Recommendation server

Abstract

The present invention provides a recommendation distribution server capable of recommending content which is appropriate while being unexpected for a user.　A recommendation distribution server 100 comprises a personality information DB 107 for storing user personality information. Then, a recommendation generation unit 104 generates recommendation information including content to be recommended on the basis of the user personality information. A recommendation transmission unit 105 transmits the recommendation information to a user terminal 200. Thus, the recommendation of the content based on the user personality information is made possible, and the user can receive unexpected or serendipitous recommendation of the content.

Description

Technical Field

[0001] The present invention relates to a recommendation distribution server for recommending content and the like.

Background Art

[0002] Patent Document 1 describes specifying video content to be recommended to one customer according to the attribute of the one customer, based on acquired program recommendation information and customer information.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of Invention

Problems to be Solved by the Invention

[0004] However, recommendations based on attribute information may not be able to provide appropriate recommendations in some cases. For example, in Patent Document 1, attributes include gender, age, and favorite content type. There are cases where appropriate recommendations cannot be made based only on gender and age. While recommendations based on the content type a user likes can be expected to be appropriate, the user has to register that content type, which is troublesome. In addition, unexpected or serendipitous video recommendations for the user cannot be expected.

[0005] Accordingly, an object of the present invention is to provide a recommendation distribution server that can make appropriate content recommendations that are unexpected for the user.

Means for Solving the Problems

[0006] The recommendation distribution server of the present invention comprises a personality information storage unit that stores user personality information, a recommendation generation unit that generates recommendation information including recommended content based on the user personality information, and a recommendation transmission unit that transmits the recommendation information to the user. [Effects of the Invention]

[0007] According to the present invention, it is possible to recommend content that is both unexpected and appropriate to the user. [Brief explanation of the drawing]

[0008] [Figure 1] This figure shows the system configuration including the recommendation delivery server 100 in this disclosure. [Figure 2] This diagram shows the functional configuration of the recommendation delivery server 100. [Figure 3] This flowchart shows the operation of the recommendation delivery server 100 described herein. [Figure 4] This flowchart shows the operation of the recommendation server 100 when the number of viewers exceeds a predetermined number. [Figure 5] This figure shows a specific example of attribute information DB108 in this disclosure. [Figure 6] This figure shows a specific example of the personality information DB107. [Figure 7] This figure shows a specific example of the viewing history DB109. [Figure 8] This figure shows a specific example of Content DB110. [Figure 9] This figure shows the probability distribution of the number of viewers during periods t and t-1. [Figure 10] This is a schematic diagram illustrating the weighted averaging process performed by the viewing probability estimation unit 103. [Figure 11] This is training data that associates viewing results (whether or not the content was viewed) with actual viewing results. [Figure 12] This figure shows the relationship between the recommendation accuracy of the recommendation server 100 described herein and the number of views of that content. [Figure 13] This figure shows an example of the hardware configuration of a recommendation delivery server 100 according to one embodiment of the present disclosure. [Modes for carrying out the invention]

[0009] Embodiments of this disclosure will be described with reference to the attached drawings. Where possible, the same parts will be denoted by the same reference numerals, and redundant descriptions will be omitted.

[0010] Figure 1 shows the system configuration including the recommendation distribution server 100 in this disclosure. As shown in the figure, the recommendation distribution server 100 is a server that distributes content such as videos to users, and prior to or separately from such distribution, it is a server that recommends content that is suited to the user's personality. The user terminal 200 is a terminal owned by the user, and by receiving and playing the content, the user can view the content.

[0011] Figure 2 shows the functional configuration of the recommendation distribution server 100. As shown in the figure, it is composed of a recommendation accuracy determination unit 101, an estimation model selection unit 102, a viewing probability estimation unit 103, an estimation model 103a, an estimation model 103b, a recommendation generation unit 104, a recommendation transmission unit 105, a viewing information acquisition unit 106, a personality information DB 107, an attribute information DB 108, a viewing history DB 109, a content DB 110, and a content distribution unit 111.

[0012] The recommendation accuracy determination unit 101 calculates the recommendation accuracy based on the viewing history information stored in the viewing history DB 109. The recommendation accuracy determination unit 101 calculates the accuracy of the recommendation by comparing the number of views (or viewing probability) of the recommended content over a certain period with a subsequent period. Further details will be described later.

[0013] The estimation model selection unit 102 is a section that selects the estimation model 103a and / or the estimation model 103b to be used based on recommendation accuracy and the number of views. Specifically, when the number of viewing histories is less than a predetermined value, the estimation model selection unit 102 selects the estimation model 103a that estimates the number of views for each content based on personality. When the number of viewing histories is equal to or greater than a predetermined value, the estimation model selection unit 102 selects the estimation model 103b that estimates the number of views for each content based on personality and viewing histories.

[0014] Note that the estimation model selection unit 102 may select both the estimation model 103a and the estimation model 103b based on the recommendation accuracy calculated by the recommendation accuracy determination unit 101. For example, when the recommendation accuracy is equal to or lower than a predetermined value, or when the recommendation accuracy has decreased from the previous period, both the estimation model 103a and the estimation model 103b are selected. This detailed processing will be described later.

[0015] The viewing probability estimation unit 103 is a section that acquires (estimates) the viewing probability for each content by inputting attribute information or personality information stored in the attribute information DB 108 or the personality information DB 107 into the selected estimation model and obtaining the output. The viewing probability estimation unit 103 is also a section that acquires (estimates) the viewing probability for each content by inputting viewing history information or the like stored in the viewing history DB 109 into the estimation model 103b.

[0016] The estimation model 103a is an estimation model that receives information from the viewing probability estimation unit 103 and outputs an estimated value in response thereto. Although the estimation model 103a is an estimation model prepared for each content, it may be configured to output estimated values (viewing probabilities) for a plurality of contents from a single estimation model. Then, when the user's personality information, user information, and content information are input, the estimation model 103a outputs a viewing probability that is the probability that the user views the content.

[0017] Estimation model 103b is an estimation model that takes information from the viewing probability estimation unit 103 as input and outputs an estimated value. Similar to the above, estimation model 103b is an estimation model prepared for each piece of content, but it may also be configured to output estimated values ​​(viewing probabilities) for multiple pieces of content from a single estimation model. Furthermore, when estimation model 103b takes user personality information, user information, content information, and viewing information as input, it outputs a viewing probability, which is the probability that the user will watch that content. This viewing information is information that indicates the content and genre that the user has watched before. If the content is a video, it may also include information about the performers.

[0018] These estimation models 103a and 103b may be prediction equations expressed by pre-prepared regression equations, or they may be machine learning models learned through machine learning.

[0019] Machine learning is performed on the estimated model 103a using user personality information, user information, and content information as explanatory variables, and whether or not the user has viewed the content as the dependent variable.

[0020] Machine learning for the estimated model 103b is performed using user personality information, user information, content information, and viewing information as explanatory variables, and whether or not the user viewed the content as the dependent variable.

[0021] These learning processes are carried out using pre-prepared training data by a learning device (not shown in the diagram), as will be described later.

[0022] The recommendation generation unit 104 selects content to be recommended based on the viewing probability estimated by the viewing probability estimation unit 103 and generates recommendations for that content. For example, the recommendation generation unit 104 decides to recommend content with a high viewing probability.

[0023] The recommendation transmission unit 105 is the part that transmits the recommendations generated by the recommendation generation unit 104 to the user terminal 200.

[0024] The viewing information acquisition unit 106 is responsible for acquiring viewing information from the user terminal 200 indicating whether or not the user has viewed the recommended content. This viewing information is generated on the user terminal 200 when the user has viewed the recommended content and is then sent to the recommendation distribution server 100.

[0025] The attribute information DB108 is a storage unit that stores user attribute information. In this disclosure, attribute information is referred to as user information. Attribute information (user information) includes age, gender, place of residence, occupation, etc. It also includes subscription services to content distribution services (a function of the recommendation distribution server 100). These subscription services include types of optional contracts related to content distribution services, such as unlimited content viewing contracts.

[0026] Figure 5 shows a specific example of the attribute information DB108 in this disclosure. As shown in the figure, for each user ID, attribute information such as age, gender, place of residence, and subscribed services is stored. In this disclosure, this information is referred to as attribute information, but it is not limited to this information, and other information may also be included. Furthermore, this information is not mandatory, and some of it may be omitted.

[0027] The personality information DB107 is the part that stores personality information that describes the user's personality. Personality information refers to the user's internal characteristics, such as extroversion, agreeableness, conscientiousness, neuroticism, and openness. Viewing tendencies for content are determined according to this personality information. This information is collected and registered in advance through questionnaires given by the user.

[0028] Figure 6 shows a specific example of the personality information DB107. As shown in the figure, the personality information DB107 includes the five so-called Big Five personality traits for each user ID: agreeableness, conscientiousness, extraversion, neuroticism, and openness. Furthermore, information such as the time discount rate, which indicates how much a person's future value is discounted to its present value, may also be included in the personality information. Of course, the information that represents a person's personality does not have to be limited to what is shown in Figure 6. Other information may also be included.

[0029] The viewing history DB 109 is the part that stores the viewing information acquired by the viewing information acquisition unit 106. This viewing history DB 109 associates user ID, content, whether or not it was recommended, and whether or not the content was viewed.

[0030] Figure 7 shows a specific example of the viewing history DB109. As shown in the figure, the viewing history DB109 includes information for each user ID, such as whether or not the user viewed content based on recommendations, the recommended content / genre, the recommendation interval, and the recommendation date. Not all of this information is necessarily required. Furthermore, other information may also be included.

[0031] Content DB110 is the part that stores the content to be distributed. In this disclosure, the content is video, but it may be other types of information. For example, comics consisting of still images, music that does not contain video, etc. It may also be applications such as games.

[0032] Figure 8 shows a specific example of the content DB 110. As shown in the figure, it includes content ID, genre, title, cast, synopsis, and content data. The content distribution unit 111 distributes the content data stored in this content DB 110 to the user terminal 200.

[0033] The content distribution unit 111 is the part that distributes content stored in the content DB 110 according to user requests. The recommendation server 100 configured in this way can recommend content tailored to the user based on the user's personality information or viewing history, and then deliver that content.

[0034] Next, the process for calculating recommendation accuracy by the recommendation accuracy determination unit 101 will be explained. Figure 9 shows the probability distribution of the number of views in period t and period t-1. For example, if we assume that this probability distribution is a binomial distribution, the shape of the distribution can be determined based on the number of recommendations n and the viewing probability p (or number of views) for a certain period. The viewing probability p is determined based on the number of recommendations and the number of views. Here, we assume that the probability distribution is a binomial distribution, but of course, it is not limited to this. Other forms of probability distributions may also be assumed.

[0035] Here, let n[t] be the number of recommendations a user has for a particular piece of content over a certain period t, and let p[t] be the viewing probability, which is the percentage of users who viewed that particular recommendation.

[0036] Now consider the following situation.

[0037] (1) The probability of viewership decreases compared to the previous period. (2) The probability distribution is significantly different from that of the previous period. When each of the above conditions (1) and (2) is met, the recommendation accuracy determination unit 101 selects a different model, determining that the recommendation accuracy has decreased. In this disclosure, in addition to the estimation model 103b, the viewing probability is determined using the estimation model 103a.

[0038] More details are as follows:

[0039] As described in (1) above, the fact that the viewing probability in period t is lower than that of the previous period t-1 means that p[t] <p[t-1]がなりたつ。

[0040] As described in (2) above, if the probability distribution of period t is significantly different from the probability distribution of the previous period t-1, then the Kullback-Leibler divergence is above a certain threshold.

[0041] The Kullback-Leibler divergence can be calculated using the following formula.

[0042]

number

[0043] Alternatively, the recommendation accuracy determination unit 1031 may calculate the recommendation accuracy based on the trend of that ratio, simply by calculating the ratio of the number of views to the number of recommendations over a certain period, without calculating the Kullback-Leibler information as described above.

[0044] The operation of the recommendation distribution server 100 configured in this way will now be described. Figure 3 is a flowchart showing the operation of the recommendation distribution server 100 of this disclosure. When the estimation model selection unit 102 receives a recommendation distribution instruction from the user terminal 200 (S100), or at a predetermined timing, it refers to the viewing history DB 109 and determines whether the number of viewing history entries by a user who is the target of the recommendation has exceeded a predetermined value (S101). The user who is the target of the recommendation may be determined based on the user terminal 200 that issued the instruction, or it may be determined according to predetermined rules.

[0045] The estimated model selection unit 102 selects estimated model 103a (estimated model A) (S102) if the number of viewing history entries for the above-mentioned user is less than a predetermined value (S101: NO).

[0046] The viewing probability estimation unit 103 obtains the user's personality information from the personality information DB 107, inputs the personality information into the selected estimation model 103a, and obtains the viewing probability for each piece of content as its output (S103). This viewing probability is an estimated value of the viewing probability of the content output by the estimation model 103a.

[0047] The recommendation generation unit 104 generates recommendations for content with a viewing probability of a predetermined value or higher. The recommendation transmission unit 105 then transmits the recommendations to the user terminal 200 of the aforementioned user (S104).

[0048] Subsequently, the viewing information acquisition unit 106 receives viewing status information transmitted from the user terminal 200 when a user has viewed content (S105). Upon receiving the viewing status information, the viewing information acquisition unit 106 updates the viewing history for a user in the viewing history DB 109 (S106).

[0049] In this way, if a user's viewing history is below a predetermined value, recommendations are generated based on their personality information, allowing them to receive recommendations for content that is appropriate, unexpected, and serendipitous.

[0050] Next, we will explain process A (processing when process S101: YES) when the number of viewing history entries exceeds a predetermined number. Figure 4 is a flowchart showing the operation of the recommendation distribution server 100 when the number of viewing history entries exceeds a predetermined number.

[0051] If a user's viewing history exceeds a predetermined value, the recommendation accuracy determination unit 101 calculates the recommendation accuracy (as described above, for example, Kullback-Leibler information) for a certain period t and the preceding period t-1, and determines whether the recommendation accuracy in that period t has decreased based on that recommendation accuracy (S201).

[0052] If the recommendation accuracy determination unit 101 determines that the recommendation accuracy has not decreased compared to the previous period (S201: NO), the estimation model selection unit 102 selects estimation model 103b (estimated model based on viewing history) (S202).

[0053] The viewing probability estimation unit 103 inputs attribute information and viewing history information into the selected estimation model 103b and obtains an estimated viewing probability for each piece of content as its output (S203).

[0054] The recommendation generation unit 104 determines content with a viewing probability of a predetermined value or higher, generates a recommendation for it, and the recommendation transmission unit 105 transmits the recommendation to a user (S204).

[0055] Furthermore, the recommendation accuracy determination unit 101 calculates the recommendation accuracy for a certain period t and the preceding period t-1. If the recommendation accuracy for that period has decreased (S201: YES), the viewing probability estimation unit 103 selects both estimation model 103a and estimation model 103b (S205). The viewing probability estimation unit 103 then inputs the personality information of a user into estimation model 103a (estimation model A) and obtains the viewing probability. The viewing probability estimation unit 103 also inputs the viewing history information of a user into estimation model 103b (estimation model B) and obtains the viewing probability (S206).

[0056] The viewing probability estimation unit 103 performs weight processing using predetermined parameters on each obtained viewing probability to estimate a viewing probability that takes into account personality information and viewing history (S207). In this disclosure, the sum of these predetermined parameters may be 1, or the sum may be less than 1. Initially, the weight parameters are set randomly. Subsequently, weight parameters learned based on these initial parameters are set. Details will be described later.

[0057] The recommendation generation unit 104 generates recommendations for content whose viewing probability, weighted according to predetermined parameters, is equal to or greater than a predetermined value, and the recommendation transmission unit 105 transmits these recommendations (S208).

[0058] Subsequently, the viewing information acquisition unit 106 receives viewing status information transmitted from the user terminal 200 when a user has viewed the content (S105). Upon receiving the viewing status information, the viewing information acquisition unit 106 updates the viewing history for a given user in the viewing history DB 109 (S106). The weight parameters, described later, are then learned using the updated viewing history.

[0059] In this way, if a certain number of viewing histories are accumulated, recommendations based on those viewing histories can be made. On the other hand, if the accuracy of recommendations based on viewing histories deteriorates, recommendations that take personality information into account can be generated, resulting in more effective recommendations.

[0060] Next, estimation models 103a and 103b will be described. Estimation models 103a and 103b are models that estimate the viewing probability for each piece of content. In the above disclosure, an estimation model is provided for each piece of content, and it is assumed that personality information, content information, viewing information, and user information are input into each estimation model to estimate the viewing probability, but this is not the only way. For example, one estimation model may output the viewing probability for each of multiple pieces of content.

[0061] These estimation models may be based on multiple regression analysis or on decision trees. For example, estimation model 103a can be expressed as a multiple regression equation as follows. Note that the user information below consists of attribute information such as gender and age, and the content information consists of metadata such as the genre and title of the content.

[0062] Prediction formula for viewing probability of content A: y_1 = β_(1_1) * personality information + β_(2_1) * user information + β_(3_1) * content information Prediction formula for viewing probability of content B: y_2 = β_(1_2) * personality information + β_(2_2) * user information + β_(3_2) * content information ... Furthermore, the estimated model 103b can be expressed as a multiple regression equation as follows. Here, the prediction equation takes viewing information into account.

[0063] Prediction formula for viewing probability of content A: y_1 = β_(1_1) * personality information + β_(2_1) * user information + β_(3_1) * content information + β_(4_1) * viewing information Prediction formula for viewing probability of content B: y_2 = β_(1_2) * personality information + β_(2_2) * user information + β_(3_2) * content information + β_(4_2) * viewing information User information and content information (metadata) are vectorized using the publicly known FastText. Furthermore, if only one estimation model is used, the above β_(1_1) will be changed for each content. For example, estimation model 103a is as follows. Estimation model 103b is similar. Here, i represents the content.

[0064] Prediction formula for viewing probability of content i: y_i = β_1[i] * personality information + β_2[i] * user information + β_3 * content information These values ​​β_(1_1), etc., may be predetermined values, or appropriate values ​​can be obtained by applying the above formula to training data consisting of whether or not the recommendation was viewed, personality information, user information, and content information (meta information) and solving it using the least squares method. This training data is pre-prepared information.

[0065] Furthermore, these can, of course, be considered as machine learning models.

[0066] In that case, when training the estimation model 103a, the training data will use user personality information, user information, and content information (meta-information) as explanatory variables, and the dependent variable will be whether or not the content was viewed. When training the estimation model 130b, the training data will also include user viewing information as an explanatory variable, and the dependent variable will be whether or not the content was viewed.

[0067] The above-mentioned training data is information pre-stored in the training data storage unit of the learning device, and is based on the viewing history of users, including other users. The learning device (not shown) uses the training data to construct an estimation model using known machine learning methods.

[0068] Next, we will explain a method for performing a weighted average on the output values ​​of estimation models 103a and 103b when the recommendation accuracy decreases. Figure 10 is a schematic diagram showing the weighted average processing by the viewing probability estimation unit 103. As shown in the figure, y represents the viewing probability of a certain content. b_1 is the weight for the viewing probability of estimation model 103a, b_2 is the weight for the viewing probability of estimation model 103b, y_1 is the viewing probability of estimation model 103a, and y_2 is the viewing probability of estimation model 103b. r_user indicates the correction based on user information.

[0069] r_user is as shown in Figure 10(b). x_user is user information representing a vector of demographics (gender, age, etc.). This vectorization process can be performed using, for example, FastText, but is not limited to that; any known vectorization method that captures the meaning is acceptable. b_user represents the weight (regression coefficient). This value should be a predetermined value.

[0070] The weights b_1 and b_2 for viewing probabilities in the estimated models 103a and 103b are initially set to random values. As mentioned above, once a certain amount of training data (number of data points) has been accumulated, appropriate values ​​are learned and these values ​​are treated as weights.

[0071] Figure 11(a) shows the training data, which associates the output value y_1 of the estimated model 103a, the output value y_2 of the estimated model 103b, and the viewing results (whether or not the user watched) for x_user. This training data includes the viewing results when weight parameters are randomly set, the viewing probability is calculated based on these parameters, and recommendations are made.

[0072] Using this training data and the multiple regression equation shown in Figure 10(a), we can find b_1, b_2, and b in Figure 10(a). Here, we extract the data for which viewing was performed, i.e., the data where viewing is indicated by 1, and by solving using the least squares method or similar with only these data, we can find the appropriate b_1, b_2, and b.

[0073] Furthermore, as shown in Figure 10(b), one may also select the combination of b_1, b_2, and b that has the most views, based on the histories of randomly selected b_1, b_2, and b, and whether or not they were viewed. Alternatively, the result may be calculated using statistical methods, such as taking the mean or median of the combinations with the highest viewing frequency.

[0074] Figure 12 shows the relationship between the recommendation accuracy of the recommendation delivery server 100 of this disclosure and the number of content views (viewing history). As shown in the figure, the estimation model is switched when the number of viewing history reaches approximately 20. The estimation model is switched when the number of content views exceeds 20. Although the recommendation accuracy improves initially, if the recommendation accuracy decreases, the settings are configured to switch the weighting for each model.

[0075] Generally, each personality type has its own tendencies regarding preferred genres. Those with high emotional instability tend to prefer investing or gambling. Those with high openness tend to prefer new technologies (works of art), etc. Those with excessively high conformity may try to fit in and put on airs to please others. The recommended content reflects these tendencies.

[0076] Based on the above, it is thought that there are viewing tendencies based on personality. For example, people with low emotional instability (i.e., calm and composed) prefer push notifications for recommendations. Conversely, people with low conscientiousness and high emotional instability do not prefer push notifications. Therefore, it may be possible to select whether recommendations are delivered via push notifications or in-app notifications based on personality.

[0077] Next, the operation and effects of the recommendation distribution server 100 of this disclosure will be described. This recommendation distribution server 100 includes a personality information DB 107 that stores the user's personality information. The recommendation generation unit 104 generates recommendation information, including recommended content, based on the user's personality information. The recommendation transmission unit 105 transmits the recommendation information to the user terminal 200.

[0078] This enables content recommendations based on user personality information, allowing users to receive unexpected and serendipitous content recommendations. Furthermore, it allows for effective recommendations even in the early stages when information such as viewing history has not yet been accumulated.

[0079] For example, hobbies and preferences, one type of attribute information, are influenced by seasonality and other factors, are highly variable, and have a wide variety of variations, making them unreliable for initial application. On the other hand, personality traits are more universal and can be categorized into a few types compared to hobbies and preferences, giving them the advantage of being easier to apply in the early stages.

[0080] Furthermore, recommendations based on personality information are based on personality data rather than attribute information or viewing history, and it can be assumed that such recommendations are more likely to be accepted by that user. The core of a user persona is, in fact, psychological characteristics centered on personality. Therefore, recommendations based on personality information are often more effective than recommendations based on attribute information or viewing history.

[0081] In this disclosure, the viewing probability estimation unit 103 estimates the viewing probability of content based on the user's personality information, and the recommendation generation unit 104 determines the content to recommend based on that viewing probability. In addition, in this disclosure, the viewing probability estimation unit 103 inputs personality information into the estimation model 103a and obtains the estimated viewing probability of content. For example, this estimation model 103a is a learning model learned based on personality information and viewing results, or a multiple regression equation in which regression coefficients are obtained based on that information.

[0082] In the disclosure described above, user information and content information are input to the estimation models 103a and 103b in addition to personality information and viewing information. However, this is not limited to this, and only personality information and viewing information may be input to the estimation model. In that case, the estimation model must be designed or trained accordingly.

[0083] Furthermore, the recommendation delivery server 100 of this disclosure further includes a viewing history DB 109, which is a viewing history storage unit that stores the user's viewing history. The viewing probability estimation unit 103 estimates the viewing probability of content based on the user's viewing history when the number of viewing history entries in the viewing history reaches a predetermined condition. The recommendation generation unit 104 then determines the content to recommend based on the viewing probability.

[0084] As described above, the recommendation delivery server 100 is equipped with an estimation model 103b. This estimation model 103b is a learning model that has been trained based on viewing history and viewing results, or a multiple regression equation whose regression coefficients have been determined based on that information.

[0085] This allows for effective recommendations to be made once a sufficient number of viewing history entries have been accumulated.

[0086] Furthermore, the recommendation distribution server 100 of this disclosure further includes a recommendation accuracy determination unit 101 that determines the recommendation accuracy of the recommendation information. The recommendation generation unit 104 then determines the content to recommend based on the recommendation accuracy, using either personality information or viewing history.

[0087] This enables recommendations based on recommendation accuracy. For example, if recommendations were previously based on viewing history, but at a certain point the recommendation accuracy has decreased, the system can use both personality information and viewing history to determine which content to recommend, allowing for recommendations of that content without compromising accuracy.

[0088] Furthermore, the viewing history DB 109 stores viewing history, including the number of recommendations for content and the number of views for those recommendations. The recommendation accuracy determination unit 101 then calculates the viewing results (for example, the viewing probability for each recommendation) based on the viewing history and determines the recommendation accuracy based on those viewing results.

[0089] Here, the recommendation accuracy determination unit 101 generates a first probability distribution and a second probability distribution that indicate the probability of viewing a content, based on the number of recommendations and the number of views for that content. The first probability distribution represents the probability distribution over a certain period t, and the second probability distribution represents the probability distribution over a period t-1 prior to that period t. The recommendation accuracy determination unit 101 then determines the recommendation accuracy by comparing the first probability distribution and the second probability distribution.

[0090] The recommendation accuracy determination unit 101 determines the recommendation accuracy by determining a decrease in viewing probability based on a first probability distribution and a second probability distribution. Furthermore, the recommendation accuracy determination unit 101 determines the recommendation accuracy by comparing the shapes of the first and second probability distributions. This shape determination is based on the Kullback-Leibler divergence.

[0091] By analyzing the shape, it becomes possible to determine a decrease in the accuracy of recommendations, etc.

[0092] The viewing probability estimation unit 103 estimates viewing probabilities based on content viewing history and viewing probabilities based on personality information using estimation models 103a and 103b. The viewing probability estimation unit 103 then calculates viewing probabilities by weighting each viewing probability with predetermined parameters. The recommendation generation unit 104 determines the content to recommend based on these viewing probabilities.

[0093] These predetermined parameters are determined randomly over a predetermined period, and the content to be recommended is determined based on the viewing probability calculated using these parameters. Viewing results are then obtained for the content to be recommended. Based on the viewing results, the predetermined parameters are learned.

[0094] The recommendation delivery server of the present invention has the following configuration.

[0095] [1] A personality information storage unit that stores the user's personality information, A recommendation generation unit that generates recommendation information including recommended content based on the user's personality information, A recommendation transmission unit that transmits the aforementioned recommendation information to the user, A recommendation server equipped with [features / equipment].

[0096] [2] The system further includes an estimation unit that estimates the probability of viewing the content based on the user's personality information. The recommendation generation unit determines the content to recommend based on the viewing probability. The recommendation server described in [1].

[0097] [3] The system further includes a viewing history storage unit that stores the user's viewing history, When the number of views in the viewing history reaches a predetermined condition, the estimation unit estimates the viewing probability of the content based on the viewing history. The recommendation generation unit determines the content to recommend based on the viewing probability. The recommendation server described in [2].

[0098] [4] The system further includes a determination unit that determines the recommendation accuracy of the aforementioned recommendation information. The recommendation generation unit, Based on the recommendation accuracy, the system determines the content to recommend using either the personality information or the viewing history. The recommendation server described in [3].

[0099] [5] The aforementioned viewing history storage unit stores the viewing history, including the number of recommendations for the content and the number of views for that content. The aforementioned determination unit calculates the viewing result for the number of recommendations based on the viewing history, and determines the recommendation accuracy based on the viewing result. The recommendation server described in [4].

[0100] [6] The determination unit generates a first probability distribution and a second probability distribution indicating the viewing probability for the content, based on the number of recommendations and the number of views for the content. The first probability distribution described above shows the probability distribution over a certain period. The second probability distribution described above represents the probability distribution for a period prior to a certain period, The recommendation accuracy is determined by comparing the first probability distribution and the second probability distribution. The recommendation server described in [5].

[0101] [7] The unit that makes the determination said, The recommendation accuracy is determined by determining a decrease in the probability of viewing based on the first and second probability distributions. The recommendation server described in [6].

[0102] [8] The unit that makes the determination said, The recommendation accuracy is determined by comparing the shapes of the first probability distribution and the second probability distribution. The recommendation server described in [6] or [7].

[0103] [9] The estimation unit, The viewing probability of the content is estimated based on the viewing history and the viewing probability is estimated based on the personality information. For each viewing probability, a viewing probability is calculated by applying weights with predetermined parameters. The recommendation generation unit determines the content to recommend based on the viewing probability. A recommendation server as described in any one of [3] through [8].

[0104]

[10] The aforementioned predetermined parameters are determined randomly over a predetermined period, the viewing probability is calculated based on those parameters, and the content to be recommended is determined. Viewing results are obtained for the content that is to be recommended. Based on the viewing results, the predetermined parameters are learned. The recommendation server described in [9].

[0105] The block diagram used in the description of the above embodiment shows functional units. These functional blocks (components) are realized by any combination of at least one of hardware and software. Furthermore, the method of realizing each functional block is not particularly limited. That is, each functional block may be realized using one device that is physically or logically coupled, or it may be realized using two or more physically or logically separated devices that are directly or indirectly connected (for example, using wired or wireless connections). A functional block may be realized by combining the above one device or the above multiple devices with software.

[0106] Functions include, but are not limited to, judgment, decision, judgment, calculation, calculation, processing, derivation, investigation, exploration, confirmation, reception, transmission, output, access, resolution, selection, selection, establishment, comparison, assumption, expectation, assumption, broadcasting, notifying, communicating, forwarding, configuring, reconfiguring, allocating (mapping), and assigning. For example, a functional block (configuration part) that enables transmission is called a transmitting unit or transmitter. As mentioned above, the method of implementation is not particularly limited.

[0107] For example, the recommendation delivery server 100 in one embodiment of the present disclosure may function as a computer that processes the recommendation delivery method of the present disclosure. Figure 13 is a diagram showing an example of the hardware configuration of the recommendation delivery server 100 according to one embodiment of the present disclosure. The recommendation delivery server 100 described above may be physically configured as a computer device including a processor 1001, memory 1002, storage 1003, communication device 1004, input device 1005, output device 1006, bus 1007, etc.

[0108] In the following explanation, the term "device" can be replaced with "circuit," "device," "unit," etc. The hardware configuration of the recommendation delivery server 100 may include one or more of the devices shown in the diagram, or it may be configured to omit some of the devices.

[0109] Each function in the recommendation distribution server 100 is realized by loading predetermined software (programs) onto hardware such as the processor 1001 and memory 1002, which causes the processor 1001 to perform calculations, control communication by the communication device 1004, and control at least one of the reading and writing of data in the memory 1002 and storage 1003.

[0110] The processor 1001 controls the entire computer, for example, by running the operating system. The processor 1001 may be composed of a central processing unit (CPU) that includes interfaces with peripheral devices, control devices, arithmetic units, registers, etc. For example, the recommendation accuracy determination unit 101, estimation model selection unit 102, and viewing probability estimation unit 103 mentioned above may be implemented by the processor 1001.

[0111] Furthermore, the processor 1001 reads programs (program code), software modules, data, etc., from at least one of the storage 1003 and the communication device 1004 into the memory 1002 and executes various processes accordingly. The program used is one that causes the computer to execute at least a part of the operations described in the above embodiment. For example, the recommendation accuracy determination unit 101 may be implemented by a control program stored in the memory 1002 and running on the processor 1001, and other functional blocks may be implemented similarly. The above-described various processes have been explained as being executed by one processor 1001, but they may be executed simultaneously or sequentially by two or more processors 1001. The processor 1001 may be implemented by one or more chips. The program may also be transmitted from a network via a telecommunications line.

[0112] Memory 1002 is a computer-readable recording medium and may consist of at least one of the following: ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), RAM (Random Access Memory), etc. Memory 1002 may also be called a register, cache, main memory, etc. Memory 1002 can store executable programs (program code), software modules, etc., for implementing the recommendation delivery method according to one embodiment of this disclosure.

[0113] Storage 1003 is a computer-readable recording medium and may consist of at least one of the following: an optical disc such as a CD-ROM (Compact Disc ROM), a hard disk drive, a flexible disk, a magneto-optical disk (e.g., a compact disc, a digital multipurpose disc, a Blu-ray® disc), a smart card, flash memory (e.g., a card, a stick, a key drive), a floppy® disk, a magnetic strip, etc. Storage 1003 may also be called an auxiliary storage device. The above-mentioned storage medium may be, for example, a database, server, or other suitable medium including at least one of memory 1002 and storage 1003.

[0114] The communication device 1004 is hardware (transmitting / receiving device) for communicating between computers via at least one of a wired network and a wireless network, and is also referred to as a network device, network controller, network card, communication module, etc. The communication device 1004 may be configured to include high-frequency switches, duplexers, filters, frequency synthesizers, etc., in order to implement at least one of frequency division duplex (FDD) and time division duplex (TDD). For example, the recommendation transmission unit 105, viewing information acquisition unit 106, content distribution unit 111, etc., described above may be implemented by the communication device 1004. The transmitting and receiving unit may be implemented in a physically or logically separated manner, with the transmitting unit and the receiving unit being separated.

[0115] The input device 1005 is an input device that accepts input from an external source (e.g., a keyboard, mouse, microphone, switch, button, sensor, etc.). The output device 1006 is an output device that outputs to an external source (e.g., a display, speaker, LED lamp, etc.). The input device 1005 and the output device 1006 may be configured as an integrated unit (e.g., a touch panel).

[0116] Furthermore, each device, such as the processor 1001 and memory 1002, is connected by a bus 1007 for communicating information. The bus 1007 may be configured using a single bus, or different buses may be configured for each device.

[0117] Furthermore, the recommendation distribution server 100 may be configured to include hardware such as a microprocessor, a digital signal processor (DSP), an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), or an FPGA (Field Programmable Gate Array), and some or all of each functional block may be realized by such hardware. For example, the processor 1001 may be implemented using at least one of these hardware components.

[0118] Information notification is not limited to the embodiments described herein and may be carried out by other means. For example, information notification may be carried out by physical layer signaling (e.g., DCI (Downlink Control Information), UCI (Uplink Control Information)), upper layer signaling (e.g., RRC (Radio Resource Control) signaling, MAC (Medium Access Control) signaling, broadcast information (MIB (Master Information Block), SIB (System Information Block))), other signals, or combinations thereof. RRC signaling may also be called RRC messages, and may be, for example, RRC Connection Setup messages, RRC Connection Reconfiguration messages, etc.

[0119] The processing procedures, sequences, flowcharts, etc., of each aspect / embodiment described herein may be reordered, provided they are consistent with each other. For example, the methods described herein present various step elements in an exemplary order and are not limited to that specific order.

[0120] Input and output information may be stored in a specific location (e.g., memory) or managed using a management table. Input and output information may be overwritten, updated, or appended to. Output information may be deleted. Input information may be transmitted to other devices.

[0121] The determination may be made by a value represented by 1 bit (0 or 1), by a boolean value (true or false), or by a numerical comparison (for example, a comparison with a predetermined value).

[0122] Each aspect / embodiment described herein may be used individually, in combination, or switched between as needed during implementation. Furthermore, notification of specific information (e.g., notification that "X is") is not limited to explicit notification, but may also be implicit (e.g., by not providing such notification).

[0123] Although the present disclosure has been described in detail above, it will be clear to those skilled in the art that the present disclosure is not limited to the embodiments described herein. The present disclosure can be implemented in modified and altered forms without departing from the intent and scope of the present disclosure as defined by the claims. Accordingly, the descriptions in the present disclosure are illustrative and not intended to be restrictive in any way.

[0124] Software should be broadly interpreted to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executable files, execution threads, procedures, functions, and so on, whether they are called software, firmware, middleware, microcode, hardware description languages, or by any other name.

[0125] Furthermore, software, instructions, information, etc., may be transmitted and received via a transmission medium. For example, if software is transmitted from a website, server, or other remote source using at least one of wired technologies (such as coaxial cable, fiber optic cable, twisted pair, or digital subscriber line (DSL)) and wireless technologies (such as infrared or microwave), then at least one of these wired and wireless technologies is included in the definition of a transmission medium.

[0126] The information, signals, etc. described in this disclosure may be represented using any of the various different techniques. For example, the data, instructions, commands, information, signals, bits, symbols, chips, etc. that may be referred to throughout the above description may be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, optical fields or photons, or any combination thereof.

[0127] In addition, terms used in this disclosure and terms necessary for understanding this disclosure may be replaced with terms having the same or similar meanings. For example, at least one of the channel and symbol may be a signal (signaling). Also, a signal may be a message. Furthermore, a component carrier (CC) may be called a carrier frequency, cell, frequency carrier, etc.

[0128] Furthermore, the information, parameters, etc., described in this disclosure may be expressed using absolute values, relative values ​​from a given value, or other corresponding information. For example, wireless resources may be indicated by an index.

[0129] The names used for the parameters described above are not restrictive in any way. Furthermore, the formulas and other expressions using these parameters may differ from those expressly disclosed in this disclosure. Various channels (e.g., PUCCH, PDCCH, etc.) and information elements can be identified by any suitable name, and therefore, the various names assigned to these various channels and information elements are not restrictive in any way.

[0130] In this disclosure, terms such as "Mobile Station (MS)," "user terminal," "User Equipment (UE)," and "terminal" may be used interchangeably.

[0131] A mobile station may also be referred to by those skilled in the art as a subscriber station, mobile unit, subscriber unit, wireless unit, remote unit, mobile device, wireless device, wireless communication device, remote device, mobile subscriber station, access terminal, mobile terminal, wireless terminal, remote terminal, handset, user agent, mobile client, client, or some other appropriate term.

[0132] As used in this disclosure, the terms “determining” and “determining” may encompass a wide variety of actions. “Determining” may include, for example, judging, calculating, computing, processing, deriving, investigating, looking up, searching, or inquiring (e.g., searching in a table, database, or other data structure), or ascertaining. “Determining” may also include, for example, receiving (e.g., receiving information), transmitting (e.g., sending information), inputting, outputting, or accessing (e.g., accessing data in memory). Furthermore, "judgment" and "decision" can include considering something as having been "judged" or "decided" after resolving, selecting, choosing, establishing, comparing, etc. In other words, "judgment" and "decision" can include considering something as having been "judged" or "decided" after some action. Also, "judgment (decision)" can be reinterpreted as "assuming," "expecting," or "considering."

[0133] The terms “connected,” “coupled,” or any variation thereof, mean any direct or indirect connection or coupling between two or more elements, and may include the presence of one or more intermediate elements between two elements that are “connected” or “coupled” with each other. The coupling or connection between elements may be physical, logical, or a combination thereof. For example, “connection” may be reinterpreted as “access.” As used in this disclosure, two elements may be considered to be “connected” or “coupled” with each other using at least one of one or more wires, cables, and printed electrical connections, and, in some non-limiting and non-exclusive examples, electromagnetic energy having wavelengths in the radio frequency domain, microwave domain, and optical (both visible and invisible) domain.

[0134] In this disclosure, the phrase "based on" does not mean "based solely on" unless otherwise specified. In other words, the phrase "based on" means both "based solely on" and "based at least on."

[0135] Any reference to elements using designations such as “first,” “second,” etc., as used in this disclosure does not generally limit the quantity or order of those elements. These designations may be used in this disclosure as a convenient way to distinguish between two or more elements. Accordingly, references to first and second elements do not imply that only two elements may be adopted, or that the first element must precede the second element in any way.

[0136] Where the terms “include,” “including,” and their variations are used in this disclosure, these terms are intended to be inclusive, as is the term “comprising.” Furthermore, the term “or” as used in this disclosure is not intended to mean exclusive OR.

[0137] In this disclosure, if articles are added by translation, such as a, an, and the in English, this disclosure may include the fact that the noun following these articles is plural.

[0138] In this disclosure, the term "A and B are different" may mean "A and B are different from each other." The term may also mean "A and B are each different from C." Terms such as "separate" and "combine" may be interpreted similarly to "different." [Explanation of symbols]

[0139] 100... Recommendation distribution server, 200... User terminal, 101... Recommendation accuracy determination unit, 102... Estimation model selection unit, 103... Viewing probability estimation unit, 103a... Estimation model, 103b... Estimation model, 104... Recommendation generation unit, 105... Recommendation transmission unit, 106... Viewing information acquisition unit, 107... Personality information DB, 108... Attribute information DB, 109... Viewing history DB, 110... Content DB, 111... Content distribution unit.

Claims

1. A personality information storage unit that stores the user's personality information, A viewing history storage unit that stores the user's viewing history, An estimation unit that estimates the probability of viewing content based on at least one of the aforementioned personality information and the aforementioned viewing history, A recommendation generation unit determines content to be recommended based on the aforementioned viewing probability and generates recommendation information that includes said content, A recommendation transmission unit that transmits the aforementioned recommendation information to the user, A determination unit that determines the recommendation accuracy of the aforementioned recommendation information, Equipped with, The aforementioned viewing history storage unit stores viewing history including the number of recommendations for content and the number of views of said content. The determination unit calculates the viewing result for the number of recommendations based on the viewing history, and determines the recommendation accuracy based on the viewing result. The estimation unit estimates the viewing probability based on the personality information if the number of views in the viewing history does not reach a predetermined condition, and estimates the viewing probability based on the recommendation accuracy, using at least the viewing history, if the number of views reaches a predetermined condition. The recommendation generation unit determines the content to recommend based on the estimated viewing probability. Recommendation server.

2. The determination unit generates a first probability distribution and a second probability distribution indicating the viewing probability for the content based on the number of recommendations and the number of views for the content. The first probability distribution above shows the probability distribution over a certain period. The second probability distribution described above represents the probability distribution in a period prior to a certain period, The recommendation accuracy is determined by comparing the first probability distribution and the second probability distribution. The recommendation server according to claim 1.

3. The unit that makes the determination said, The recommendation accuracy is determined by determining a decrease in viewing probability based on the first and second probability distributions. The recommendation server according to claim 2.

4. The unit that makes the determination said, The recommendation accuracy is determined by comparing the shapes of the first probability distribution and the second probability distribution. The recommendation server according to claim 2.

5. A personality information storage unit that stores user personality information, A viewing history storage unit that stores the user's viewing history, An estimation unit that estimates the probability of viewing content based on at least one of the user's personality information and the viewing history, A recommendation generation unit determines recommended content based on the aforementioned viewing probability and generates recommendation information that includes said content, A recommendation transmission unit that transmits the aforementioned recommendation information to the user, Equipped with, When the number of views in the viewing history reaches a predetermined condition, the estimation unit estimates the viewing probability of the content based on the viewing history and the viewing probability based on the personality information. For each viewing probability, a viewing probability is calculated by applying weights with predetermined parameters. The recommendation generation unit determines the content to recommend based on the viewing probability, The aforementioned predetermined parameters are determined randomly over a predetermined period, the viewing probability is calculated based on those parameters, and the content to be recommended is determined. Viewing results are obtained for the content that is to be recommended. Based on the viewing results, the predetermined parameters are learned. Recommendation server.

Images