JP5408570B2 - Attribute information update method and information update method - Google Patents

Description

  The present invention relates to a system that provides content to a user and evaluates the content and the characteristics of the user using a user's reaction to the content.

  With the development of the Internet, it has become easy for content providers to transmit content created by themselves to an unspecified number of users. On the other hand, since it is difficult for many content providers to charge usage fees individually to users who use content, there are services that perform content management and billing.

  As a representative example of the service described above, Amazon. There is an EC site such as com (registered trademark, the same applies hereinafter) or an official content service of a mobile phone.

  In addition, as one of the content management services described above, the appearance of a content management service that utilizes user electricity usage data is expected. Conventionally, electricity usage and charges are confirmed once a month by the meter reader checking the power meter and inputting the value read from the power meter into the handheld computer for meter reading. Was printed, and the notice slip was posted on a post or newspaper tray.

  However, with the introduction of AMI (Advanced Metering Infrastructure) that automatically transmits electricity usage to the power company using a communication line, it is not necessary to check the meter reader, and detailed usage such as 15 minutes or 30 minutes The amount can be grasped.

  As a result, content such as energy-saving consulting services tailored to users' electricity usage patterns, proposals for optimal power rate plans, and proposals for devices with high replacement effects (power usage-linked advertisements) are provided by electric power companies and electrical equipment manufacturers. It is thought that it will come to be provided from such as.

  One function of the above-described content management service is a function (recommendation function) for recommending content suitable for a specific user from a large number of contents registered in the content management service.

  As a method for realizing the above-mentioned recommendation function, another user group having a similar selection tendency (hereinafter referred to as usage trend information) is specified based on content selected by the user in the past, and the specified user group is Based on the collaborative filtering method that recommends the selected content, information that expresses the characteristics of the content (hereinafter referred to as feature information), and information that represents the user's preferences (hereinafter referred to as preference information) There is a content-based recommendation method that recommends content that matches the user's preference.

  The collaborative filtering method does not require the feature information and the preference information to be determined in advance, but requires a large amount of usage trend information for each user.

  On the other hand, in the content-based recommendation method, it is necessary to prepare feature information and preference information in advance. As a method of acquiring feature information, a method in which the content provider sets the feature information by itself, a method in which the selected monitor (evaluator) uses the content, sets the feature information from the answer result of the questionnaire, and A method of extracting feature information by analyzing the content itself is known. On the other hand, with regard to a preference information setting method, a method is known in which a user selects an object of interest at the time of service use registration.

  However, the feature information set by the content provider or the preference information self-reported by the user does not necessarily reflect the actual situation. Moreover, since the user who uses arbitrary contents and a user's preference tendency change with progress of time, it is necessary to update feature information and preference information sequentially.

  As a method of solving the above-described problem, a method of dynamically correcting feature information and preference information based on a response to content used by a user is known (see, for example, Patent Document 1).

  In the method described in Patent Literature 1, user evaluation information for a recommended video content is estimated from a user's reaction (skip, repeated browsing, etc.) to the recommended video content, and the feature is based on the estimated evaluation information. Information and preference information are corrected sequentially.

  Further, in the method described in Patent Document 1, the degree of sequential correction is adjusted using a reliability that is a parameter indicating whether or not feature information and preference information are appropriate. That is, feature information and preference information with low reliability are positively corrected, and feature information and preference information with high reliability are not corrected unnecessarily.

JP 2004-54768 A

  In the technique described in Patent Literature 1, evaluation information is estimated based on a user's reaction to content recommended based on a user's preference, and feature information and preference information are corrected.

  Therefore, the content presented to the user is biased in the sense that it matches the user's preference. That is, only the content estimated to be of interest to the user is presented. For this reason, when a user negatively evaluates the presented content, it is difficult to estimate what kind of preference the user actually has.

  Further, the invention described in Patent Document 1 has the following problems. One problem is that erroneous correction may be performed when the reliability of either the feature information or the preference information is low. Another problem is that when the reliability of either the feature information or the preference information is low, the update frequency is adjusted based on the reliability, so that the feature information or the preference information is not easily corrected.

  Further, in the content management service for content utilizing the power consumption, there are the following problems when the conventional method is applied.

  That is, in the content management service, there are many users, but since the usage frequency may be lowered once a month, it takes a very long time to correct the feature information and the preference information. In order to solve the above-described problem, a method of requesting content evaluation from all or a large number of users can be considered. However, this method has a problem that the reliability of the evaluation information is lowered although the cost is high.

  An object of the present invention is to collect highly reliable evaluation information (user attribute information and content attribute information) for correcting feature information and preference information at a low cost and in a short time.

  A typical example of the present invention is as follows. That is, a computer used by a user, a content providing server that transmits content to be provided to the user, the content transmitted from the content providing server is selected, and the content to be provided to the user is selected, An attribute information update method in a computer system, comprising: an evaluation system that transmits the selected content to a computer and updates the user attribute information representing the characteristics of the user or the content attribute information representing the characteristics of the content. Includes a first processor, a memory connected to the first processor, and a first network interface connected to the first processor, and the content providing server includes a second processor, A memory connected to the second processor; and the second processor A second network interface connected to the processor, wherein the evaluation system includes a third processor, a memory connected to the third processor, and a third network connected to the third processor And the evaluation system includes a user attribute information reliability indicating the reliability of the user attribute information and the user attribute information, and a content attribute information reliability indicating the reliability of the content attribute information and the content attribute information. The method includes a first step in which the evaluation system selects an evaluation target user who is the user to update the user attribute information based on the user attribute information reliability, and the evaluation system For updating the user attribute information based on the content attribute information reliability A second step of selecting evaluation content as the content; and the evaluation system transmits the evaluation content to the evaluation target user, and acquires operation details of the evaluation target user with respect to the evaluation content. 3 and a fourth step in which the evaluation system updates the user attribute information and the user attribute information reliability of the evaluation target user based on the acquired operation content. And

  According to the present invention, a user with low user attribute information reliability is positively selected, and the user attribute information and user attribute information reliability of the user can be updated using content with high content attribute reliability. Further, content attribute information and content attribute information reliability of the content can be updated by positively selecting content with low content attribute information reliability and evaluating the content by a user with high user attribute reliability. Therefore, highly reliable evaluation information (user attribute information and content attribute information) can be acquired at a low cost and in a short time.

It is a block diagram explaining the computer system configuration | structure and software configuration | structure of embodiment of this invention. It is explanatory drawing which shows an example of the content information stored in the content information recording part of embodiment of this invention. It is explanatory drawing which shows an example of the user information stored in the user information recording part of embodiment of this invention. It is a flowchart explaining the process which the evaluation object content selection part of embodiment of this invention performs. It is a flowchart explaining the process which the evaluation object content selection part of embodiment of this invention performs. It is a flowchart explaining the process which the user selection part for evaluation of the 1st Embodiment of this invention performs. It is a flowchart explaining an example of the process performed in step 603 of embodiment of this invention. It is a flowchart explaining the process which the content selection part for evaluation of embodiment of this invention performs. It is a sequence diagram explaining the flow of a process of each structure in the content and user evaluation system of the 1st Embodiment of this invention.

  FIG. 1 is a block diagram illustrating a computer system configuration and a software configuration according to the embodiment of this invention.

  The computer system according to the embodiment of the present invention includes a content and user evaluation system 101, a content provider terminal 102, and a user terminal 103.

  The content and user evaluation system 101 and the content provider terminal 102 are connected via the Internet such as a LAN or a dedicated line. The content and user evaluation system 101 and the user terminal 103 are connected to each other via the Internet such as a LAN or a dedicated line.

  The content provider terminal 102 is a terminal used by the content provider, and transmits various types of content provided to the user who uses the user terminal 103. The content provider terminal 102 is a computer including a CPU (not shown), a memory such as RAM or ROM (not shown), a nonvolatile storage medium such as HDD or SSD (not shown), and a network interface (not shown). . The above-described configuration is connected using an internal bus or the like.

  The user terminal 103 is a terminal used by the user, and uses various contents transmitted from the content provider terminal 102. The user terminal 103 is a computer that includes a CPU (not shown), a memory such as RAM or ROM (not shown), a nonvolatile storage medium (not shown) such as an HDD or SSD, and a network interface (not shown). The above-described configuration is connected using an internal bus or the like.

  The content and user evaluation system 101 selects one or more contents from the contents received from the content provider terminal 102, transmits the selected contents to the user terminal 103, and sends the selected content to the content provider terminal 102. To the user, the usage status of the content and the user's evaluation of the content are transmitted.

  The content and user evaluation system 101 is a computer including a CPU (not shown), a memory such as RAM or ROM (not shown), a nonvolatile storage medium such as HDD or SSD (not shown), and a network interface (not shown). is there.

  In order to implement the above-described functions, the content and user evaluation system 101 includes a content registration unit 104, a content information recording unit 105, a user registration unit 106, a user information recording unit 107, a content recommendation unit 108, an information presentation unit 109, an evaluation A target content selection unit 110, an evaluation content selection unit 111, an evaluation target user selection unit 112, an evaluation user selection unit 113, a reaction acquisition unit 114, an attribute update unit 115, and a reporting unit 116 are provided.

  Each component described above is stored as software on a memory (not shown), and each function is realized by the software being executed by a CPU (not shown). Each configuration described above may be realized using hardware.

  The content registration unit 104 receives the content registration request transmitted from the content provider terminal 102 and outputs the content and content information included in the content registration request to the content information recording unit 104.

  The content information includes content attribute information, attribute reliability of the content attribute information, and the like. Here, the content attribute information is information representing the feature of the content, and includes one or more attribute values.

  The content information recording unit 105 stores the content information input from the content registration unit 104. The content information recording unit 105 can be realized by, for example, a relational database. The content information stored in the content information recording unit 105 will be described later with reference to FIG.

  The user registration unit 106 receives the user registration request transmitted from the user terminal 103 and outputs the user information included in the user registration request to the user information recording unit 107.

  The user information includes user attribute information, attribute reliability of the user attribute information, and the like. Here, the user attribute information is information representing the characteristics of the user and includes one or more attribute values.

  The user information recording unit 107 stores user information input from the user registration unit 106. The user information recording unit 107 can be realized by, for example, a relational database. The user information stored in the user information recording unit 107 will be described later with reference to FIG.

  The content recommendation unit 108 reads information from the content information recording unit 105 and the user information recording unit 107, and selects one or more contents estimated to be of interest to the user who uses the user terminal 103. Hereinafter, the content selected by the content recommendation unit 108 is referred to as recommended content.

  The content recommendation unit 108 executes processing when the user accesses the system using the user terminal 103 or periodically. The period for executing the process may be, for example, a notification of a monthly electricity bill or a certain campaign. When the process is periodically executed, the content recommendation unit 108 selects the optimum content for each user for all users.

  In the present embodiment, it is assumed that the content selection method recommended to the user uses a content-based recommendation method. That is, content that is content attribute information similar to user attribute information of an arbitrary user is selected as content that is likely to be of interest to the user.

  As a method for calculating the similarity between the user attribute information and the content attribute information, for example, a method of calculating an inner product of the multidimensional vector using attribute values included in the attribute information as multidimensional vectors can be considered. In this case, the content recommendation unit 108 selects, as recommended content, upper content that is sorted in descending order of the inner product value.

  The evaluation target content selection unit 110 selects content whose content attribute information is to be updated. Hereinafter, the content selected by the evaluation target content selection unit 110 is referred to as evaluation target content.

  The specific processing of the evaluation target content selection unit 110 will be described later with reference to FIGS. 4 and 5.

  The evaluation user selection unit 113 selects a user who requests evaluation of the content in order to update the content attribute information of the evaluation target content. Hereinafter, a user selected by the evaluation user selection unit 113 is referred to as an evaluation user.

  The evaluation target user selection unit 112 selects a user whose user attribute information is to be updated. Hereinafter, a user selected by the evaluation target user selection unit 112 is referred to as an evaluation target user.

  The evaluation content selection unit 111 selects content for evaluating the user in order to update the user attribute information of the evaluation target user. Hereinafter, the content selected by the evaluation content selection unit 111 is referred to as evaluation content.

  The information presenting unit 109 transmits information for presenting the recommended content, information for presenting the evaluation target content to the evaluation user, or information for presenting the evaluation content to the evaluation target user. As a result, various contents can be presented to the user who uses the user terminal 103.

  In addition, as a presentation method in the user terminal 103, a method of presenting a URL link for accessing various contents in an e-mail or the like, a URL link for accessing various contents on a screen of a website accessed by the user is presented. And a method of presenting a URL link for accessing various contents on a wall-mounted display that always displays the amount of electricity used.

  In the present embodiment, a method used in an existing content-based recommendation method can be used as a method for expressing content attribute information and user attribute information, a method for implementing the content recommendation unit 108, and the like.

  The reaction acquisition unit 114 acquires user responses to various contents presented on the user terminal 103.

  The attribute update unit 115 estimates content attribute information or user attribute information based on the user's reaction to the content acquired by the reaction acquisition unit 114, and updates the content information or user information based on the estimation result.

  The reporting unit 116 transmits the registered content attribute information to the content provider terminal 102 periodically or when the content attribute information is updated by the attribute update unit 115. That is, the content attribute information of the content registered by the content provider is reported to the content provider.

  Information to be transmitted includes current content attribute information, update history, attribute reliability, and the like.

  As a result, the content provider can grasp what kind of user attribute information the registered content is highly evaluated by the user, and is consistent with the user layer assumed by the content provider. It can be confirmed whether or not.

  In addition, when the attribute reliability 302 is low, if the reference number is high, it is considered that the user group is popular regardless of the user group. However, when the reference number is low, the interest of the user group assumed to use the content is captured. I can think of it not.

  An object of the present invention is to determine attribute information that appropriately represents each attribute of each content and each user. That is, an object of the present invention is to set content attribute information and user attribute information appropriately and update them according to the situation.

  For example, when the content provider has not set the content attribute information, the content provider can acquire the estimated content attribute information as a report by the process described later.

  In addition, when the content provider sets content attribute information based on the expectation that the user with the preference wants to use the content, or the user with the preference of the content attracts interest When the content information is set based on the prediction, the initial value of the set content attribute information does not always match the reputation from the actual user.

  According to the present invention, the attribute value of the content suitable for the actual user evaluation is calculated by the process described later, and the calculated attribute value can be fed back to the content provider.

  In addition, when the content provider conducts a questionnaire on the content in advance to the user and sets the initial value of each attribute information based on the questionnaire result, the initial value of the set attribute information should be respected. It is desirable to set a high initial value of the attribute reliability 202 for each content provider.

  For example, a method in which the content provider specifies that the attribute value is set based on a questionnaire or the like at the time of registration, or a method in which the content provider directly sets the attribute reliability 202 can be used.

  The content registration unit 104, content information recording unit 105, user registration unit 106, user information recording unit 107, content recommendation unit 108, and information presentation unit 109 are the same as those used in a general content-based recommendation method. belongs to. Also, the content attribute information and user attribute information expression method, and the content recommendation unit 108 mounting method can use those used in the existing content-based recommendation method.

  Hereinafter, details of each component of the present invention will be described.

  FIG. 2 is an explanatory diagram illustrating an example of content information stored in the content information recording unit 105 according to the embodiment of this invention.

  The content information includes a content ID 201, attribute reliability 202, attribute 203, and last evaluated date 204.

  The content ID 201 is an identifier for identifying content. The actual content is stored in a memory (not shown) area and can be referred to based on the content ID 201.

  The attribute reliability 202 is information indicating whether or not the content attribute information of the content corresponding to the content ID 201 is appropriate. That is, when the content attribute information is appropriate, it indicates that the content attribute information is difficult to update. When the update frequency or the difference between the updated values is large, it indicates that the content attribute information is not appropriate.

  In the attribute reliability 202, for example, a numerical value between “0” and “100” is stored. In the example illustrated in FIG. 2, “0” represents the lowest reliability, and “100” represents the highest reliability.

  At the time of content registration, a fixed value such as “50” or an attribute reliability assigned to each content provider using the content provider terminal 102 may be stored as an initial value of the attribute reliability 202. If the content attribute information is not set, “0” is stored in the attribute reliability 202.

  An attribute 203 is content attribute information of content corresponding to the content ID 201. That is, an attribute value representing the feature of the content is stored. Any attribute value may be used as long as it represents the user's interests and preferences.

  In the present embodiment, attribute values representing the user's interests and preferences regarding power usage are used. Specifically, the attribute 203 includes a short-term cost 210, a long-term cost 211, an environment 212, an action burden 213, a device purchase 214, a novelty 215, and a comparison / competition 216.

  The short-term cost 210 represents a degree of interest of a user who places importance on the short-term cost.

  The long-term cost 211 represents the degree of interest of the user who places importance on the long-term cost.

  The environment 212 represents the degree of interest of the user who places importance on the global environment or the local environment.

  The action burden 213 represents the degree to which the user feels a burden for performing a specific action (for example, turning off the lighting frequently for energy saving).

  Device purchase 214 represents the degree of user's willingness to purchase power related devices.

  The novelty 215 represents the degree of interest of the user with respect to a new service or device.

  The comparison / competition 216 represents the degree of tendency to determine the user's own behavior based on comparison with other users.

  Each attribute value included in the attribute 203 represents the degree of content suitability in terms of each attribute. For example, values from “0” to “1” are stored in each attribute value. In the example shown in FIG. 2, “0” represents the lowest fitness level, and “1” represents the highest fitness level.

  Further, when the content attribute information is not set by the content provider and the content attribute information is not estimated, “−” is stored in each attribute value. In this case, the attribute reliability 202 is “0”.

  In the example illustrated in FIG. 2, the content whose content ID 201 is “1” represents that the degree of matching with respect to a user who is interested in short-term costs is “0.2182179”. In the present embodiment, the attribute value is considered as an attribute value vector, and each attribute value of the attribute 203 is normalized so that the vector length is 1.

  For example, content that recommends replacement of an incandescent bulb with an LED lamp has a high value of the long-term cost 211 because it is considered that users who place importance on the long-term cost are interested. On the other hand, since it is considered that users who place importance on short-term costs are not interested, the value of long-term cost 211 of the content is low. In addition, a user with a high value of comparison / competition 216 tends to improve motivation for energy saving by referring to comparison data with other users who have similar power devices.

  The final evaluated date and time 204 stores the date and time when the content corresponding to the content ID 201 is evaluated by the evaluation user as the evaluation target content.

  The above is an example of content information stored in the content information recording unit 105.

  FIG. 3 is an explanatory diagram illustrating an example of user information stored in the user information recording unit 107 according to the embodiment of this invention.

  The user information includes content information 200 including a user ID 301, an attribute reliability 302, an attribute 303, a final evaluated date 204, and a final evaluation date 205.

  The user ID 301 is an identifier for identifying a user.

  The attribute reliability 302 is information indicating whether or not the user attribute information of the user corresponding to the user ID 301 is appropriate. That is, when the user attribute information is appropriate, it indicates that the user attribute information is difficult to be updated. When the update frequency or the difference between updated values is large, it indicates that the user attribute information is not appropriate.

  As in the attribute reliability 202, the attribute reliability 302 stores a numerical value between “0” and “100”. In the example shown in FIG. 3, “0” represents the lowest reliability, and “100” represents the highest reliability.

  In the attribute reliability 302, a fixed value such as “50” may be stored as an initial value at the time of user registration, or an initial value corresponding to the amount of answers to a questionnaire may be set. When the user attribute information is not set, “0” is set in the attribute reliability 302.

  An attribute 303 is user attribute information of a user corresponding to the user ID 301. That is, attribute values representing user characteristics are stored. Any attribute value may be used as long as it represents the user's interests and preferences.

  In the present embodiment, attribute values representing the user's interests and preferences regarding power usage are used. Specifically, the attributes 303 include a short-term cost 310, a long-term cost 311, an environment 312, an action burden 313, a device purchase 314, a novelty 315, and a comparison / competition 316.

  The short term cost 310 represents the degree to which the user corresponding to the user ID 301 places importance on the short term cost.

  The long-term cost 311 represents the degree to which the user corresponding to the user ID 301 places importance on the long-term cost.

  The environment 312 represents the degree to which the user corresponding to the user ID 301 places importance on the global environment or the local environment.

  The action burden 313 represents the degree of burden on the user corresponding to the user ID 301 performing a specific action (for example, frequently turning off the lighting for energy saving).

  The device purchase 314 represents the degree of willingness for the user corresponding to the user ID 301 to purchase the power-related device.

  The novelty 315 represents the degree to which the user corresponding to the user ID 301 is interested in a new service or device.

  The comparison / competition 316 represents the degree of tendency that the user corresponding to the user ID 301 determines his / her behavior based on comparison with other users.

  Each attribute value included in the attribute 303 represents the fitness of content in terms of each attribute. For example, values from “0” to “1” are stored in each attribute value. In the example illustrated in FIG. 3, “0” represents the lowest fitness level, and “1” represents the highest fitness level.

  If user attribute information is not set and user attribute information is not estimated, “-” is stored in each attribute value. In this case, the attribute reliability 302 is “0”.

  In this embodiment, the attribute value is considered as an attribute value vector, and each attribute value of the attribute 303 is normalized so that the vector length is 1.

  For example, content that recommends replacement of an incandescent light bulb with an LED lamp has a high value of the long-term cost 311 because it is considered that users who place importance on the long-term cost are interested. On the other hand, since a user who places importance on short-term costs is considered not interested in the content, the value of the long-term cost 211 of the content is low. In addition, a user with a high value of comparison / competition 316 tends to improve the motivation for energy saving by referring to comparison data with other users who have similar power devices.

  As the value of each attribute included in the attribute 303, for example, the user may directly input the initial value of each attribute at the time of user registration, or the initial value may be determined by referring to a correspondence table between questionnaire responses and attribute values. It may be set.

  The final evaluated date and time 304 stores the date and time when the user corresponding to the user ID 301 is evaluated using the evaluation content.

  The final evaluation date and time 305 stores the date and time when the evaluation target content is evaluated as the evaluation user.

  The above is an example of user information stored in the user information recording unit 107.

  Next, processing executed by the evaluation target content selection unit 110 will be described.

  The evaluation target content selection unit 110 executes two processes (modes) in parallel. Specifically, a mode for periodically selecting content with low attribute reliability 202 (hereinafter referred to as a first mode) and a mode for selecting content satisfying a predetermined condition (hereinafter referred to as a second mode). Are described in parallel).

  The first mode is executed at a timing such as once a week, once a day, a timing when a monthly electricity usage fee is notified, or a combination thereof. In the first mode, content having the lowest attribute reliability 202 is selected from all the content. A plurality of contents may be selected.

  On the other hand, in the second mode, processing is executed when the following three conditions are satisfied. That is, (Condition 1) At the time of content registration, (Condition 2) If the attribute reliability 202 is below the reference value for a predetermined period, (Condition 3) Within the predetermined period, the attribute reliability 202 is below the reference value It is three when it falls.

  In (Condition 1), processing is executed to set an initial value of content attribute information of the content at the time of content registration. (Condition 1) is a processing start condition for estimating an initial value of content in which the content attribute 203 and the attribute reliability 202 are not set.

  (Condition 2) is not selected as the evaluation target content in the first mode because the attribute reliability 202 is not low in all content, but the content is re-evaluated when the state with the low attribute reliability 202 continues. This is a start condition for the processing to be performed.

  (Condition 3) is a condition for starting the process of evaluating the content again when content with high attribute reliability 202 is used from a user layer that has not been used before. This is because the attribute reliability 202 is lowered because the content has been used from the user layer that has not been used before, and needs to be evaluated again.

  At the initial setting of the computer system or at the time of content registration, processing is executed to satisfy the condition of (Condition 1). Further, when the computer system is in operation, processing is executed according to the conditions (condition 2) and (condition 3). That is, the present invention can be applied to a newly constructed computer system or a computer system in operation.

  Hereinafter, the process performed by the evaluation target content selection unit 110 will be described separately for the first mode and the second mode.

  FIG. 4 is a flowchart illustrating processing executed by the evaluation target content selection unit 110 according to the embodiment of this invention. FIG. 4 is a flowchart for explaining processing in the first mode.

  The evaluation target content selection unit 110 starts processing when a selection event occurs when the above-described period has elapsed (step 401).

  The evaluation target content selection unit 110 reads the content information 200 stored in the content information recording unit 105, and selects the content having the lowest attribute reliability 202 based on the read content information 200 (step 402). A plurality of contents may be selected.

  The evaluation target content selection unit 110 outputs the selected content, that is, the evaluation target content to the information presentation unit 109 (step 403), and ends the process.

  In step 402, the evaluation target content selection unit 110 determines whether or not a certain period has elapsed since the last evaluated date and time 204. If the certain period has not elapsed since the last evaluated date and time 204, It does not have to be selected as content. This can prevent the same content from being frequently selected as the evaluation target content.

  FIG. 5 is a flowchart illustrating processing executed by the evaluation target content selection unit 110 according to the embodiment of this invention. FIG. 5 is a flowchart for explaining the processing in the second mode.

  The evaluation target content selection unit 110 starts processing when an attribute reliability change event occurs (step 501).

  Here, the attribute reliability change event is an event that occurs when the attribute reliability 202 of the content is changed. Further, when the content provider sets the initial value of the attribute reliability 202 at the time of initial content registration, or when the content provider sets “0” for the attribute reliability 202, the attribute reliability change event is also generated. Occur.

  Hereinafter, content whose attribute reliability 202 has been changed is referred to as processing target content.

  The evaluation target content selection unit 110 determines whether or not the processing target content is initially registered content (step 502). In step 502, it is determined whether or not (condition 1) is met.

  For example, the evaluation target content selection unit 110 refers to the content information 200 read from the content information recording unit 105 and determines whether or not a value is set for each attribute value of the attribute 203 of the processing target content. When no value is set for each attribute value of the attribute 203, that is, when “−” is stored in each attribute value, the evaluation target content selection unit 110 determines that the processing target content is content that is initially registered. judge.

  When it is determined that the processing target content is initially registered content, the evaluation target content selection unit 110 selects the processing target content as the evaluation target content (step 506). The evaluation target content selection unit 110 outputs the selected evaluation target content to the information presentation unit 109 and ends the process.

  When it is determined that the content to be processed is not initially registered content, the evaluation target content selection unit 110 refers to the last evaluated date and time 204 of the content information 200 and is constant from the date and time indicated in the final evaluated date and time 204. It is determined whether or not the period has elapsed (step 503).

  This is to prevent the same content from being frequently selected. Accordingly, content that has not passed a certain period from the date and time indicated in the final evaluated date and time 204 is not selected as the content to be evaluated.

  When it is determined that a certain period has not elapsed since the date / time indicated in the final evaluated date / time 204, the evaluation target content selection unit 110 ends the process.

  When it is determined that a certain period has elapsed since the date and time indicated in the last evaluated date and time 204, the evaluation target content selection unit 110 refers to the attribute reliability 202 of the content information 200, and for a predetermined period, the attribute reliability 202 Is less than or equal to the reference value (step 504). In step 504, it is determined whether or not (condition 2) is met.

  The reference value is a preset value, and the evaluation target content selection unit 110 holds the reference value.

  When it is determined that the attribute reliability 202 is equal to or lower than the reference value for a predetermined period, the evaluation target content selection unit 110 selects the processing target content as the evaluation target content (step 506). The evaluation target content selection unit 110 outputs the selected evaluation target content to the information presentation unit 109 and ends the process.

  When it is determined that the attribute reliability 202 is not less than or equal to the reference value for a predetermined period, the evaluation target content selection unit 110 refers to the attribute reliability 202 of the content information 200, and the attribute reliability 202 is a predetermined value within the predetermined period. It is determined whether or not the number has decreased (step 505). In step 505, it is determined whether or not (Condition 3) is satisfied.

  If it is determined that the attribute reliability 202 has not decreased by a predetermined value or more within the predetermined period, the evaluation target content selection unit 110 ends the process.

  When it is determined that the attribute reliability 202 has decreased by a predetermined value or more within the predetermined period, the evaluation target content selection unit 110 selects the processing target content as the evaluation target content (step 506). The evaluation target content selection unit 110 outputs the selected evaluation target content to the information presentation unit 109 and ends the process.

  In addition, you may perform the process of step 504 and step 505 from any process.

  The evaluation target user selection unit 112 can use the same process as the process executed by the evaluation target content selection unit 110. In this case, the selection target is the user. Therefore, the processing is the same as that in the flowcharts shown in FIGS. Therefore, description of the process of the evaluation target user selection unit 112 is omitted.

  Next, processing executed by the evaluation user selection unit 113 will be described.

  The evaluation user selection unit 113 executes processing in conjunction with the evaluation target content selection unit 110. That is, the process is started when the evaluation target content selection unit 110 selects the evaluation target content.

  FIG. 6 is a flowchart illustrating processing executed by the evaluation user selection unit 113 according to the first embodiment of this invention.

  When starting the process (step 601), the evaluation user selection unit 113 refers to the user information 300 read from the user information recording unit 107, the attribute reliability 302 is equal to or greater than a predetermined value, and the final evaluation date and time 305. The users who have passed for a certain period from the date and time shown in (2) are extracted (step 602).

  The reason why the user with high attribute reliability 302 is extracted is that the evaluation with respect to highly reliable content can be obtained by the user with high attribute reliability 302 evaluating. On the other hand, even if the evaluation for the content of the user having the low attribute reliability 302 is collected, the collected evaluation cannot be estimated correctly because the reliability is low.

  The final evaluation date and time 305 is considered in order to prevent a user with a high attribute reliability 302 from being frequently selected as an evaluation user. As a result, the burden on the user with high attribute reliability 202 can be reduced.

  The predetermined value and the predetermined period are set in advance, and these are held by the evaluation user selection unit 113.

  Next, the evaluation user selecting unit 113 selects a user whose extracted attribute value vector is dissimilar (step 603).

  In step 603, a user whose attribute values of the attribute 303 are separated is selected. That is, a plurality of users having high attribute reliability 302 and different attribute values (preference tendencies) of the attribute 303 are selected.

  This is because the evaluation of content may be biased when a user with a similar preference tendency, that is, a user with a similar attribute value vector is selected as the user for evaluation. Therefore, useful information for estimating the attribute information of the evaluation target content can be acquired by selecting a user whose attribute value vector is dissimilar.

  Note that the user selection method in step 603 will be described later with reference to FIG.

  The evaluation user selection unit 113 outputs the selected user, that is, the evaluation user to the information presentation unit 109 (step 604), and ends the process.

  Through the process described above, a user having a high attribute reliability 202 is selected as an evaluator for updating the content attribute information. In order to calculate highly reliable content attribute information, many users with high attribute reliability 302 are required. Therefore, in the present invention, the number of users having high attribute reliability 302 can be increased by the evaluation content selection unit 111 and the evaluation target user selection unit 112 described later.

  FIG. 7 is a flowchart illustrating an example of processing executed in step 603 according to the embodiment of this invention.

  In FIG. 7, a K-means clustering method will be described as a method of selecting users with different attribute values. K-means clustering is a method of classifying a plurality of data expressed by multidimensional vectors into K clusters so that separation between clusters is improved.

When the evaluation user selection unit 113 starts the processing (step 701), the attribute value of the extracted user attribute 303 is set to the attribute value vector x _i (i = 1... N), and randomly assigned to K clusters. Is assigned an attribute value vector x _i (step 702). Here, n represents the number of extracted users.

Next, the evaluation user selection unit 113 calculates the center V _j (j = 1... K) of each cluster (step 703). For example, the center of the cluster can be calculated using an average calculation or the like.

The evaluation user selection unit 113 calculates the distance between each attribute value vector x _i (i = 1... N) and the cluster center V _j (j = 1... K) (step 704).

The evaluation user selection unit 113 changes the assignment of each attribute value vector x _i (i = 1... N) to the cluster based on the distance calculated in step 704 (step 705).

Specifically, the attribute value vector _{x i (i = 1 ... n} ) and the distance between the center V _j of the cluster is the smallest, i.e., the cluster center V _j and distance closest cluster attribute value vector x _i ( The allocation of i = 1... n) is changed.

The evaluation user selection unit 113 determines whether or not the assignment of each attribute value vector x _i (i = 1... N) to the cluster has been changed (step 706).

If it is determined that the assignment of each attribute value vector x _i (i = 1... N) to the cluster has been changed, the evaluation user selection unit 113 returns to step 703 and executes the same processing (step 703 to step 703). 708).

If it is determined that the assignment to the cluster has not been changed, the evaluation user selection unit 113 calculates the finally determined cluster center V _j (j = 1... K) (step 707).

The evaluation user selection unit 113 selects the attribute value vector closest to the center V _j (j = 1... K) of each cluster calculated in step 707 from the attribute value vectors belonging to each cluster (step 708). The process is terminated. That is, the user corresponding to the attribute value vector is selected.

  Through the above processing, a predetermined number (K people) of users whose attribute value vectors vary greatly is selected.

  In the present invention, a user with a high attribute reliability 302 is selected in order to update content attribute information of content with a low attribute reliability 202. As a result, content attribute information having a high attribute reliability 202 can be set.

  In the present invention, in order to update the user attribute information of a user having a low attribute reliability 302, a content having a high attribute reliability 202 is selected. As a result, user attribute information having a high attribute reliability 302 can be set.

  That is, both the content and the user can improve the attribute reliability 202 and 302 of the attribute information of each other. Therefore, appropriate content attribute information and user attribute information can be set for the entire computer system.

  Next, processing executed by the evaluation content selection unit 111 will be described.

  The evaluation content selection unit 111 is different in that the evaluation user selection unit 113 selects the evaluation content based on a criterion different from the selection criterion for the evaluation user.

  Specifically, the evaluation content selection unit 111 next selects evaluation content based on two criteria.

  (Criteria 1) The user group to be used is stable, that is, the attribute reliability 202 is high, and (Criteria 2) variation in attribute values of content attribute information is large.

  The evaluation content selection unit 111 does not consider information corresponding to the final evaluation date 305 when selecting content. This is because there is no problem even if the evaluation content is frequently selected.

  The evaluation content selection unit 111 can select K evaluation contents having high attribute reliability 202 and large variation in attribute values based on (Criteria 1) and (Criteria 2). Here, K is different from the number of evaluation users.

  FIG. 8 is a flowchart illustrating processing executed by the evaluation content selection unit 111 according to the embodiment of this invention.

  When the evaluation content selection unit 111 starts the processing (step 801), the content selection unit 111 refers to the content information 200 read from the content information recording unit 105, and extracts content whose attribute reliability 202 is a predetermined value or more (step). 802).

  As described above, information corresponding to the final evaluation date 305 is not considered. It should be noted that the method for extracting the content having the attribute reliability 202 of a predetermined value or more can use the same method as the processing executed by the evaluation user selection unit 113.

  Next, the evaluation content selection unit 111 selects content whose attribute value vectors of the extracted content are dissimilar (step 803).

  The process in step 803 is the same as that in step 603. Useful content for estimating the user attribute information of the evaluation target user can be acquired by selecting the content having dissimilar attribute value vectors.

  In step 803, content can be selected using a K-means clustering method (see FIG. 7).

  The evaluation content selection unit 111 outputs the selected content, that is, the evaluation content to the information presentation unit 109 (step 804), and ends the process.

  FIG. 9 is a sequence diagram illustrating the flow of processing of each component in the content and user evaluation system 101 according to the first embodiment of this invention.

  When the evaluation target content selection unit 110 executes the process according to a certain cycle, the evaluation user selection unit 113 also executes the process in conjunction with the process (step 901).

  The evaluation target content selection unit 110 and the evaluation user selection unit 113 respectively output the evaluation target content and the evaluation user to the information presentation unit 109 (step 902).

  Based on the inputs from the evaluation target content selection unit 110 and the evaluation user selection unit 113, the information presentation unit 109 transmits information for presenting the evaluation target content to the user terminal 103 used by the evaluation user (step). 903).

  In addition, the process of step 916-step 918 and step 925-step 927 is also the same process.

  When an attribute reliability change event occurs due to content registration or the like, the evaluation target content selection unit 110 executes processing, and the evaluation user selection unit 113 also executes processing in conjunction (step 904). The attribute reliability change event has been described with reference to FIG.

  The evaluation target content selection unit 110 and the evaluation user selection unit 113 each output the evaluation target content and the evaluation user to the information presentation unit 109 (step 905).

  Based on the inputs from the evaluation target content selection unit 110 and the evaluation user selection unit 113, the information presentation unit 109 transmits information for presenting the evaluation target content to the user terminal 103 used by the evaluation user (step). 906).

  In addition, the process of step 907-step 909, step 913-step 915, and step 922-step 924 is also the same process.

  When an attribute reliability change event occurs due to a decrease in the attribute reliability 202, the evaluation target content selection unit 110 executes the process, and the evaluation user selection unit 113 also executes the process (step 910). The attribute reliability change event has been described with reference to FIG.

  The evaluation target content selection unit 110 and the evaluation user selection unit 113 respectively output the evaluation target content and the evaluation user to the information presentation unit 109 (step 911).

  Based on the inputs from the evaluation target content selection unit 110 and the evaluation user selection unit 113, the information presentation unit 109 transmits information for presenting the evaluation target content to the user terminal 103 used by the evaluation user (step). 912).

  In addition, the process of step 919-step 921 and step 928-step 930 is also the same process.

  Next, content attribute information and user attribute information update processing will be described.

  In the update process of the content attribute information and the user attribute information, an evaluation user response to the evaluation target content or an evaluation target user response to the evaluation content is used. The information regarding the reaction of each user described above is acquired by the reaction acquisition unit 114.

  The reaction acquisition unit 114 can acquire, as a user reaction, whether or not the URL link presented on the e-mail or the Web site has been accessed as the user response to the recommended content. Moreover, a user's content utilization time can also be acquired as a user's reaction.

For example, the evaluation value μ _ij of the user j for the content i is “0” when the content is not used, “1” when the content is used, and “2” when the content is used for a predetermined time. It is possible to decide.

  In addition, at the end of content use, an evaluation questionnaire can be presented to the user, and the answer can be acquired as the user's evaluation. In this case, for example, a method is conceivable in which the user designates evaluation values from “−2” which is not interested at all to “2” which is very interested.

  The attribute update unit 115 estimates the content information 200 or the user information 300 and calculates the attribute reliability 202 or the attribute reliability 302.

  Further, the attribute updating unit 115 stores the content information 200 or the user information 300 in the content information recording unit 105 or the user information recording unit 107 based on the calculated attribute reliability 202 or attribute reliability 302. Updated information.

  Hereinafter, a case where the content attribute information is updated will be described as an example.

When the attribute value vector of the evaluation target content is y _i , the attribute value vector y _i is updated as shown in Expression (1) by the evaluation by the evaluation user.

Here, α represents a control parameter for the update rate. E _i represents an estimated value of each attribute value, and is calculated using Expression (2).

Here, K represents the number of users for evaluation. X _j represents the attribute value vector of the evaluation user j. Here, it is assumed that evaluation is periodically performed and content attribute information is updated. When the attribute value y _i is not set by the content provider, e _i is set as the updated attribute value.

When the attribute reliability 202 of the evaluation target content i is set to c _i , the attribute reliability 202 is updated using Expression (3).

Here, β represents an update rate.

If the attribute reliability 202 is low and there is a difference between the pre-update attribute value y _i and the estimated value e _i , it is considered that the content attribute information has been corrected, and the attribute reliability 202 is changed to a high value. Also, if the attribute reliability 202 is high and there is a difference between the pre-update attribute value y _i and the estimated value e _i , it is considered that the user group using the content has changed, and the attribute reliability 202 is changed to a low value Is done.

Note that the user attribute information of the evaluation target user i can also be updated using the equations (1) and (2). Here, K represents the number of evaluation contents, and x _j represents an attribute value vector of the evaluation contents j. Μ _ij represents an evaluation value for the evaluation content j of the evaluation target user i. Also, the attribute reliability 302 of the evaluation target user i can be updated by using Expression (3).

  According to one aspect of the present invention, when the attribute reliability 202 of content provided by an electric power company or a third party is low, or when the attribute reliability 202 decreases, a reliable user is requested to evaluate the content. Thus, content attribute information with high attribute reliability 202 can be set.

  As a result, it is not necessary to perform an unnecessary evaluation, and further, since the content evaluation is not requested for a user having a low attribute reliability 302, it is related to the content evaluation such as monetary cost and computer system resources. Costs can be reduced and accurate content attribute information can be set quickly.

  Further, according to another aspect of the present invention, it is possible to evaluate a user whose usage frequency is low and whose preference tendency is not determined using a plurality of contents having high attribute reliability 202 and varying attribute tendencies. It becomes possible. Thereby, user attribute information having a high attribute reliability 302 can be set quickly. Therefore, a user's preference tendency can be analyzed quickly.

[Modification]
Hereinafter, modifications of the present invention will be described.

  When executing the process of selecting the evaluation target content or the evaluation target user at a periodic timing and the timing overlaps with the timing of the fee notification once a month, the content recommendation unit 108 selects the evaluation target user and the evaluation target For a user other than the user, content based on the content-based recommendation method is selected, and the information presenting unit 109 can transmit the selected content to the user.

  At this time, in addition to the evaluation target content, the information presenting unit 109 can also transmit content based on the content-based recommendation method to the user terminal 103 used by the evaluation user.

  Thus, at the timing of the charge notification, (1) the attribute reliability 302 is low, so the attribute reliability 302 is high, and the evaluation target user who is presented with the content for evaluation with a large variation in attribute values, (2) attribute reliability Since the degree 302 is high, the evaluation target content whose attribute reliability 202 is low and the evaluation user who presents the content according to the user's preference, (3) In addition, only the content according to the user's preference is presented There will be a user.

  As described above, according to one aspect of the present invention, in a situation where collaborative filtering does not function accurately because the number of references to the user's content is small, the content and user evaluation system 101 has a high attribute reliability 302 for evaluation. By requesting the user to evaluate the evaluation target content having a low attribute reliability 202 or presenting the evaluation content having a high attribute reliability 202 to the evaluation target user having a low attribute reliability 302, the attribute trust Content attribute information having a high degree 202 and attribute reliability 302 can set user attribute information.

  Thereby, functions useful for both the content provider and the user, such as a recommendation function and a report function, can be realized.

DESCRIPTION OF SYMBOLS 101 User evaluation system 102 Content provider terminal 103 User terminal 104 Content registration part 105 Content information recording part 106 User registration part 107 User information recording part 108 Content recommendation part 109 Information presentation part 110 Evaluation object content selection part 111 Evaluation content selection part 112 Evaluation target user selection unit 113 Evaluation user selection unit 114 Reaction acquisition unit 115 Attribute update unit 116 Reporting unit 200 Content information 300 User information

Claims (10)

A computer used by the user, a content providing server that transmits the content to be provided to the user, the content transmitted from the content providing server, the content to be provided to the user is selected, and the computer is selected. An attribute information update method in a computer system including an evaluation system that transmits the selected content and updates the user attribute information representing the characteristics of the user or the content attribute information representing the characteristics of the content,
The computer includes a first processor, a memory connected to the first processor, and a first network interface connected to the first processor,
The content providing server includes a second processor, a memory connected to the second processor, and a second network interface connected to the second processor,
The evaluation system includes a third processor, a memory connected to the third processor, and a third network interface connected to the third processor,
The evaluation system holds user attribute information reliability indicating the reliability of the user attribute information and the user attribute information, and content attribute information reliability indicating the reliability of the content attribute information and the content attribute information,
The method
A first step in which the evaluation system selects an evaluation target user who is the user to update the user attribute information based on the user attribute information reliability;
A second step in which the evaluation system selects content for evaluation, which is the content for updating the user attribute information, based on the content attribute information reliability;
A third step in which the evaluation system transmits the evaluation content to the evaluation target user and acquires operation details of the evaluation target user with respect to the evaluation content;
A fourth step in which the evaluation system updates the user attribute information and the user attribute information reliability of the evaluation target user based on the acquired operation content;
The attribute information update method characterized by including. A fifth step in which the evaluation system selects, based on the content attribute information reliability, an evaluation target content that is the content for updating the content attribute information;
A sixth step in which the evaluation system selects an evaluation user who is the user for updating the content attribute information based on the user attribute information reliability;
A seventh step in which the evaluation system transmits the evaluation target content to the evaluation user and acquires operation details of the evaluation user with respect to the evaluation target content;
The evaluation system includes an eighth step of updating the content attribute information and the content attribute information reliability of the evaluation target content based on the acquired operation content. The attribute information update method described in 1. The user attribute information includes a plurality of user feature values representing the features of the user,
The content attribute information includes a plurality of content feature values representing features of the content,
The user attribute information reliability is numerical information indicating the update frequency of the user attribute information or the change amount of the user feature value, and the change amount of the user feature value is small when the update frequency of the user attribute information is low. In the case, the user attribute information reliability is set high,
The content attribute information reliability is numerical information indicating the update frequency of the content attribute information or the change amount of the content feature value, and when the update frequency of the content attribute information is low or the change amount of the content feature value is small. In the case, the content attribute information reliability is set high,
In the second step, the content attribute information reliability value is equal to or greater than a predetermined value, and the content having different characteristics of the content feature value included in the content attribute information is selected as the evaluation content. Including
The sixth step is a step of selecting, as the evaluation user, the user having a user attribute information reliability value equal to or greater than a predetermined value and having a different characteristic of the user feature value included in the user attribute information. The attribute information update method according to claim 2, further comprising: The evaluation system retains time information related to the time selected as the evaluation user,
The attribute information updating method according to claim 3, wherein the sixth step further selects, as the evaluation user, a user who has passed a predetermined period from the time indicated in the time information. The user attribute information includes a plurality of user feature values representing the features of the user,
The content attribute information includes a plurality of content feature values representing features of the content,
The user attribute information reliability is numerical information indicating the update frequency of the user attribute information or the change amount of the user feature value, and the change amount of the user feature value is small when the update frequency of the user attribute information is low. In the case, the user attribute information reliability is set high,
The content attribute information reliability is numerical information indicating the update frequency of the content attribute information or the change amount of the content feature value, and when the update frequency of the content attribute information is low or the change amount of the content feature value is small. In the case, the content attribute information reliability is set high,
The first step includes periodically referring to the user attribute information reliability, and selecting the user whose value of the user attribute information reliability is a predetermined value or less as the evaluation target user,
The fifth step includes a step of periodically referring to the content attribute information reliability and selecting the content having a content attribute information reliability value of a predetermined value or less as the evaluation target content. The attribute information update method according to claim 2. The first step includes a step of selecting the user included in the user registration request as the evaluation target user when receiving a user registration request for registering a new user from the computer;
The fifth step includes a step of selecting the content included in the content registration request as the evaluation target content when receiving a content registration request for registering new content from the content providing server. The attribute information update method according to claim 2, wherein: The user attribute information includes a plurality of user feature values representing the features of the user,
The content attribute information includes a plurality of content feature values representing features of the content,
The user attribute information reliability is numerical information indicating the update frequency of the user attribute information or the change amount of the user feature value, and the change amount of the user feature value is small when the update frequency of the user attribute information is low. In the case, the user attribute information reliability is set high,
The content attribute information reliability is numerical information indicating the update frequency of the content attribute information or the change amount of the content feature value, and when the update frequency of the content attribute information is low or the change amount of the content feature value is small. In the case, the content attribute information reliability is set high,
The first step refers to the user attribute information reliability, and selects, as the evaluation target user, the user for which a state where the value of the user attribute information reliability is equal to or less than a predetermined value continues for a predetermined period. Including steps,
In the fifth step, the content attribute information reliability is referred to, and the content in which the value of the content attribute information reliability is not more than a predetermined value continues for a predetermined period is selected as the evaluation target content The attribute information update method according to claim 2, further comprising steps. The user attribute information includes a plurality of user feature values representing the features of the user,
The content attribute information includes a plurality of content feature values representing features of the content,
The user attribute information reliability is numerical information indicating the update frequency of the user attribute information or the change amount of the user feature value, and the change amount of the user feature value is small when the update frequency of the user attribute information is low. In the case, the user attribute information reliability is set high,
The content attribute information reliability is numerical information indicating the update frequency of the content attribute information or the change amount of the content feature value, and when the update frequency of the content attribute information is low or the change amount of the content feature value is small. In the case, the content attribute information reliability is set high,
The first step refers to the user attribute information reliability, and selects the user whose user attribute information reliability value has decreased by a predetermined value or more within a predetermined period as the evaluation target user. Including
The fifth step is a step of referring to the content attribute information reliability and selecting, as the evaluation target content, the content whose content attribute information reliability value has decreased by a predetermined value or more within a predetermined period. The attribute information update method according to claim 2, further comprising: The user attribute information includes a plurality of user feature values representing the features of the user,
The content attribute information includes a plurality of content feature values representing features of the content,
The user attribute information reliability is numerical information indicating the update frequency of the user attribute information or the change amount of the user feature value, and the change amount of the user feature value is small when the update frequency of the user attribute information is low. In the case, the user attribute information reliability is set high,
The content attribute information reliability is numerical information indicating the update frequency of the content attribute information or the change amount of the content feature value, and when the update frequency of the content attribute information is low or the change amount of the content feature value is small. In the case, the content attribute information reliability is set high,
The fourth step includes
Based on the content attribute information of the evaluation content, the user attribute information of the evaluation target user, and the operation content of the evaluation target user with respect to the evaluation content, the user attribute information of the evaluation target user is included in the user attribute information Calculating an estimated value of the user feature value;
Calculating an update value of the user feature value included in the user attribute information of the evaluation target user based on the estimated value of the user feature value and the user attribute information of the evaluation target user;
Calculating an update value of the user attribute information reliability based on the estimated value of the user feature value, the update value of the user feature value, and the reliability of the user attribute information, and
The eighth step includes
The content included in the content attribute information of the evaluation target content based on the user attribute information of the evaluation user, the content attribute information of the evaluation target content, and an operation content of the evaluation user for the evaluation target content Calculating an estimate of the feature value;
Calculating an update value of the content feature value included in the content attribute information of the evaluation target content based on the estimated value of the content feature value and the content attribute information of the evaluation target content;
And a step of calculating an update value of the content attribute information reliability based on the estimated value of the content feature value, the update value of the content feature value, and the reliability of the content attribute information. 2. The attribute information update method according to 2. An information updating method in a computer system comprising a first computer, a second computer, and an update server for updating first information related to the first computer,
The first computer includes a first processor, a memory connected to the first processor, and a first network interface connected to the first processor,
The second computer includes a second processor, a memory connected to the second processor, and a second network interface connected to the second processor,
The update server includes a third processor, a memory connected to the third processor, and a third network interface connected to the third processor,
The update server includes the first information and the reliability of the first information indicating the update degree of the first information, and the second information indicating the update degree of the second information and the second information related to the second computer. Keep the confidence of
The method
The update server selects the evaluation target first information that is the first information for updating the reliability of the first information based on the reliability of the first information;
The update server selecting second evaluation information that is the second information for updating the reliability of the first information based on the reliability of the second information;
The update server updating the reliability of the evaluation target first information and the evaluation target first information using the evaluation second information;
An information update method comprising:

Images