from the estimated sensibilities, with each numerical value of the sensibility factors estimated from the user's image or impression of the content and each numerical value of the sensibility factors estimated from the user's image or impression of the product or service. The inventor of the present application has found that there is a strong correlation between the sensibilities when the user selects their favorite music content and the sensibilities when the user selects their favorite fashion brand. Therefore, in the present embodiment, an information processing apparatus will be described that estimates the user's sensibilities from content data related to the music content selected by the user, with the content being music content and the product or service being a fashion brand, and estimates the user's preference (preferability) for a plurality of fashion brands from the estimated sensibilities.

[0011] FIG. 1 is a block diagram showing a configuration example of an information processing apparatus. The information processing apparatus 10 is a computer capable of performing various information processes and information transmission and reception, and is a server computer, a personal computer, a tablet terminal, or the like. The information processing apparatus 10 includes a control unit 11, a storage unit 12, a communication unit 13, an input unit 14, a display unit 15, a reading unit 16, etc., and these units are mutually connected via a bus. The control unit 11 is configured using one or a plurality of processors such as a CPU (Central Processing Unit), an MPU (Micro-Processing Unit), a GPU (Graphics Processing Unit), a TPU (Tensor Processing Unit), or an AI chip (semiconductor for AI). The control unit 11 executes various information processes and control processes to be performed by the information processing apparatus 10 by appropriately executing the program P stored in the storage unit 12. When the control unit 11 includes a plurality of processors, the control unit 11 may execute each process by different processors.

[0012] The storage unit 12 includes a RAM (Random Access Memory), a flash memory, a hard disk, an SSD (Solid State Drive), etc. The storage unit 12 stores the program P (program product) executed by the control unit 11 and various data necessary for the execution of the program P. Also, the storage unit 12 temporarily stores data and the like generated when the control unit 11 executes the program P. Further, the storage unit 12 stores, for example, the learned learning models M1 and M2 of the training data by machine learning. The learning models M1 and M2 are assumed to be used as program modules constituting artificial intelligence software. The learning models M1 and M2 perform a predetermined operation on the input value and output the operation result. In the storage unit 12, data such as the coefficients and thresholds of the function defining this operation are stored as the learning models M1 and M2. The storage unit 12 may be composed of a plurality of storage devices, and a part of the storage unit 12 may be another storage device connected to the information processing device 10, or may be another storage device to which the information processing device 10 can communicate. Also, instead of the configuration in which the learning models M1 and M2 are stored in the storage unit 12, the information processing device 10 may be configured to access and read the learning models M1 and M2 from a server that stores the learning models M1 and M2.

[0013] The communication unit 13 is a communication module for performing processing related to wired communication or wireless communication, and transmits and receives information to and from other devices via a network. The network may be the Internet or a public telephone network, or may be a LAN (Local Area Network) constructed within the facility where the information processing device 10 is provided. The input unit 14 receives an operation input by the user and sends a control signal corresponding to the operation content to the control unit 11. The input unit 14 includes, for example, a keyboard and a mouse, a microphone for voice input, etc. The display unit 15 is a liquid crystal display or an organic EL (electroluminescence) display, etc., and displays various information according to an instruction from the control unit 11. A part of the input unit 14 and the display unit 15 may be a touch panel configured integrally.

[0014] The reading unit 16 reads information stored in a portable storage medium 10a such as a CD (Compact Disc), DVD (Digital Versatile Disc), USB (Universal Serial Bus) memory, or SD (Secure Digital) card. The program P and various data stored in the storage unit 12 may be read by the control unit 11 from the portable storage medium 10a via the reading unit 16 and stored in the storage unit 12. Also, the program P and various data may be written to the storage unit 12 at the manufacturing stage of the information processing apparatus 10, or may be downloaded by the control unit 11 from another apparatus via the communication unit 13 and stored in the storage unit 12.

[0015] In the present embodiment, the information processing apparatus 10 may be a multi-computer composed of a plurality of computers, or may be a virtual machine virtually constructed by software within a single apparatus. Also, when the information processing apparatus 10 is configured as a server computer, the information processing apparatus 10 may be a local server installed within the facility where the information processing apparatus 10 is provided, or may be a cloud server communicatively connected via a network such as the Internet. Hereinafter, the information processing apparatus 10 will be described as being a single computer. Also, the program P may be arranged on a single computer or at one site, or may be distributed across a plurality of sites and executed in a distributed manner on a plurality of computers interconnected via a network. Further, the input unit 14 and the display unit 15 of the information processing apparatus 10 are not essential, and the apparatus may be configured to receive operations through a connected computer, or may be configured to output information to be displayed to an external display device.

[0016] The information processing apparatus 10 according to this embodiment performs a process of estimating the preference (selectivity) of a user for a fashion brand from the answers to a questionnaire regarding the music content (piece of music) that the user likes the most. FIG. 2 is an explanatory diagram showing an example of a questionnaire screen regarding music content. On the screen shown in FIG. 2, there are provided input fields for the title and artist name of the piece of music that the user likes the most (the most favorite piece of music) or the piece of music with the highest listening frequency (the most frequently listened-to piece of music). Also, on the screen shown in FIG. 2, for the piece of music entered in the input fields for the title and artist name, a plurality of questionnaire items by the SD method (Semantic Differential Method) are displayed, and for each questionnaire item, the user is configured to answer (input) the image or impression that the user has of the piece of music by a numerical value of any one of 1 to 5. Note that the number of numerical values selectable for each questionnaire item is not limited to 5 of 1 to 5, and can be any number. Also, the number of questionnaire items can be, for example, about 50 items, but can be any number as well.

[0017] The information processing apparatus 10 of the present embodiment acquires answers to a plurality of questionnaire items regarding music content that the user most prefers via a questionnaire screen as shown in FIG. 2. Then, the information processing apparatus 10 estimates numerical values for a plurality of aesthetic factors representing the user's aesthetics from the acquired answers, and estimates the user's preference (preferability) for a plurality of fashion brands from the numerical values for the estimated aesthetic factors. The preference for a fashion brand of each estimated user can be used for promoting the fashion brand to each user. Therefore, the image or impression (answer to the questionnaire item) that the user has for the music content the user likes can be used as marketing information for the fashion brand. Note that the information processing apparatus 10 of the present embodiment uses the first learning model M1 when estimating the numerical values of the aesthetic factors representing the user's aesthetics from the answers to the questionnaire items regarding music content, and uses the second learning model M2 when estimating the preference (preferability) for a fashion brand from the numerical values of each aesthetic factor. Further, the information processing apparatus 10 may be configured to estimate the user's preference for a fashion brand from the questionnaire answers regarding the music content arbitrarily selected by the user instead of the music content that the user most prefers or the music content with the highest listening frequency of the user.

[0018] FIG. 3 is an explanatory diagram showing a configuration example of the first learning model M1 and the second learning model M2. The first learning model M1 takes, as content data regarding music content selected by a user (for example, music content that the user likes most), answers to questionnaire items regarding the music content as input, and based on the input content data, performs an operation to estimate numerical values for each of a plurality of types of sensitivity factors regarding the user's sensitivity, and is trained to output the result of the operation. In the example of FIG. 3A, answers to a plurality of questionnaire items including answers to the first questionnaire item (vulgar, elegant) shown in FIG. 2 (answer to the first item), answers to the second questionnaire item (childish, adult-like), etc. are input, and it is configured to output numerical values for each of a plurality of preset sensitivity factors. As the sensitivity factors, for example, those that can express the user's sensitivity such as a gentle sense of healing, a sense of cheerfulness, a sense of independence, an authentic sense, a sense of passion, a wild sense, a crisp sense, a naive sense, a glittering sense, etc. are used. The sensitivity factors may use some or all of these, and are not limited thereto. The number of sensitivity factors is, for example, about 13, but is not limited to this number. The numerical values for the sensitivity factors can be, for example, numerical values from -1 to 1 (for example, numerical values up to the second decimal place).

[0019] The first learning model M1 can be configured using algorithms such as multiple regression analysis, SVM (Support Vector Machine), random forest, decision tree, XGBoost (eXtreme Gradient Boosting), LightGBM (Light Gradient Boosting Machine), neural network, deep learning, etc., and may also be configured by combining multiple algorithms. In addition to the above-mentioned algorithms, the first learning model M1 may be configured using dimensionality reduction algorithms such as factor analysis, principal component analysis (PCA: Principal Component Analysis), UMAP (Uniform Manifold Approximation and Projection), LSI (Latent Semantic Indexing), SVD (Singular Value Decomposition), LDA (Linear Discriminant Analysis), ICA (Independent Component Analysis), PLS (Partial Least Squares), t-SNE (t-distributed Stochastic Neighbor Embedding), etc. When the first learning model M1 is configured using a dimensionality reduction algorithm, it is generated by unsupervised learning.

[0020] The first learning model M1 includes an input layer to which questionnaire responses regarding music content are input, an output layer that outputs numerical values for each emotional factor, and an intermediate layer that calculates output values based on the input data. The input layer of the first learning model M1 has a plurality of input nodes, and questionnaire responses regarding music content answered by the user are input through each input node. The intermediate layer of the first learning model M1 extracts feature quantities of the input data using various functions and thresholds, etc., and outputs the extracted feature values to the output layer. The output layer of the first learning model M1 has a plurality of output nodes, and preset emotional factors, such as a gentle healing feeling, a cheerful feeling, a sense of independence, an authentic feeling, a passionate feeling, a wild feeling, a refreshing feeling, a naive feeling, a glittering feeling, etc., are associated with each output node, and numerical values for the associated emotional factors are output from each output node. With such a configuration, when a predetermined number of questionnaire responses are input, the first learning model M1 outputs numerical values of the user for each emotional factor.

[0021] The first learning model M1 is generated by performing machine learning using training data in which training questionnaire responses (questionnaire responses regarding music content) are associated with numerical values (correct numerical values) of emotional factors representing the user's emotions calculated based on the questionnaire responses. The training data for the learning models M1 and M2 of this embodiment is generated based on questionnaire responses by the user. FIG. 4 is an explanatory diagram showing an example of a questionnaire screen for generating training data. In the screen shown in FIG. 4, in addition to the questionnaire regarding the music shown in FIG. 2, it includes a questionnaire regarding the degree of preference for fashion brands. In the questionnaire regarding fashion brands, preset fashion brands are displayed, and for each fashion brand, the user is configured to answer (input) the degree of preference for the fashion brand with a numerical value from 1 to 7.

[0022] The training data of the first learning model M1 is generated by associating the questionnaire responses regarding the music obtained via the questionnaire shown in FIG. 4 with the numerical values for each sentiment factor calculated by an operation using the maximum likelihood estimation method with a preset linear equation or the like based on this questionnaire response. For the linear equation for calculating the numerical value of each sentiment factor, coefficients for each sentiment factor are set for the responses to each questionnaire item (numerical values from 1 to 5 in the example of FIG. 4). The first learning model M1 learns so that when the questionnaire response included in the training data is input, the correct numerical value of each sentiment factor is output from the output node corresponding to each sentiment factor. In the learning process, the first learning model M1 performs an operation based on the input questionnaire response and calculates the output value from each output node. Then, the first learning model M1 compares the calculated output value of each output node with the correct numerical value and optimizes the parameters used in the operation process so that the two approximate each other. For example, parameters such as the weights (coupling coefficients) between nodes in the first learning model M1 are optimized using the error backpropagation method, the steepest descent method, or the like. Thereby, when the questionnaire response regarding the music of a certain user is input, the first learning model M1 that outputs the numerical value for the sentiment factor representing the sentiment of the user is obtained.

[0023] The first learning model M1 is configured to output a numerical value represented by a continuous value, for example, from -1 to 1, as a numerical value for a perceptual factor. However, it may also be configured to discriminate any of the numerical values selectable as the perceptual factor. In this case, for each perceptual factor in the first learning model M1, a plurality of output nodes corresponding to a plurality of numerical values selectable as the numerical value of the perceptual factor may be provided, and each output node may be configured to output the confidence level for the associated numerical value. In this case, for each perceptual factor, an output value, for example, from 0 to 1, is output from each output node, and the sum of the output values from each output node becomes 1 (100%). In such a configuration, when a questionnaire response regarding a piece of music is input to the first learning model M1, for each perceptual factor, the confidence level for the numerical value assigned to each output node is output from each output node. When using the first learning model M1 having such a configuration, the information processing apparatus 10 can estimate the numerical value associated with the output node that outputs the maximum output value (confidence level) as the numerical value of the corresponding perceptual factor for each perceptual factor. Note that the first learning model M1 having such a configuration may have one output node that outputs the numerical value with the highest confidence level instead of having a plurality of output nodes that output the confidence level for each numerical value for each perceptual factor. Further, the first learning model M1 can be configured using a dimensionality reduction algorithm. In this case, by performing factor analysis on the training questionnaire responses (for example, questionnaire responses for about 50 items regarding music content for 2000 people) by unsupervised learning, it learns to output the numerical values of, for example, about 13 perceptual factors. Note that the first learning model M1 learns to output perceptual factors that can contribute to the estimation of the preference for a fashion brand while maintaining the information amount of the 50-item questionnaire response.

[0024] The second learning model M2 shown in FIG. 3B takes as input the numerical values for each of a plurality of types of sentiment factors representing the user's sentiment, and based on the numerical values of the input sentiment factors, performs an operation to estimate the user's preference degree (preference information) for each of a plurality of types of fashion brands, and is trained to output the result of the operation. As the numerical values for the sentiment factors input to the second learning model M2, for example, the numerical values for the sentiment factors estimated using the first learning model M1 in FIG. 3A can be used. The preference degree for a fashion brand can be, for example, numerical values from 1 to 7 used in the questionnaire shown in FIG. 4. The second learning model M2 can also be configured using algorithms such as multiple regression analysis, SVM, random forest, decision tree, XGBoost, LightGBM, neural network, deep learning, etc., and may be configured by combining multiple algorithms.

[0025] The second learning model M2 includes an input layer to which numerical values for a plurality of sentiment factors are input, an output layer that outputs the degree of preference for each fashion brand, and an intermediate layer that calculates an output value based on the input data. The input layer of the second learning model M2 has a plurality of input nodes, and through each input node, numerical values for a plurality of sentiment factors are input. The intermediate layer of the second learning model M2 extracts feature amounts of the input data using various functions, thresholds, etc., and outputs the extracted feature values to the output layer. The output layer of the second learning model M2 has a plurality of output nodes, and each output node is associated with a preset fashion brand, and from each output node, the degree of preference for the associated fashion brand is output. Specifically, the output layer of the second learning model M2 has a plurality (here, seven) of output nodes associated with respective numerical values (for example, degrees of preference from 1 to 7) that can be selected as the degree of preference for each fashion brand, and from each output node, the confidence level for the associated degree of preference is output. For each fashion brand, output values from 0 to 1 are output from each output node, and the sum of the output values from each output node is 1 (100%). When using the second learning model M2 with such a configuration, the information processing apparatus 10 can estimate, for each fashion brand, the degree of preference (for example, any one of 1 to 7) associated with the output node that outputs the maximum output value (confidence level) as the degree of preference for the fashion brand. Note that the second learning model M2 with such a configuration may have a configuration with one output node that outputs the degree of preference with the highest confidence level instead of having a plurality of output nodes that output the confidence level for each degree of preference for each fashion brand.

[0026] The second learning model M2 is generated by performing machine learning using training data that associates each numerical value of the emotional factors for training with the preference level (correct preference level) for fashion brands by the same user as the questionnaire responses used to calculate each numerical value of the emotional factors. Specifically, the training data for the second learning model M2 is generated by associating numerical values for each emotional factor calculated based on questionnaire responses regarding music with the preference level for fashion brands (numerical values from 1 to 7 in the example of FIG. 4) obtained via the questionnaire shown in FIG. 4. The second learning model M2 learns so that when the numerical values of each emotional factor included in the training data are input, the correct preference level for each fashion brand is output from the output node corresponding to each fashion brand. In the learning process, the second learning model M2 performs operations based on the input numerical values of each emotional factor and calculates the output values from each output node. Then, for each fashion brand, the second learning model M2 compares the calculated output value of each output node with a value corresponding to the correct preference level (1 for the output node corresponding to the correct preference level and 0 for other output nodes), and optimizes the parameters used in the operation process so that the two are approximated. Here too, parameters such as the weights (coupling coefficients) between nodes in the second learning model M2 are optimized using the error backpropagation method, the steepest descent method, etc. Thereby, when the numerical values of the emotional factors representing the emotions of a certain user are input, the second learning model M2 that outputs the preference level of the user for fashion brands is obtained.

[0027] The learning of the learning models M1 and M2 may be performed by the information processing device 10 or may be performed by other learning devices. The learned learning models M1 and M2 generated by learning in other learning devices are downloaded from the learning device to the information processing device 10 via a network or via the portable storage medium 10a and stored in the storage unit 12.

[0028] Next, a process of learning the training data as described above to generate learning models M1 and M2 will be described. FIG. 5 is a flowchart showing an example of the generation process procedure of the learning models M1 and M2. The following process is executed by the control unit 11 of the information processing apparatus 10 according to the program P stored in the storage unit 12, but it may be performed by other learning apparatuses. In the following process, the control unit 11 first generates training data based on the answers to the questionnaire shown in FIG. 4, and uses the generated training data to learn the learning models M1 and M2. It is assumed that the questionnaire answers used for generating the training data are those answered by a large number of respondents and stored in a predetermined area (predetermined DB) of the storage unit 12.

[0029] The control unit 11 of the information processing apparatus 10 acquires the questionnaire answer for one person stored in the storage unit 12 (S11). In addition to reading from the storage unit 12, the questionnaire answer may be acquired, for example, from the terminal of the questionnaire respondent via a network, a terminal that collects the questionnaire answer, or the like. The acquired questionnaire answer includes the questionnaire answer for the music and the questionnaire answer for the preference degree of the fashion brand.

[0030] The control unit 11 extracts the questionnaire answer for the music from the questionnaire answer, and based on the questionnaire answer for the music, calculates the numerical value for each sensitivity factor representing the sensitivity of this questionnaire respondent (S12). Here, the control unit 11 calculates the numerical value (maximum likelihood estimator) of each sensitivity factor representing the sensitivity of the questionnaire respondent by the maximum likelihood estimation method for the sensitivity of the questionnaire respondent. Specifically, the control unit 11 calculates the numerical value of the sensitivity factor representing the sensitivity of the questionnaire respondent by a linear equation with coefficients preset for each questionnaire item. The control unit 11 associates the extracted questionnaire answer for the music with the calculated numerical values (correct numerical values) of each sensitivity factor, generates training data for learning the first learning model M1, and stores it in the storage unit 12 (S13). The control unit 11 stores the generated training data, for example, in a training DB (not shown) for the first learning model M1 prepared in the storage unit 12.

[0031] The control unit 11 extracts the questionnaire responses regarding the preference for fashion brands among the questionnaire responses, associates the numerical values of each aesthetic factor calculated in step S12 with the preference (correct preference) for the extracted fashion brand, generates training data to be used for training the second learning model M2, and stores it in the storage unit 12 (S14). Here too, the control unit 11 stores the generated training data in, for example, a training DB (not shown) for the second learning model M2 prepared in the storage unit 12.

[0032] The control unit 11 determines whether there is any unprocessed questionnaire response among the questionnaire responses stored in the storage unit 12 that has not been used in the generation process of the training data (S15). If it is determined that there is an unprocessed questionnaire response (S15: YES), the control unit 11 returns to the process of step S11 and performs the processes of steps S11 to S14 for the unprocessed questionnaire response. The control unit 11 repeats the processes of steps S11 to S15 until it determines that there is no unprocessed questionnaire response. As a result, based on the questionnaire responses stored in the storage unit 12, training data to be used for training the learning models M1 and M2 is generated and accumulated in the training DB (the training DB for the first learning model M1 and the training DB for the second learning model M2).

[0033] When the control unit 11 determines that there is no unanswered questionnaire (S15: NO), it performs learning of the learning models M1 and M2 using the training data stored in the training DB as described above. For example, the control unit 11 reads out one of the training data stored in the training DB for the first learning model M1, and performs learning processing of the first learning model M1 based on the read training data (S16). Here, the control unit 11 inputs the questionnaire answer for the music included in the training data to the first learning model M1, and acquires the output value output from the first learning model M1 when the questionnaire answer is input. The control unit 11 compares the output value of each output node output from the first learning model M1 with the correct numerical value of each emotional factor included in the training data, and learns the first learning model M1 so that both are approximated. In the learning process, the control unit 11 optimizes the parameters such as the weights between the nodes in the first learning model M1 using the error backpropagation method that sequentially updates the parameters from the output layer toward the input layer.

[0034] The control unit 11 determines whether there is any unprocessed training data among the training data stored in the training DB for the first learning model M1 for which learning processing has not been performed (S17). If it is determined that there is unprocessed training data (S17: YES), the control unit 11 returns to the processing of step S16 and performs the learning processing of the first learning model M1 using the unprocessed training data for which learning processing has not been performed. If it is determined that there is no unprocessed training data (S17: NO), the control unit 11 reads out one of the training data accumulated in the training DB for the second learning model M2, and based on the read training data, performs the learning processing of the second learning model M2 (S18). Here, the control unit 11 inputs the numerical values of each sensitivity factor included in the training data into the second learning model M2, and acquires the output values output from the second learning model M2 when the numerical values of each sensitivity factor are input. The control unit 11 compares the output values of each output node output from the second learning model M2 with the correct preference degrees for each fashion brand included in the training data, and learns the second learning model M2 so that the two are approximated. Specifically, for each fashion brand, the control unit 11 compares the output value of each output node with a value corresponding to the correct preference degree (1 for the output node corresponding to the correct preference degree, and 0 for other output nodes), and learns so that the two are approximated. Also here, the control unit 11 optimizes the parameters such as the weights between the nodes in the second learning model M2 using the error backpropagation method that sequentially updates the parameters from the output layer toward the input layer.

[0035] The control unit 11 determines whether there is any unprocessed training data among the training data stored in the training DB for the second learning model M2 for which learning processing has not been performed (S19). If it is determined that there is unprocessed training data (S19: YES), the control unit 11 returns to the processing of step S18 and performs the learning processing of the second learning model M2 using the unprocessed training data for which learning processing has not been performed. If it is determined that there is no unprocessed training data (S19: NO), the control unit 11 ends the series of processing.

[0036] Through the above learning process, when a questionnaire response regarding a piece of music is input, a first learning model M1 that outputs the numerical values of the perceptual factors of the questionnaire respondent, and a second learning model M2 that outputs the preference degree for the fashion brand of the questionnaire respondent when the numerical values of each perceptual factor are input are generated. In addition, in the above-described process, the generation process of the training data by steps S11 to S15 and the generation process of the learning models M1 and M2 by steps S16 to S19 may be performed by separate devices. The learning models M1 and M2 can be further optimized by repeatedly performing the learning process using the training data as described above. Also, for the already learned learning models M1 and M2, by re-learning them with the above-described learning process, learning models M1 and M2 with further improved discrimination accuracy can be generated. When the first learning model M1 is configured with a dimensionality reduction algorithm, in step S13, the control unit 11 stores the questionnaire response regarding the music as the training data of the first learning model M1. Then, in step S16, the control unit 11 performs learning of the first learning model M1 by performing factor analysis on the training data (questionnaire response).

[0037] Hereinafter, a process in which the information processing apparatus 10 of the present embodiment estimates the preference degree for the fashion brand of an arbitrary user based on a questionnaire response regarding music by the user will be described. FIG. 6 is a flowchart showing an example of a procedure for estimating the preference degree for a fashion brand.

[0038] In the present embodiment, when a user starts using or is using a music streaming service (music distribution service), or when the user accesses various sites via the network, etc., a questionnaire as shown in FIG. 2 is conducted at an arbitrary timing. The information processing apparatus 10 estimates the preference degree for the fashion brand of the user from the response to the questionnaire conducted at an arbitrary timing.

[0039] The control unit 11 of the information processing apparatus 10 acquires a questionnaire response from an arbitrary user (S21). The control unit 11 may acquire user information including the contact information of the questionnaire respondent together with the questionnaire response. Note that the questionnaire response may be acquired in advance via the communication unit 13 or the input unit 14 and stored in the storage unit 12. In this case, the control unit 11 reads the questionnaire response from the storage unit 12. Further, the control unit 11 may be configured to acquire the questionnaire response from the terminal of the questionnaire respondent, the terminal that collects the questionnaire response, etc. via the network. Based on the acquired questionnaire response regarding the music, the control unit 11 estimates the numerical value of each sensitivity factor representing the sensitivity of the questionnaire respondent (S22). Here, the control unit 11 inputs the questionnaire response regarding the music into the first learning model M1, and acquires the output value from each output node of the first learning model M1 as the numerical value of the sensitivity factor associated with each output node.

[0040] Next, based on the estimated numerical value of each sensitivity factor, the control unit 11 estimates the preference degree of the questionnaire respondent for each fashion brand (S23). Here, the control unit 11 inputs the numerical value of each sensitivity factor into the second learning model M2, and acquires the preference degree for each fashion brand based on the output value from each output node of the second learning model M2. Specifically, for each fashion brand, the control unit 11 acquires the preference degree associated with the output node that outputs the maximum output value (confidence level) as the preference degree for the fashion brand. When the control unit 11 has acquired user information together with the questionnaire response, it associates the user information with the brand information including the preference degree for each fashion brand estimated in step S23 and stores it in the storage unit 12 (S24). Note that, in addition to the user information and the fashion information, the control unit 11 may store at least one of the questionnaire response acquired in step S21 and the numerical value of each sensitivity factor estimated in step S22 in the storage unit 12.

[0041] The control unit 11 determines whether there is an unprocessed questionnaire response for which the user's preference for a fashion brand has not been estimated (S25). If it is determined that there is an unprocessed questionnaire response (S25: YES), the control unit 11 returns to the process of step S21 and performs the processes of steps S21 to S24 for the unprocessed questionnaire response. Thereby, for each user who has answered the questionnaire regarding the music, the preference for the fashion brand can be estimated. The control unit 11 repeats the processes of steps S21 to S25 until it is determined that there is no unprocessed questionnaire response. If it is determined that there is no unprocessed questionnaire response (S25: NO), the series of processes is terminated. Thereby, based on the questionnaire response regarding the music obtained from an arbitrary user, the preference of the user for the fashion brand is estimated. Note that the preference for each fashion brand estimated in step S23 is represented by a numerical value from 1 to 7 (1: dislike very much, 7: like very much), and the levels of preference for each fashion brand can be compared.

[0042] By the above-described process, the preference of the user for the fashion brand can be estimated from the questionnaire response regarding the music content (for example, the most preferred music content) selected by the user. Therefore, it becomes possible to provide, for example, publicity and advertisements regarding fashion brands with high preference, product recommendation information, etc. to the user, and it becomes possible to provide, for example, information on users with high preference for each fashion brand to fashion brand stores and the like. Note that the user can have the fashion brand preferred by the user estimated only by answering the questionnaire regarding the music without being conscious of the preference for the fashion brand.

[0043] FIG. 7 is an explanatory diagram showing a comparison between the degree of preference of a user for a fashion brand and the degree of preference for a fashion brand estimated from the user's sensitivity to music. The graph shown in FIG. 7 has the actual degree of preference of the user for the fashion brand on the horizontal axis, and the degree of preference for the fashion brand estimated from the sensitivity (numerical values of each sensitivity factor) estimated from the questionnaire responses for the music selected by the user on the vertical axis. It is a graph in which the actual degree of preference of each user and the degree of preference estimated from the sensitivity to music are plotted. The actual degree of preference is determined by each user ranking their preferences for a plurality of brands and assigning a score corresponding to the rank (the higher the rank, the higher the score) to each brand, and subtracting the score of the second brand from the score of the first brand (score of the first brand - score of the second brand). That is, the actual degree of preference shown on the horizontal axis of the graph in FIG. 7 indicates the difference in the degree of preference of each user for two brands, showing that the higher the preference for the first brand is towards the right side, and the higher the preference for the second brand is towards the left side. The degree of preference estimated from the sensitivity to music is determined by ranking the preferences of the user for a plurality of brands estimated from the sensitivity (numerical values of each sensitivity factor) of the user calculated from the questionnaire responses regarding the music by the processing of this embodiment, assigning a score corresponding to the rank (the higher the rank, the higher the score) to each brand, and subtracting the score of the second brand from the score of the first brand. That is, the estimated degree of preference shown on the vertical axis of the graph in FIG. 7 indicates the difference in the degree of preference of each user for two brands estimated from the sensitivity to music, showing that the higher the preference for the first brand is towards the upper side, and the higher the preference for the second brand is towards the lower side.

[0044] From the graph of FIG. 7, it can be seen that there is a strong correlation between the actual preference of users for fashion brands and the preference for fashion brands estimated from the sensitivity of users to music. Therefore, in the present embodiment, from the questionnaire responses to the music selected by the user, the sensitivity of the user related to music is represented by numerical values of a plurality of sensitivity factors, and the preference of the user for fashion brands is estimated using the numerical values of the plurality of sensitivity factors representing the sensitivity of the user, whereby the preference (liking) of the user for fashion brands can be accurately estimated.

[0045] As described above, since the user makes various selections according to his or her own sensitivity, it may be configured to estimate the sensitivity from the preference (liking) of the user for various contents, not limited to music content. Also, it may be configured to estimate the preference (liking) of the user for various products and services based on the sensitivity of the user, not limited to fashion brands. For example, it may be configured to estimate the numerical values of the sensitivity factors representing the sensitivity of the user from questionnaire responses to any one or a plurality of contents such as video content (movies, TV programs, plays, YouTube, etc.), articles, books, photos, comics, animations, illustrations, computer graphics, games, etc. In addition, for the questionnaire responses used for estimating the sensitivity, in addition to the questionnaire responses regarding the most preferred music content as shown in FIG. 2, questionnaire responses regarding the user's own music sensitivity or music values, questionnaire responses regarding the image and impression of the music, artist, etc. that affect the score (numerical value) of the sensitivity factor, and the rearrangement result in the order of preference for a predetermined number of music, artists, or melody lines, etc. that affect the score of the sensitivity factor may also be used.

[0046] Also, based on the user's sentiment estimated as described above, it may be configured to estimate the user's preference (liking) for various products and services including buildings such as restaurants, shopping centers, shopping malls, landmarks, places, streets, travel destinations such as tourist attractions, types of pets, etc. Further, it may be configured to estimate the user's preference for people such as colleagues at work, friends, and the opposite sex based on the user's sentiment (each numerical value of sentiment factors). In this case, it becomes possible to estimate the compatibility with others based on the preference for others. Note that the combination of the content used for estimating the sentiment (each numerical value of sentiment factors) and the product or service for which the user's preference (liking) is estimated using the estimated sentiment is preferably a combination having a strong correlation between each numerical value of the sentiment factors estimated from the questionnaire responses for the content and each numerical value of the sentiment factors estimated from the similar questionnaire responses for the product or service.

[0047] In the above-described process, the process of estimating each numerical value of the sentiment factors representing the respondent's sentiment from the questionnaire responses regarding music using the first learning model M1, and / or the process of estimating the respondent's preference for fashion brands from each numerical value of the sentiment factors using the second learning model M2 are not limited to the configuration in which the information processing apparatus 10 performs them locally. For example, these processes may be executed by a server that provides the learning models M1 and M2. For example, the information processing apparatus 10 may be configured to transmit the received questionnaire responses to the server and receive each numerical value of the sentiment factors estimated from the questionnaire responses at the server. Further, the information processing apparatus 10 may be configured to transmit each numerical value of the sentiment factors representing the respondent's sentiment to the server and receive the respondent's preference for fashion brands estimated from each numerical value of the sentiment factors at the server.

[0048] (Embodiment 2) In the above-described Embodiment 1, an information processing apparatus that estimates a numerical value of a sensitivity factor representing a user's sensitivity based on the music included in a playlist provided by a music streaming service (music distribution service) instead of the questionnaire answers for the music will be described. Since the information processing apparatus of the present embodiment has the same configuration as the configuration of the information processing apparatus 10 of Embodiment 1 shown in FIG. 1, the description of the configuration will be omitted.

[0049] FIG. 8 is a flowchart showing an example of a preference estimation processing procedure for a fashion brand in Embodiment 2. The processing shown in FIG. 8 is obtained by adding steps S31 to S32 instead of step S21 and adding step S33 instead of step S25 in the processing shown in FIG. 6. The description of the same steps as those in FIG. 6 will be omitted. In the present embodiment, the control unit 11 of the information processing apparatus 10 acquires a playlist used by the user in the music distribution service (S31). For example, when the information processing apparatus 10 is a server that provides a music distribution service, the control unit 11 can acquire a playlist from the user's terminal by transmitting and receiving information to and from the user's terminal via a network. Also, when the user's terminal can transmit a playlist via a network, the control unit 11 may acquire the playlist from the user's terminal via the network. The playlist includes the song title and artist name of each song. Note that the control unit 11 may be configured to acquire information on each song registered as a favorite and information on each song included in the playback history (listening history) in the music distribution service instead of the playlist.

[0050] Based on the acquired playlist, the control unit 11 estimates answers to the questionnaire items as shown in FIG. 2 (S32). For example, for the music itself or music information such as the tempo, rhythm, pitch, melody line, etc. of the music, answers (for example, numerical values from 1 to 5) to each questionnaire item that a user who likes the music is likely to give are associated. Then, the control unit 11 extracts the tendency of the music included in the playlist or the music information such as the tempo, rhythm, pitch, melody line, etc. of the music, and identifies each questionnaire answer corresponding to the extracted tendency of the music or music information, thereby estimating the questionnaire answers by the user who uses the playlist. Also, when the playlist includes the number of playbacks (number of listens), the control unit 11 may extract the music with the most playbacks and estimate the questionnaire answers by the user by identifying each questionnaire answer associated with this music or the music information of this music.

[0051] After that, the control unit 11 executes the processing after step S22. Thereby, also in this embodiment, based on the questionnaire answers estimated from the playlist, the numerical values of each sensitivity factor representing the sensitivity of the user of the playlist are estimated, and the preference degree of the user of the playlist for each fashion brand can be estimated. The control unit 11 of this embodiment determines whether there is an unprocessed playlist (S33). If it is determined that there is (S33: YES), it returns to step S31. If it is determined that there is not (S33: NO), the processing ends.

[0052] By the above-described processing, in this embodiment, based on the playlist used by the user, the sensitivity of the user is represented by the numerical values of a plurality of sensitivity factors, and the preference degree of the user for the fashion brand can be estimated based on the sensitivity of the user. Therefore, the preference degree for a fashion brand that matches the user's sensitivity (taste) can be estimated. Also, in this embodiment, since the preference degree of the user for the fashion brand is estimated based on the playlist used by the user in the music distribution service, the user only needs to use the music distribution service and the operation burden does not increase.

[0053] Also in this embodiment, not limited to the playlist of music in the music distribution service, for example, in the video distribution service such as video content, or in the playlists, playback histories, or user preferences such as favorite registrations of various types of content, a configuration for estimating the sensibility from the user's preferences may be adopted. In this case, for various types of content or information related to each content, the answers to each questionnaire item are associated, and the content registered in the user's playlist, playback history, or favorite registration, or each questionnaire answer corresponding to the information related to the content is specified. Further, based on the sensibility of the user estimated from the user's preferences such as playlists, playback histories, or favorite registrations of various types of content, a configuration for estimating the user's preferences (degree of preference) for various products and services may be adopted. Also in this embodiment, the process of estimating each numerical value of the sensibility factor representing the user's sensibility from the questionnaire answers related to music (questionnaire answers estimated from the playlist) using the first learning model M1, and / or the process of estimating the degree of preference of the user for the fashion brand from each numerical value of the sensibility factor using the second learning model M2 is not limited to the configuration in which the information processing apparatus 10 performs locally. These processes may also be executed by a server that provides the learning models M1 and M2.

[0054] In this embodiment, based on the music (content) registered in the user's playlist, playback history, or favorite registration, the user's answers to the questionnaire items related to the music (content) are estimated, and except for the process of estimating each numerical value of the sensibility factor representing the user's sensibility from the estimated questionnaire answers, the same processes as in Embodiment 1 are performed, and the same effects as in Embodiment 1 can be obtained. Also in this embodiment, the application of the modification examples appropriately described in Embodiment 1 above is possible.

[0055] (Embodiment 3) The information processing apparatus 10 of the above-described Embodiments 1 to 2 estimates the degree of preference of the user for a fashion brand from the questionnaire answers by the user or the questionnaire answers estimated from the user's playlist using the first learning model M1 and the second learning model M2. In the present embodiment, an information processing apparatus that estimates the degree of preference of the user for a fashion brand from the questionnaire answers by the user or the questionnaire answers estimated from the user's playlist using one third learning model will be described. Since the information processing apparatus of the present embodiment has the same configuration as the configuration of the information processing apparatus 10 of Embodiment 1 shown in FIG. 1, a detailed description of the configuration will be omitted. Note that the information processing apparatus 10 of the present embodiment stores a third learning model M3 in the storage unit 12 instead of the first learning model M1 and the second learning model M2.

[0056] FIG. 9 is an explanatory diagram showing a configuration example of the third learning model M3. Similar to the input data of the first learning model M1, the third learning model M3 takes as input a questionnaire answer regarding music content selected by the user (for example, the music content that the user likes the most), and based on the input data, performs an operation to estimate the degree of preference (preference information) of the user for each of a plurality of types of fashion brands, and is trained to output the degree of preference of the user for the fashion brand, similar to the output data of the second learning model M2. The third learning model M3 can be configured using algorithms such as multiple regression analysis, SVM, random forest, decision tree, XGBoost, LightGBM, neural network, deep learning, etc., or may be configured by combining a plurality of algorithms such as combining the first learning model M1 and the second learning model M2 of Embodiments 1 to 2.

[0057] The third learning model M3 includes an input layer to which questionnaire responses regarding music content are input, similar to the first learning model M1, an output layer that outputs the degree of preference for each fashion brand, similar to the second learning model M2, and an intermediate layer that calculates output values based on the input data. With such a configuration, when questionnaire responses regarding a piece of music are input, the third learning model M3 outputs, for each fashion brand, the degree of confidence in the associated degree of preference from each output node. When using such a third learning model M3, the information processing apparatus 10 acquires, for each fashion brand, the degree of preference (for example, any one of 1 to 7) associated with the output node that outputs the maximum output value (degree of confidence) as the degree of preference for the fashion brand.

[0058] The third learning model M3 is generated by performing machine learning using training data that associates questionnaire responses regarding training music with the degree of preference (correct degree of preference) for fashion brands by the questionnaire respondents. The training data for the third learning model M3 is generated by associating questionnaire responses regarding music content obtained via a questionnaire (numerical values from 1 to 5 for each questionnaire item in the example of FIG. 4) with the degree of preference for fashion brands obtained via a questionnaire (numerical values from 1 to 7 in the example of FIG. 4). Note that the degree of preference used in the training data may be the degree of preference for each fashion brand estimated from the questionnaire responses using the first learning model M1 and the second learning model M2 of Embodiment 1. The third learning model M3 learns so that when a questionnaire response included in the training data is input, the correct degree of preference for each fashion brand is output from the output node corresponding to each fashion brand. In the learning process, the third learning model M3 performs calculations based on the input questionnaire response and calculates the output value from each output node. Then, for each fashion brand, the third learning model M3 compares the calculated output value of each output node with the value corresponding to the correct degree of preference (1 for the output node corresponding to the correct degree of preference and 0 for other output nodes), and optimizes the parameters used in the calculation process so that the two are approximated. Here too, parameters such as the weights between nodes in the third learning model M3 are optimized using the error backpropagation method, the steepest descent method, etc. As a result, when a questionnaire response regarding a user's music is input, a third learning model M3 that outputs the degree of preference of the user for fashion brands is obtained. The learning of the third learning model M3 may also be performed by the information processing apparatus 10 or may be performed by another learning apparatus.

[0059] The information processing apparatus 10 of the present embodiment configured as described above is capable of executing the same processing as that shown in FIG. 6. In the information processing apparatus 10 of the present embodiment, the control unit 11 does not perform the processing of step S22, and estimates the degree of preference for each fashion brand of the questionnaire respondent based on the questionnaire response regarding the music acquired in step S21 (S23). Here, the control unit 11 inputs the questionnaire response into the third learning model M3, and acquires, as the degree of preference for the fashion brand associated with each output node, the output value from each output node of the third learning model M3. Specifically, the control unit 11 estimates, as the degree of preference for the fashion brand, the degree of preference (for example, any one of 1 to 7) associated with the output node that outputs the maximum output value (confidence) for each fashion brand.

[0060] By the above-described processing, also in the present embodiment, the degree of preference for the fashion brand of the user can be estimated from the questionnaire response regarding the music content most preferred by the user. Therefore, the user can estimate the degree of preference for the fashion brand that matches the user's sensibility (preference) only by answering the questionnaire regarding the selected music (for example, favorite music).

[0061] The configuration of the present embodiment is applicable to the information processing apparatuses 10 of the above-described Embodiments 1 to 2, and the same effects can be obtained even when applied to the information processing apparatuses 10 of Embodiments 1 to 2. Also in the present embodiment, the application of the modification examples appropriately described in the above-described Embodiments 1 to 2 is possible.

[0062] (Embodiment 4) A modification example of the third learning model M3 of Embodiment 3 will be described. The information processing apparatus of the present embodiment has the same configuration as the information processing apparatus 10 of Embodiment 3, except that the storage unit 12 stores a fourth learning model instead of the third learning model M3. The information processing apparatus 10 of the present embodiment stores in the storage unit 12 a DB (hereinafter referred to as a training data use DB) in which data that can be used for training data for learning the fourth learning model is stored.

[0063] FIG. 10 is an explanatory diagram showing an example of a record layout of a training data DB. The training data DB shown in FIG. 10 includes a data ID column, an artist column, a music column, a music questionnaire answer column, a sentiment factor column, a brand questionnaire answer column, etc., and stores information about each user in association with a data ID assigned to the data collected from each user. The artist column, the music column, the music questionnaire answer column, and the brand questionnaire answer column store, for example, the artist name of the favorite music, the music name, the answers to each questionnaire item regarding the music (for example, numerical values from 1 to 5), and the questionnaire answers regarding the preference for each fashion brand (for example, numerical values from 1 to 7) that the user answered via the questionnaire shown in FIG. 4. The sentiment factor column stores numerical values for each sentiment factor calculated by a predetermined calculation based on the questionnaire answers regarding the music, or numerical values for each sentiment factor estimated using the first learning model M1 in Embodiment 1. Note that the brand questionnaire answer column may store, in addition to the questionnaire answers regarding the preference for each fashion brand, the preference for each fashion brand estimated for each user using the second learning model M2 in Embodiment 1. Also, when estimating the preference for each fashion brand using the second learning model M2, the numerical values of each sentiment factor input to the second learning model M2 may be numerical values for each sentiment factor calculated by a predetermined calculation set in advance based on the questionnaire answers regarding the music, or may be numerical values for each sentiment factor estimated from the questionnaire answers regarding the music using the first learning model M1.

[0064] FIG. 11 is an explanatory diagram showing a configuration example of the fourth learning model M4. In the third learning model M3, the fourth learning model M4 uses content data regarding music content selected (designated) by the user as input data instead of the questionnaire responses regarding music content. The content data includes, for example, the song names and artist names of music content that the user likes or music content with a high listening frequency. The fourth learning model M4 performs an operation to estimate the degree of preference (preference information) of the user for each of a plurality of types of fashion brands based on the input content data, and is trained to output the degree of preference of the user for fashion brands. The fourth learning model M4 can be configured using algorithms such as multiple regression analysis, SVM, random forest, decision tree, XGBoost, LightGBM, neural network, and deep learning, or may be configured by combining multiple algorithms.

[0065] The fourth learning model M4 includes an input layer into which content data is input, an output layer that outputs the degree of preference for each fashion brand, and an intermediate layer that calculates an output value based on the input data. The input layer of the fourth learning model M4 has, for example, two input nodes, and through each input node, the song name and artist name of the music content preferred by the user or the music content with a high listening frequency are input. The input layer of the fourth learning model M4 may be configured such that either the song name or the artist name is input. The intermediate layer of the fourth learning model M4 extracts the features of the input data using various functions, thresholds, etc., and outputs the extracted feature values to the output layer. The output layer of the fourth learning model M4 has a plurality of output nodes, and each output node is associated with a preset fashion brand. From each output node, the degree of preference (for example, any value from 1 to 7) for the associated fashion brand is output. Specifically, the output layer of the fourth learning model M4 has, for each fashion brand, for example, a softmax function (not shown) and a plurality (here, seven) of output nodes associated with each numerical value (for example, 1 to 7) that can be selected as the degree of preference, and from each output node, the confidence level for the associated degree of preference is output. For each fashion brand, an output value from 0 to 1 is output from each output node, and the sum of the output values from each output node is 1 (100%). In the fourth learning model M4 with such a configuration, for each fashion brand, the degree of preference (for example, any value from 1 to 7) associated with the output node that outputs the maximum output value (confidence level) can be estimated as the degree of preference for the fashion brand. By using the fourth learning model M4 with such a configuration, when the information processing apparatus 10 inputs the content data of the music content selected by the user, the information processing apparatus 10 can specify the degree of preference of the user for each fashion brand based on each output value output from the fourth learning model M4. Note that the fourth learning model M4 may have a configuration in which, instead of having a plurality of output nodes that output the confidence level for each degree of preference for each fashion brand, it has one output node that outputs the degree of preference with the highest confidence level.

[0066] The fourth learning model M4 is generated by performing machine learning using training data that associates content data related to training music content with the degree of preference (correct degree of preference) for each fashion brand of the user who selected the music content. The training data of the fourth learning model M4 is generated using the data stored in the training data DB of the storage unit 12. Specifically, the training data of the fourth learning model M4 is generated by associating the artist name and song name in each data stored in the training data DB with the questionnaire answers for fashion brands. In the learning process, when the artist name and song name (content data) included in the training data are input, the fourth learning model M4 performs operations based on the input data and calculates the output values from each output node. Then, for each fashion brand, the fourth learning model M4 compares the calculated output values of each output node with the values corresponding to the correct degree of preference (1 for the output node corresponding to the correct degree of preference and 0 for other output nodes), and optimizes the parameters used in the arithmetic processing so that the two are approximated. Here too, parameters such as the weights between nodes in the fourth learning model M4 are optimized using the error backpropagation method, the steepest descent method, etc. As a result, when content data of a song preferred by a certain user is input, the fourth learning model M4 that outputs the confidence level for each degree of preference from the output nodes of each fashion brand is obtained. The learning of the fourth learning model M4 may also be performed by the information processing apparatus 10 or may be performed by another learning apparatus.

[0067] The content data, which is the input data of the fourth learning model M4, may include, in addition to the song title and artist name of the music content preferred by the user, information such as the tempo, rhythm, pitch, and melody line of the music content. Further, it may include the result of rearranging a plurality of songs, artists, or melody lines that affect sensibility in the order of the user's preference. Also, the input data of the fourth learning model M4 may be the song information of the songs (music content) registered in the playlist, playback history, or favorite registration used in the music distribution service. The fourth learning model M4 with such a configuration is obtained by learning using the content data stored in the training data use DB and the training data based on the questionnaire responses for the corresponding fashion brands by storing the content data as described above in the training data use DB.

[0068] When using the fourth learning model M4 with the above-described configuration, the information processing apparatus 10 can execute the same processing as the processing shown in FIG. 6. In the information processing apparatus 10 of the present embodiment, the control unit 11 does not perform the processing of step S22, and estimates the degree of preference for each fashion brand of the user based on the artist name and song title of the song preferred by the user among the questionnaire responses regarding the songs acquired in step S21 (S23). Here, the control unit 11 inputs the artist name and song title (content data) of the song preferred by the user into the fourth learning model M4, and acquires the degree of preference for each fashion brand based on the output value from each output node of the fourth learning model M4. Specifically, for each fashion brand, the control unit 11 identifies the output node that outputs the maximum output value among the output values from each output node, and specifies the degree of preference associated with the identified output value as the degree of preference for the fashion brand.

[0069] By the above-described process, when using the fourth learning model M4, it is possible to estimate the degree of preference of the user for a fashion brand from the artist name and song name of the songs preferred by the user. In this embodiment, in step S21, it is not necessary to obtain an answer to a questionnaire as shown in FIG. 2, and it is sufficient to obtain only the artist name and song name of the songs preferred by the user. Also, the artist name and song name of the songs preferred by the user may be extracted from, for example, a playlist, a playback history, or a favorite registration of a music distribution service. In this case, the user can estimate the degree of preference for a fashion brand that matches the user's sensibility (taste) simply by using a music distribution service. Also in this embodiment, the modified examples appropriately described in the above-described Embodiments 1 to 3 can be applied.

[0070] (Embodiment 5) In the above-described Embodiments 1 to 4, the degree of preference of the user for a fashion brand is estimated from the user's questionnaire answers regarding music or the information of the songs selected by the user. Therefore, it is possible to provide recommendation information for products or services in consideration of the degree of preference of the user for a fashion brand. In addition to this, in this embodiment, an information processing apparatus that can compare the sensibilities (each numerical value of the sensibility factors) of users with a high degree of preference for each of a plurality of fashion brands will be described. Since the information processing apparatus of this embodiment has the same configuration as the configuration of the information processing apparatus 10 of Embodiment 1 shown in FIG. 1, the description of the configuration will be omitted.

[0071] FIG. 12 is a flowchart showing an example of a procedure for comparing the sensibilities of users with a high preference for each fashion brand, and FIG. 13 is an explanatory diagram showing an example of a screen. The information processing apparatus 10 according to the present embodiment can execute the same processing as the processing shown in FIG. 5. In step S14 in FIG. 5, numerical values for each sensibility factor representing the sensibility of the questionnaire respondent are associated with the preference of the questionnaire respondent for the fashion brand and stored in the storage unit 12. In the present embodiment, using the numerical values for each sensibility factor stored in the storage unit 12 in this way and the preference of the questionnaire respondent for the fashion brand (that is, the training data for the second learning model M2), the sensibilities of users with a high preference for each fashion brand are compared. When the information processing apparatus 10 stores the training data DB shown in FIG. 10 in the storage unit 12, the numerical values for each sensibility factor stored in the training data DB and the questionnaire responses for the fashion brand may be used.

[0072] The control unit 11 of the information processing apparatus 10 selects one of the fashion brands (S41), and acquires the numerical values of each sensitivity factor of the questionnaire respondents with a high preference for the selected fashion brand (S42). Here, the control unit 11 extracts the questionnaire respondents with the highest preference for the selected fashion brand among the plurality of fashion brands, and reads out the numerical values of each sensitivity factor of the extracted questionnaire respondents from the storage unit 12. Then, the control unit 11 calculates the average value of the numerical values of each sensitivity factor based on the numerical values of each sensitivity factor of the read questionnaire respondents (S43). Note that the control unit 11 stores the average value of the numerical values of each sensitivity factor in the storage unit 12 in association with the information of the fashion brand. The control unit 11 determines whether there is an unprocessed fashion brand for which the above-described processing has not been performed (S44). If it is determined that there is an unprocessed fashion brand (S44: YES), the control unit 11 returns to the process of step S41, selects one of the unprocessed fashion brands (S41), and performs the processes of steps S42 to S43 for the selected fashion brand. The control unit 11 repeats the processes of steps S41 to S44 until it is determined that there is no unprocessed fashion brand. Thereby, for each fashion brand, the average value of the numerical values of each sensitivity factor in the questionnaire respondents with the highest preference can be obtained.

[0073] When the control unit 11 determines that there is no unprocessed fashion brand (S44: NO), it displays the average value of the numerical values of the respective sensitivity factors corresponding to each fashion brand calculated in step S43 and stored in the storage unit 12 for each fashion brand (S45). For example, the control unit 11 generates a screen as shown in FIG. 13A and displays it on the display unit 15. The screen of FIG. 13A displays the numerical values (average values of the numerical values of each sensitivity factor) of the 1st to 15th sensitivity factors for each fashion brand. Thereby, for each fashion brand, the average sensitivity (numerical values of each sensitivity factor) of users with a high preference for the fashion brand can be displayed in a graph. From the graph shown in FIG. 13A, the sensitivities (numerical values of each sensitivity factor) of users who like each fashion brand can be compared. For example, in the example of FIG. 13A, it can be seen that the sensitivities of users who like the 1st brand and the 3rd brand are similar. Also, it can be seen that the sensitivity of users who like the 4th brand is not similar to the sensitivities of users who like the 1st to 3rd brands.

[0074] Next, the control unit 11 selects two of the fashion brands (S46), and acquires the numerical values of each sensitivity factor of questionnaire respondents with a high preference for the two selected fashion brands (S47). Also here, the control unit 11 extracts the questionnaire respondent with the highest preference for each of the two selected fashion brands, reads out the numerical values of each sensitivity factor of the extracted questionnaire respondents from the storage unit 12, and calculates the average value of the numerical values of each sensitivity factor. Then, the control unit 11 calculates the similarity of the average values of the numerical values of each sensitivity factor between the two fashion brands (S48). As the similarity, for example, a correlation coefficient or a cosine similarity can be used, but other similarities may also be used. The control unit 11 stores the similarity of the average value of each sensitivity factor in the storage unit 12 in association with the pair of fashion brands.

[0075] The control unit 11 determines whether there is an unprocessed pair of fashion brands (a pair of two fashion brands) for which the calculation process of the similarity of the perceptual factors has not been performed (S49). If it is determined that there is an unprocessed pair of fashion brands (S49: YES), the process returns to step S46. Then, the control unit 11 selects an unprocessed pair of fashion brands (S46), and performs the processes of steps S47 to S48 for the selected pair of fashion brands. The control unit 11 repeats the processes of steps S46 to S49 until it determines that there is no unprocessed pair of fashion brands. Thereby, for each pair of fashion brands, the similarity of the perceptual factors (similarity of perception) in the questionnaire respondents with the highest preference for each fashion brand can be obtained.

[0076] If the control unit 11 determines that there is no unprocessed pair of fashion brands (S49: NO), it displays the similarity of the perceptual factors corresponding to each pair of fashion brands calculated in step S48 and stored in the storage unit 12 for each pair (S50). For example, the control unit 11 generates and displays on the display unit 15 a screen as shown in FIG. 13B. The screen in FIG. 13B displays the similarity of the perceptual factors for each pair of fashion brands. Thereby, for each pair of fashion brands, the similarity of the perception (numerical values of each perceptual factor) of the users with high preference for each fashion brand can be listed and displayed. From the table shown in FIG. 13B, the similarity of the perception (numerical values of each perceptual factor) of the users who like each fashion brand can be grasped.

[0077] By the above-described processing, in the present embodiment, it is possible to grasp the similarity of the sensibilities of users who prefer each fashion brand (users with a high degree of preference for each fashion brand). Therefore, for example, when it is possible to estimate a fashion brand with a high degree of preference of a user from the questionnaire answers of the user for music, in addition to the estimated fashion brand, other fashion brands that the user is likely to prefer can be estimated from the similarity of the sensibilities of users who prefer each fashion brand. Therefore, not only fashion brands with a high degree of preference of the user but also other fashion brands that the user is likely to prefer can be estimated and used for marketing information.

[0078] The configuration of the present embodiment is applicable to the information processing apparatus 10 of the above-described Embodiments 1 to 4, and the same effects can be obtained even when applied to the information processing apparatus 10 of Embodiments 1 to 4. Also, in the present embodiment as well, it is possible to apply the modification examples appropriately described in the above-described Embodiments 1 to 4.

[0079] The embodiments disclosed this time should be considered to be illustrative in all respects and not restrictive. The scope of the present invention is shown not by the above meaning but by the scope of the claims, and it is intended that all modifications within the meaning and scope equivalent to the scope of the claims are included.

[0080] The matters described in each of the above-described embodiments can be combined with each other. Also, the independent claims and dependent claims described in the scope of the claims can be combined with each other in all possible combinations regardless of the citation form. Further, although the scope of the claims uses a form (multi-claim form) of describing a claim that cites two or more other claims, it is not limited to this. It may be described using a form of describing a multi-claim (multi-multi-claim) that cites at least one multi-claim.

Description of Reference Numerals

[0081] 10 Information processing apparatus 11 Control unit 12 Memory unit 13 Communication unit 14 Input unit 15 Display unit M1 First learning model M2 Second learning model M3 Third learning model M4 Fourth learning model

Claims

1. A program that causes a computer to execute a process of acquiring content data related to content selected by a user and outputting preference information regarding the user's preference for products or services based on the acquired content data.

2. The acquired content data is input to a first learning model trained to output values for each of a plurality of types of sensitivity factors representing the user's sensitivity when the content data is input, and values for each of the plurality of types of sensitivity factors are obtained. Then, the values for each of the plurality of types of sensitivity factors obtained from the first learning model are input to a second learning model trained to output preference information regarding the preference for products or services when the values for each of the plurality of types of sensitivity factors are input, and the preference information is obtained. The program according to claim 1, which causes the computer to execute the above process.

3. The acquired content data is input to a third learning model trained to output preference information regarding the preference for products or services based on the sensitivity of the user who selected the content when the content data is input, and the preference information is obtained. The program according to claim 1, which causes the computer to execute the above process.

4. The content data includes information related to the content or artist selected by the user. The program according to any one of claims 1 to 3.

5. The content data includes information related to the content or artist included in the playlist or viewing history provided by the content streaming service. The program according to any one of claims 1 to 3.

6. The preference information includes the degree of preference for a plurality of brands. The program according to any one of claims 1 to 3.

7. The content is music content selected by the user. When content data related to the music content is input, the acquired content data is input to a fourth learning model trained to output the degree of preference for a plurality of brands based on the sensitivity of the user who selected the music content, and the degree of preference for the plurality of brands is obtained. The program according to claim 1, which causes the computer to execute the above process.

8. ​ ​ ​ ​ For each of the plurality of brands, obtain values for each of a plurality of sentiment factors representing the sentiment of users who prefer each brand. For each of the plurality of brands, output values for the plurality of sentiment factors. The program according to claim 6, wherein the computer executes the process.

9. For each of the plurality of brands, obtain values for each of a plurality of sentiment factors representing the sentiment of users who prefer each brand. Calculate the similarity of values for the plurality of sentiment factors in two of the plurality of brands. Output the similarity between the two brands. The program according to claim 6, wherein the computer executes the process.

10. Obtain content data regarding content selected by a user. Based on the obtained content data, output preference information regarding the preference of the user for products or services. An information processing method in which a computer executes the process.

11. An information processing apparatus having a control unit, wherein the control unit obtains content data regarding content selected by a user. Based on the obtained content data, outputs preference information regarding the preference of the user for products or services. Information processing apparatus.

Images