JP5962456B2 - Information selection device, information selection method, and information selection program - Google Patents

Description

  The present invention relates to an information selection technique for selecting recommendation information.

There is a growing need for a technique for recommending items useful for users among a large number of items. As one of the methods for recommending items, a method for recommending in units of categories is known. When recommending by category, it is necessary to determine the recommended category. Patent Document 1 discloses a program recommendation device that determines a recommended category and a recommended program belonging to the recommended category. The program recommendation apparatus of patent document 1 determines a recommendation category using the user preference degree for every program category. Here, the user preference level for each program category is the degree of relevance between each program category and the TV program included in the user's recording history, or each program category and the user's EPG (Electronic
Shows the degree of association with the TV program selected from Program Guide.

  More specifically, in order to determine a program category to be recommended, the program recommendation device disclosed in Patent Document 1 first uses a TV program recording reservation history (or selection history from the EPG) by a user for each program category. The user preference level is calculated. Next, the program recommendation device determines a program category having a user preference level exceeding a threshold as a recommended program category. Then, the program recommendation device determines a recommended program from a group of programs corresponding to the determined recommendation category (see FIG. 8 of Patent Document 1). The determined recommended program category and the recommended program are displayed on the display in units of categories, for example (see FIG. 10 of Patent Document 1).

International Publication No. 2005/048587

  As described above, the technique disclosed in Patent Document 1 is calculated using a program included in the recording reservation history (or selection history from the EPG) of the TV program by the user in order to determine the recommended program category. The user preference level for each program category is used. However, the technique disclosed in Patent Document 1 does not sufficiently consider information of television programs belonging to each program category when determining a recommended program category. For this reason, in the technique disclosed in Patent Document 1, there are not always recommended programs associated with the recommended program category.

  Therefore, even if the technique disclosed in Patent Document 1 can present the recommended program category based on the program included in the recording reservation history (or selection history from the EPG) of the TV program by the user, the recommended program is sufficiently presented. It may not be possible. For the user, the program that is the target of direct use is more important than the program category, and the program category is merely auxiliary information. That is, even if the recommended program category can be presented, viewing of the program that is the original purpose of the recommendation is not promoted unless the recommended program category is sufficiently associated with the recommended program.

  This problem is not limited to providing a recommended category related to a television program, but can occur widely when providing a recommended category related to some item.

  In order to solve the above-mentioned problems, the present invention uses an information regarding a period during which an item can be provided to the user or a usage history in which the user uses the item to record the item to the user. A recommendation usability determination unit (1073) that determines usefulness to recommend and generates a recommended candidate item set that is a set of items determined to be useful, and belongs to the recommended candidate item set and belongs to a candidate category A category priority calculation unit (category selection unit 1071) that calculates a category priority using an item relevance indicating the strength of relevance between an item and the user, and a plurality of items calculated for each of a plurality of candidate categories A category for selecting a recommended category to be provided to the user based on the category priority of the plurality of candidate categories. Providing information selection device (10) characterized by having a re-selection unit.

  In order to solve the above-mentioned problem, the present invention provides information on a period of time during which an item can be provided to a user or a usage history in which information on the use of an item is recorded. Determining a usefulness of recommending an item, generating a recommended candidate item set that is a set of items determined to be useful, an item belonging to the recommended candidate item set and belonging to a candidate category, and the user And providing to the user based on the plurality of category priorities calculated for each of the plurality of candidate categories, using the item relevance indicating the strength of relevance And selecting a recommended category to be selected from the plurality of candidate categories.

  In order to solve the above-mentioned problem, the present invention provides information on a period of time during which an item can be provided to a user or a usage history in which information on the use of an item is recorded. A process for determining usefulness of recommending an item, generating a recommended candidate item set that is a set of items determined to be useful, an item belonging to the recommended candidate item set and belonging to a candidate category, and the user Provided to the user based on the processing of calculating the category priority using the item relevance indicating the strength of the relevance to the user and the plurality of category priorities calculated for each of the plurality of candidate categories An information selection program that causes a computer to execute a process of selecting a recommended category to be selected from among the plurality of candidate categories. To provide a beam.

  According to each aspect mentioned above, it can contribute to provision of the recommendation information which shows the recommendation category with high possibility that the recommendation item which has high relevance with a user main body exists.

It is a block diagram which shows the structural example of the network system containing an information selection apparatus. It is a block diagram which shows the other structural example of the network system containing an information selection apparatus. It is a block diagram which shows the structural example of an information selection apparatus. It is a flowchart which shows an example of the selection procedure of the recommendation category by the information selection apparatus which concerns on 1st Embodiment. It is a flowchart which shows an example of the selection procedure of the recommendation category by the information selection apparatus which concerns on 2nd Embodiment. It is a block diagram which shows the structural example of the information selection apparatus which concerns on 3rd Embodiment. It is a figure which shows the example of the item information table, the category information table, and the item-category correspondence table which are stored in the metadata storage unit. It is a figure which shows the example of the item utilization history table and category utilization history table which are stored in a utilization history storage part. It is a figure which shows the example of the category relevance table stored in a category relevance storage part. It is a figure which shows the example of the recommendation information table stored in a recommendation information storage part. It is a figure which shows the example of the item relevance degree table stored in an item relevance degree storage part. It is a block diagram which shows the structural example of an item provision server. It is a block diagram which shows the structural example of a terminal device. It is a flowchart explaining the specific example of operation | movement of the whole system. It is a figure which shows the example of a display screen based on the response data from an item provision server. It is a figure which shows the example of a screen which displays a recommendation list. It is a flowchart explaining the specific example of the operation | movement which an information selection apparatus produces recommendation information. It is a flowchart explaining the specific example of an item relevance degree calculation process. It is a flowchart explaining the specific example of a category relevance calculation process. It is a flowchart explaining the 1st specific example of a recommendation information creation process. It is a flowchart explaining the 2nd specific example of recommendation information creation processing. It is a figure which shows the example of a characteristic of the conversion function F (X) which makes a category relevance the input X, and makes the expansion relevance the output Y. It is a figure for demonstrating the specific example of the recommendation item determination process in the 1st specific example of a recommendation information creation process. It is a figure which shows the example of the temporary table in the 2nd specific example of recommendation information creation processing. It is a figure which shows the 1st modification of the screen which displays a recommendation list. It is a figure which shows the example of a screen which adjusts an expansion relevance degree. It is a figure for demonstrating the adjustment of the characteristic of the conversion function F (X) used for calculation of an extended relevance degree. It is a figure which shows the 2nd modification of the screen which displays a recommendation list. It is a figure which shows the example of the screen which displays item information in the 2nd modification of the screen which displays a recommendation list.

  Hereinafter, specific embodiments will be described in detail with reference to the drawings. In each drawing, the same or corresponding elements are denoted by the same reference numerals, and redundant description is omitted as necessary for clarification of the description.

First, definitions of terms used in this specification will be described. As used herein, the term “item” refers to various objects that are utilized (eg, purchased, viewed, viewed, rated, selected, or searched) by a user. In other words, the term item means a generic concept that includes goods, information, and services, whether tangible or intangible. The item includes, for example, at least one of the following (a) to (d).
(A) Digital contents as intangibles (for example, text data, audio data, music data, photo data, video data, computer programs, web pages), or information related thereto (for example, product information, advertisements)
(B) Non-electronic products (for example, goods, financial products) as tangible or intangible objects, or information related thereto (for example, product information, advertisements)
(C) Information (for example, ticket sales information, reservation information, TV program information) regarding services (for example, sports performances, concert performances, movie performances, television broadcasting, network services)
(D) Search keywords and search results

  In this specification, the term “category” means an item for classifying items. Categories can also be referred to as groups or classes. The correspondence between the item and the category (for example, the category to which the item belongs) may be explicitly specified by metadata attached to the item, a correspondence table, or the like. Further, the correspondence between items and categories may be determined by evaluating the relevance, similarity, or correlation between items and categories.

In addition, the term “use subject” in this specification means a user who uses an item or a terminal device used when a user uses an item. Further, the term “user entity identifier” means a user identifier for identifying a user who uses an item or a terminal identifier for identifying a terminal device used when a user uses an item.
In this specification, the term “relevance” refers to the strength of relationship, similarity, preference, or correlation between elements (for example, between a user and an item, between a user and a category, or between an item and a category). It means a scale indicating.

<First Embodiment>
The present embodiment includes an information selection device 10. FIG. 1 is a block diagram illustrating a configuration example of a system 1 including an information selection device 10 according to the present embodiment. The information selection device 10 is used in combination with the item providing server 20 or at least one terminal device 30 and operates to select a recommended category recommended to the user. The information selection device 10 may further determine recommended items recommended for the user. Thereby, recommendation information indicating a recommendation category or a recommendation category and a recommendation item is provided to the user via the terminal device 30.

The information selection device 10 includes at least one processor (for example, a microprocessor, an MPU (Micro Processing Unit), a CPU (Central
It may be configured using a computer system having a Processing Unit)). The computer system can function as the information selection device 10 by executing one or a plurality of computer programs including an instruction group for performing an algorithm related to selection of a recommendation category, which will be described later.

This program can be stored using various types of non-transitory computer readable media and supplied to a computer. Non-transitory computer readable media are various types of tangible recording media (tangible
storage medium). Examples of non-transitory computer-readable media include magnetic recording media (for example, flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (for example, magneto-optical disks), optical disks (for example, CD-ROM (Read
Only Memory), CD-R, CD-R / W, DVD-ROM, DVD-R), semiconductor memory (for example, mask ROM, PROM (Programmable ROM), EPROM (Erasable)
PROM), flash ROM, RAM (random access memory)). The program can also contain various types of temporary computer readable media (transitory).
computer readable medium). Examples of transitory computer readable media include electrical signals, optical signals, and electromagnetic waves. The temporary computer-readable medium can supply the program to the computer via a wired communication path such as an electric wire and an optical fiber, or a wireless communication path.

  The information selection device 10 may be configured using a plurality of computers. For example, a plurality of computers each responsible for the same processing block of the information selection device 10 may be used for load distribution. Further, a certain computer may be in charge of a part of processing blocks of the information selection apparatus 10, and another computer may be in charge of another part of processing blocks.

The item providing server 20 transmits an item (for example, digital content, a computer program) or information about the item (for example, information about electronic or non-electronic products / services) to the terminal device 30 via the network 40. Specific examples of the item providing server 20 include a digital content providing server, a digital content sharing server, a search server, and an EC (Electronic
Commerce) system, including TV broadcasting stations.

The terminal device 30 is used by a user. Specific examples of the terminal device 30 include a mobile phone terminal, a smartphone, a digital audio player, a tablet computer, a notebook PC (Personal
Computer), desktop PC, TV broadcast receiver, TV program recorder.

  The terminal device 30 receives the item or information related to the item transmitted from the item providing server 20, and outputs an image or sound indicating the item or information related to the item or a combination thereof from an output device (for example, a display or a speaker). Operate. Further, the terminal device 30 receives recommendation information including a recommended category and a recommended item from the information selection device 10 or the item providing server 20, and outputs an image or sound indicating the recommended category and the recommended item, or a combination thereof as an output device (for example, It operates to output from the display and speakers.

  The network 40 is a communication network or a combination of a communication network and a broadcast network. The network 40 may be a wired network, a wireless network, or a combination thereof.

Note that the configuration example of FIG. 1 is merely an example relating to the usage mode of the information selection device 10. For example, the system 1 including the information selection device 10 may be configured as shown in FIG. In the example of FIG. 2, the information selection device 10 communicates with the item providing server 20 via a network 42 different from the network 40. The network 42 is, for example, a LAN (Local
Area network), and direct access from the terminal device 30 to the information selection device 10 may be restricted from the viewpoint of ensuring security.

  In the example of FIGS. 1 and 2, the information selection device 10 is physically separated from the item providing server 20 and the terminal device 30. However, the information selection apparatus 10 may be physically integrated with any one of these. That is, the information selection device 10 may be configured as a device (for example, a computer system) that is physically the same as the item providing server 20. The information selection device 10 may be configured as a device (for example, a computer system) that is physically the same as the terminal device 30.

  Subsequently, a configuration example of the information selection device 10 and an example of an operation for selecting a recommended category and a recommended item by the information selection device 10 will be described. FIG. 3 is a block diagram illustrating a configuration example of the information selection device 10. The information selection device 10 includes an information selection unit 107. The information selection unit 107 determines a recommended category recommended to the user. The information selection unit 107 may further determine recommended items. In the example of FIG. 3, the information selection unit 107 determines the usefulness when a category selection unit 1071 that selects a recommended category, an item selection unit 1072 that selects a recommended item, and an item is recommended. A recommendation usability determination unit 1073 that creates a recommended candidate item set that is a set of determined items is included. However, the selection of the recommended item corresponding to the recommendation category determined by the information selection device 10 may be performed by another device different from the information selection device 10, for example, the item providing server 20 or the terminal device 30.

  The recommended usefulness determination unit (useful item determination unit) 1073 is a set of items that are determined to be useful by determining the usefulness of each item as a recommended item. A recommended candidate item set is created. For example, an item that is already frequently used by the user can be assumed to be sufficiently recognized by the user, and thus it is unlikely that the user will use the item more than before even if it is recommended again. For this reason, it determines with the item with very high use frequency of a user having no usefulness (low usefulness). Further, in an item providing service in which the user often uses (for example, purchases and browses) the same item only once, it is determined that the item used in the past by the user has no recommended utility. Further, an item that has been provided to the user in the past but is difficult to provide to the user in the future, for example, an item for which production / sale has been stopped or an item that is not in stock, is determined to have no recommended utility. Further, regarding items provided from the item providing server 20 for a limited time, the recommendation usefulness is determined in consideration of the time and period when recommendation information is provided to the user and the period during which the item can be provided.

For example, it is determined that there is no usefulness for an item whose provision period has ended or an item whose provision period is in the future. Then, an item determined to have recommended usefulness (an item other than an item determined to have no recommended usefulness) is added to the recommended candidate item set.
When expressing the presence / absence or degree of usefulness, a binary expression such as “yes / no” may be used, or it may be expressed as three or more multi-values, or expressed as a numerical value. Also good. For example, the usefulness of an item may be expressed by a numerical value (usefulness), and an item whose value is a predetermined value or more may be included in the recommended candidate item set. Alternatively, a predetermined number of items may be selected in descending order of usefulness, or a predetermined ratio (ratio to the total number of items) may be selected in descending order of usefulness, and the selected items may be included in the recommended candidate item set. .

  The category selection unit 1071 operates to select a recommended category for the use subject identifier from the plurality of candidate categories based on the category priority calculated for each of the plurality of candidate categories. Here, each of the plurality of category priorities is calculated using an item relevance level that indicates the strength of relevance between an item belonging to the candidate candidate item set and each candidate category and the user. That is, the category priority is used as a scale for determining the recommendation category and the ranking of the recommendation category. When the category selection unit 1071 calculates the category priority, the category selection unit 1071 considers the item relevance between the items belonging to the category and the use subject.

  The category selection unit 1071 determines a recommended category from among a plurality of candidate categories in consideration of the category priority. For example, the category selection unit 1071 may preferentially select a category with a high category priority as a recommended category. For example, the category selection unit 1071 may select a predetermined number of categories as recommended categories in descending order of category priority. Further, the category selection unit 1071 may select a category having a category priority of a predetermined value or more from among a plurality of candidate categories as a recommended category.

  Subsequently, various parameters and information required for calculating the category priority and an example of calculating the category priority will be described below. The item relevance necessary for calculating the category priority may be derived by the information selection device 10 or supplied to the information selection device 10 from the outside. For example, the item relevance determined by the administrator or operator may be supplied to the information selection device 10. The item relevance level is, for example, from an external device to the information selection device 10 as a table (ie, item relevance level table) in which a user entity identifier, an item identifier, and an item relevance level indicating the relevance level of the two are associated It may be supplied.

  In order to derive the degree of item relevance, the information selection device 10 or an external device may analyze the relevance, similarity, preference, or correlation between the user and the item. Specifically, the information selection device 10 or an external device uses a plurality of usage subjects (users) history (for example, usage history, purchase history, browsing history, viewing history, evaluation history, or search history), The item relevance level may be calculated. For example, the information selection device 10 or an external device calculates the number of items that a certain user x and another user x2 have used together, and the greater the number, the higher the similarity between users. The similarity between users is calculated. Next, with respect to the user subject x, a similar user set that is a set of other user subjects having a high similarity between users is created. Then, the item y used by the user who belongs to the similar user set is extracted, and the use frequency D1 for which the user who belongs to the similar user set uses the item y is calculated. Then, the use count D1 is set as an item relevance W [x] [y] between the use subject x and the item y.

  For items that are not used by a user who belongs to a similar user set, the degree of association may be set to “0”. The information selection device 10 or an external device calculates the similarity between users of the user x and the other user x2 by using a co-occurrence scale used in the field of statistics. May be. For example, Jaccard coefficient, Simpson coefficient, cosine measure (cosine distance), or Pearson product moment correlation coefficient for two sets of items used by user x and items used by user x2 The co-occurrence measure such as can be calculated, and the similarity between users can be calculated using the measure.

  Further, the total number D2 of the number of times that the use subject belonging to the similar user set uses any item is calculated, the use subject belonging to the similar user set calculates the use count D1 using the item y, and D1 is represented by D2. The divided value (D1 ÷ D2) may be used as the item relevance W [x] [y]. Further, without limiting the use subject to the similar user set, any use subject calculates the use count D3 of the item y, and the value obtained by dividing D1 by D3 (D1 ÷ D3) is the item relevance W [ x] [y] may be used. Also, the item relevance W [x] [y] may be calculated using the number of users who use the item y (the number of users) among the users who belong to the similar user set. In addition, although the item relevance may be calculated for all items, it is more efficient to calculate the item relevance only for items belonging to the recommended candidate item set.

  The correspondence relationship between each candidate category and item (hereinafter, item-category correspondence relationship) may be derived by the information selection device 10 or may be supplied to the information selection device 10 from the outside. The item-category correspondence may be defined in advance as, for example, a table in which item identifiers and category identifiers are associated (that is, item-category correspondence table). The item-category correspondence may be defined by defining a category to which the item belongs as one of item attributes in a table or database indicating item information. One item may correspond to (belongs to) only one category, or one item may correspond to a plurality of categories without such a restriction.

  The information selection device 10 or an external device may analyze which category each item is similar to and derive an item-category correspondence. For example, the item-category correspondence may be derived using the dissimilarity degree indicating the strength of the relationship between each item and each category. Specifically, for each item, a category having a heterogeneous relevance level of a predetermined value or more is selected, and it is determined that the item belongs to the selected category. However, if there is no category greater than or equal to the predetermined value, it is determined that the item belongs to the category having the highest degree of dissimilarity. Thereby, one item belongs to at least one category. Another method is to select a predetermined number of categories in descending order of dissimilarity for each item, determine that the item belongs to the selected category, and determine that the item does not belong to any other category. Also good. Note that the predetermined number may be set to 1, and the correspondence relationship between the category having the highest degree of dissimilarity and the item may be derived. In this case, one item corresponds to only one category.

  The category priority used to determine the recommended category for the user reflects the degree of item relevance, which belongs to the recommended candidate item set and indicates the strength of the relationship between the item belonging to each candidate category and the user If you do. Accordingly, each of the plurality of category priorities may be calculated as a function of the item relevance. For example, each of the plurality of category priorities may be a value corresponding to a value obtained by adding the item relevance between each of the plurality of items belonging to the recommended candidate item set and each of the candidate categories and the user. Further, each of the plurality of category priorities is a value corresponding to the number of items belonging to the recommended candidate item set and having an item relevance with the main user exceeding a predetermined threshold among the plurality of items belonging to each candidate category. It may be. In addition, the maximum value of the item relevance regarding items belonging to the recommended candidate item set and belonging to each candidate category may be used. The plurality of category priorities may be calculated by the information selection device 10 (that is, the category selection unit 1071) or may be supplied to the information selection device 10 from the outside.

  FIG. 4 is a flowchart showing an example of a recommended category selection procedure by the category selection unit 1071. FIG. 4 shows an example in which the category selection unit 1071 calculates the category priority. In step S10, the category selection unit 1071 selects a processing subject user who is a target for providing recommendation information. As will be described in detail in the third embodiment, the information selection device 10 may receive a recommendation information request from the terminal device 30 and extract a use subject identifier included therein. In addition, when processing is performed at a timing asynchronous with the reception of the recommendation information request, in the database (user management database) that stores the registered user or terminal device information when the user starts using the item providing service. From the above, one or more user entity identifiers may be selected. Further, one or more user entity identifiers may be selected from user entity identifiers included in the user history (for example, user history, purchase history, browsing history, viewing history, evaluation history, or search history). .

In step S11, the category selection unit 1071 acquires a recommended candidate item set relating to the user selected in step S10.
In step S12, the category selection unit 1071 acquires the item relevance level of the item corresponding to the recommended candidate item set acquired in step S11 as the item relevance level between the user selected in step S10 and the item. As already described, the item relevance may be dynamically calculated by the information selection device 10 or may be set statically by an operator or the like.

  In step S13, the category selection unit 1071 calculates the category priority for each candidate category using the item relevance acquired in step S12. In step S14, the category selection unit 1071 determines a recommended category from a plurality of categories for the user selected in step S10 in consideration of the category priority. In step S10, a plurality of usage subjects may be selected as processing targets, and in each step of steps S11 to S14, processing may be performed for all the selected usage subjects. Further, in step S10, one user is selected, and in each step of steps S11 to S14, the process is performed on the one user, and after step S14 is executed, the process returns to step S10 as necessary. And processing for a plurality of users may be performed.

  As described above, in this embodiment, the category priority, which is a scale for determining the recommended category, reflects the magnitude of the degree of item relevance between each item belonging to each candidate category and the user. Therefore, there is a high possibility that the recommended category determined based on the category priority is associated with an item having relevance, similarity, preference, or correlation with the user. Therefore, according to the information selection apparatus 10 and the recommended category determination method using the information selection apparatus 10 described in the present embodiment, a recommended category that has a high possibility that a recommended item having a high relevance with a user is present is determined. be able to.

  Next, the item selection unit 1072 in FIG. 3 will be described. The item selection unit 1072 operates to select a recommended item from items belonging to the recommended candidate item set determined by the recommendation usability determination unit 1073 and belonging to the recommended category determined by the category selection unit 1071. The item selection unit 1072 may select a recommended item based on the item relevance level. Thereby, the possibility that an item having high relevance with the main user can be selected as a recommended item is improved.

  Specifically, the item selection unit 1072 may preferentially select, as a recommended item, an item that belongs to the recommended candidate item set and belongs to the recommended category and that has a high degree of item relevance with the user. For example, the item selection unit 1072 may select a predetermined number of items as recommended items in order from items that belong to the recommended candidate item set and belong to the recommendation category in descending order of the degree of item relevance with the user. In addition, the item selection unit 1072 may use, as a recommended item, an item that belongs to the recommended candidate item set and belongs to the recommended category and whose item relevance with the user is greater than or equal to a predetermined value.

  However, the item selection unit 1072 may appropriately include an item having a low item relevance among a plurality of recommended items. As a result, the item selection unit 1072 can select an item having a relatively low item relevance with the user as a recommended item compared to other items belonging to the recommended category, and therefore can recommend a highly unexpected item to the user. There is a case. This can potentially create new usage opportunities for items (eg, purchase opportunities, viewing opportunities).

<Second Embodiment>
In the present embodiment, a modified example of the recommended category determination procedure by the information selection apparatus 10 described in the first embodiment will be described. In the first embodiment, the example in which the category priority of each candidate category is obtained using the item relevance between each item belonging to the candidate category and the user subject has been described. On the other hand, the information selection apparatus 10 according to the present embodiment, in order to obtain the category priority of each candidate category, not only the item relevance but also the category relation indicating the strength of relevance between the use subject and the candidate category Further use degrees.

  The category relevance may be derived by the information selection device 10 or may be supplied to the information selection device 10 from the outside. As with the derivation of the item relevance, in order to derive the category relevance, the information selection device 10 or an external device analyzes the relevance, similarity, preference, or correlation between the user and the category. That's fine. For example, the category relevance may be calculated based on data indicating the degree of interest in each category of the user collected by questionnaire survey or the like. Further, the information selection device 10 or an external device uses the history of a plurality of users (users) (for example, usage history, purchase history, browsing history, viewing history, evaluation history, or search history) to determine the degree of category relevance. May be calculated. For example, the information selection device 10 or an external device calculates the number of items or categories that a certain user x and another user x2 have used together, and the greater the number, the more similar the degree of similarity between users. The degree of similarity between users is calculated so as to increase.

  Next, with respect to the user subject x, a similar user set that is a set of other user subjects having a high similarity between users is created. Then, referring to the item-category correspondence table, the category p corresponding to the item used by the user belonging to the similar user set is specified, and the user corresponding to the category p uses the item corresponding to the category p. Use count D5 is calculated. Then, the usage count D5 may be the category relevance H [x] [p] between the user x and the category p. For a category that is not used by a user belonging to a similar user set, the category relevance may be set to “0”.

  Further, a total value D2 of the number of times that the use subject belonging to the similar user set uses any item is calculated, and the use number D5 that the use subject belonging to the similar user set uses the item corresponding to the category p is calculated, A value obtained by dividing D5 by D2 (D5 ÷ D2) may be set as the category relevance H [x] [p]. Further, without limiting the use subject to the similar user set, any use subject calculates the use frequency D6 using the item corresponding to the category p, and the value obtained by dividing D5 by D3 (D5 ÷ D6) is the category. The degree of relevance may be H [x] [p]. In addition, the category relevance may be calculated using the number of users (number of users) using items corresponding to the category p among the users belonging to the similar user set.

  Note that when calculating the category relevance, all stored histories may be used, or when the use date is recorded in the history, only a relatively new history may be used. When all the histories are used, the long-term preference of the user is more easily reflected in the category relevance. On the other hand, when the number of user histories is large, it may be preferable to limit to only relatively new histories because the user's recent preferences are reflected in the category relevance. In addition, when a recommended candidate item set is created without excluding items used by the user (excluding items that have been discontinued), the recommended candidate item set is included in the history. The category relevance may be calculated by using item information only for the items to which it belongs.

  The category relevance calculation method using the above-described history can be said to be a method using the information of the item provided by the item providing server 20, but the category is based on information other than the items provided by the item providing server 20. The degree of association may be calculated. For example, it is possible to select the user's interest level for each category from “very interested”, “somewhat interested”, “cannot say”, “not much interested”, “not interested at all” You can choose from the answers and digitize the answers to obtain the category relevance. In the case of a questionnaire, it is better to have the user answer the item assortment of items in the item providing server 20 (what items are provided corresponding to each category, etc.) without much awareness of the user. The original interest level of the user category can be accurately reflected in the questionnaire data. Furthermore, the same category system is used as the category system used by the item information selection device 10 and the item providing server 20 (for example, “Rock”, “Jazz”, “Classic”, “Folk” is used as the music genre). If so, the category relevance may be calculated using questionnaire data or the like answered by the user using another service. Similarly, other services that have the same category system as that used by the information selection device 10 and the item providing server 20 and that use the same user entity identifier as that used by the information selection device 10 are used. If it exists, the category relevance may be calculated using the user history of the service. In that case, if other services provide items including items other than items provided by the item providing server 20 (non-provided items), information on non-provided items is also reflected in the category relevance. Become.

  The category priority can be obtained as follows, for example. The item relevance between the user x and the item y corresponding to the category p is W [x] [y], and the category relevance between the user x and the candidate category p is H [x] [p]. The product of W [x] [y] and H [x] [p] (W [x] [y] × H [x] [p]) corresponds to the candidate category p and recommends the use subject x A value V [x] [p] added for the item y belonging to the candidate item set may be calculated, and V [x] [p] may be set as the category priority of the candidate category p. A predetermined number of items corresponding to the candidate category p and belonging to the recommended candidate item set in descending order of W [x] [y], or items having W [x] [y] equal to or larger than a predetermined value Is extracted, and a value V [x] [p] obtained by adding (W [x] [y] × H [x] [p]) is calculated for the extracted item, and the value is calculated as a candidate category. It is good also as the category priority of p. Also, a value WA obtained by adding W [x] [y] to all items belonging to the candidate category p and corresponding to the candidate category p is calculated, and weighted addition of WA and H [x] [p] is performed. The value may be the category priority.

  Further, when the heterogeneous association degree S [y ′] [p] between the candidate category p and an arbitrary item y ′ belonging to the recommended candidate item set can be used, the recommendation is not limited to the item corresponding to the candidate category p. The category priority calculation process may be performed for all items belonging to the candidate item set. For example, the product of W [x] [y ′], H [x] [p] and S [y ′] [p] (W [x] [y ′] × H [x] [p] × S [ A value V ′ [x] [p] obtained by adding y ′] [p]) to all items y ′ belonging to the recommended candidate item set is calculated, and V ′ [x] [p] is given priority to the category of the candidate category p. It may be a degree. Further, the product of W [x] [y ′] and S [y ′] [p] (W [x] [y ′] × S [y ′] [p]) is set to all the recommended candidate item sets. A value V ′ [x] [p] added for the item y ′ may be calculated, and V ′ [x] [p] may be set as the category priority of the candidate category p. Furthermore, as described in the first embodiment, when the category priority is calculated without using the category relevance, the product of W [x] [y ′] and S [y ′] [p] A value V ′ [x] [p] obtained by adding (W [x] [y ′] × S [y ′] [p]) to all items y ′ belonging to the recommended candidate item set is calculated, and V ′ [ x] [p] may be set as the category priority of the candidate category p.

  In such a calculation method, a binary determination is not made as to whether or not the item y ′ corresponds to (belongs to) the candidate category p, and the heterogeneity relevance S [y ′] [p] is set to the item y. Is used as correspondence information between 'and the candidate category p. That is, the item relevance W [x] [y ′] between the user x and the item y ′ and the heterogeneity relevance S [y ′] [p], which is correspondence information between the item y ′ and the candidate category p, are used. The category priority is calculated. Note that if S [y ′] [p] is greater than 0, the item y ′ and the candidate category p belong, and if S [y ′] [p] is 0, it is determined that they do not belong. Good. Further, when S [y ′] [p] is equal to or greater than a predetermined value, it may be determined that the item y ′ belongs to the candidate category, and when it is less than the predetermined value, it may be determined not to belong. That is, it can be said that the information on whether an item belongs to a category is one of the correspondence information.

  The configuration example of the information selection apparatus 10 according to the present embodiment is the same as the configuration example (FIG. 3) shown in the first embodiment. FIG. 5 is a flowchart showing an example of a recommended category selection procedure by the information selection apparatus 10 (ie, the category selection unit 1071) according to the present embodiment. The processes in steps S20 to S22 in FIG. 5 are the same as the processes in steps S10 to S12 in FIG. In step S23 of FIG. 5, the category relevance between the user and each candidate category is acquired. As described above, the category relevance may be dynamically calculated by the information selection device 10 or may be statically set by an operator or the like. In step S24, the category selection unit 1071 calculates a category priority for each candidate category. In step S25, the category selection unit 1071 determines a recommended category from among a plurality of candidate categories in consideration of the category priority.

  As described above, the category priority used in the present embodiment reflects the magnitude of the item relevance between each item corresponding to each candidate category and the user, as in the first embodiment. . For this reason, this embodiment can reduce possibility that the item which has high relationship with a user main body does not exist in a recommendation category.

  Furthermore, in this embodiment, the example which calculates | requires a category priority as a function of the category relevance of a user main body and a candidate category was shown. As a result, a recommendation category that has a strong relationship, similarity, preference, or correlation with the user and that is highly likely to be associated with an item that has a high relationship with the user is determined. can do.

  Note that the category priority of the candidate category may be determined so as to increase as the category relevance between the use subject and the candidate category increases. Thereby, the category priority of a candidate category with strong relevance, similarity, palatability, or correlation with a user can be increased. However, the relationship between the category priority of the candidate category and the category relevance is such that the category priority of the candidate category having a low category relevance or a medium candidate category is higher than the candidate category having a very high relevance to the user. The degree may be determined to be high. Thereby, there is a possibility that highly recommended recommendation information can be provided to the user. For example, if a category corresponding to a large number of items having a high degree of item relevance and a category relevance not so high is recommended, the user is likely to be surprised by the recommendation information. The category priority of the candidate category may be calculated as a function of the extended relevance obtained by inputting the category relevance with the user to a predetermined conversion rule. A specific example of the expanded relevance will be described in detail in a third embodiment described below.

<Third Embodiment>
In the present embodiment, a specific example of the recommended category determination procedure by the information selection device 10 described in the second embodiment will be described. In this embodiment, a specific example of calculation of category priority and item relevance necessary for determining a recommended category will also be described. Furthermore, in this embodiment, some variations regarding the category priority calculation method will be described.

<Overall system configuration>
A configuration example of the system 1 including the information selection apparatus 10 according to the present embodiment is the same as the example described in the first embodiment (for example, FIGS. 1 and 2). Hereinafter, the configuration example illustrated in FIG. 1 will be described as an example.

<Information selection device>
FIG. 6 is a block diagram illustrating a configuration example of the information selection device 10 according to the third embodiment. The information selection device 10 includes the information selection unit 107 as described in the first embodiment. Further, in the configuration example of FIG. 6, the information selection device 10 includes a metadata storage unit 101, a usage history storage unit 102, a category relevance storage unit 103, a relevance calculation unit 104, an item relevance storage unit 105, and display control information creation. Unit 106, recommendation information storage unit 108, transmission / reception unit 109, and control unit 110. In addition, the information selection device 10 in FIG. 6 includes a display device 120 (for example, LCD (Liquid) for displaying information necessary for the administrator of the information selection device 10.
A crystal display), an OELD (Organic Electroluminescent Display), a CRT (Cathode Ray Tube) display) and an input device 130 (for example, a keyboard, a mouse, and a touch panel) for operation by an administrator are connected.

As described above, the information selection apparatus 10 may be configured using a general computer system. For example, the information selection device 10 includes a CPU, a RAM, a ROM, a nonvolatile storage device (for example, an HDD (Hard
Disc drive), flash memory), network interface, etc., can be used for configuration. That is, a general computer functions as the information selection device 10 according to the present embodiment by executing a computer program for performing processing / algorithms related to selection of recommended categories and recommended items as described below. Will be able to.

  The metadata storage unit 101 stores item information, category information, and information indicating the correspondence between items and categories. The item information includes an item identifier and item attributes (for example, item name, item description, category to which the item belongs). The category information includes a category identifier and category attributes (category name, category type, category description). The correspondence relationship between items and categories may be defined as an item-category correspondence table as described in the first embodiment. Further, the correspondence between items and categories may be defined by defining the category to which the item belongs as one of the item attributes in the item information.

  Fig.7 (a) has shown an example (101A-1) of the item information table 101A which records item information. As shown in the figure, the item information table 101A is a table that associates item identifiers (item IDs) with item attribute information. In the example of FIG. 7A, the item attribute information includes “title (item name)”, “description”, and “item time” of the item. “Description” of the item attribute information indicates an outline or summary of the item, a background explanation produced, and the like.

  The “item time” is composed of three items: an item creation time (time), an item provision start time (time), and an item provision end time (time). When there is no information about the creation time of the item, the time when the item was registered in the item providing server 20 may be stored instead, or “NULL” indicating no data may be registered. The item provision start time is a point in time when provision of the item is started from the item provision server 20. Further, the item provision end time is a point in time when the item provision from the item provision server 20 ends. When the provision end time is not yet determined, a distant future date may be stored, such as “9999/12/31”. In the example of FIG. 7A, a date such as “January 1, 2010” is used as the expression format of the time (time), but other expression formats may be used. For example, a date and time up to a second unit such as “January 1, 2010 10:15:20” or a date up to a millisecond unit may be used. Alternatively, a monthly expression format such as “January 2010”, a quarterly expression format such as “2010 1Q”, an annual expression format such as “2010”, “2000s”, etc. The expression format may be a rough age than the year unit. By storing the information about the provision start time and the provision end time, even if the item is provided for a limited time, the information can be appropriately included in the recommendation information.

  FIG. 7B shows another example (101A-2) of the item information table 101A. The difference between the tables 101A-2 and 101A-1 is that category information (for example, category name, category identifier) is stored as one item attribute information. In the example shown in FIG. 7B, a maximum of two categories are associated with one item. Of course, three or more categories may correspond to one item, or only one category may correspond to one item. The category 2 of the item identifier “ItemID-2” in FIG. 7B is associated with “NULL” which means that there is no corresponding data. In the following description of the present embodiment, the item information table 101A-1 is used. However, when the item information table 101A-2 is used instead, the item-category correspondence table 101C described later is omitted. Can do. In addition, the “item time” in FIG. 7B is also the same as FIG. 7A, in three items of the item creation time (time point), the item provision start time (time point), and the item provision end time (time point). Although it consists of items, details are omitted in this figure.

  FIG. 7C shows an example of a category information table 101B that records category information. As shown in the figure, the category information table 101B is a table that associates category identifiers (category IDs) with category attribute information. The category means an index for classifying items as described above. At least one category is set for one item. In other words, one item is associated with at least one category.

  In the example of FIG. 7C, the category attribute information includes “category type”, “category name”, and “category description”. The category type (category type) represents the type of category (major classification). For example, an item “creator” can be a category type. The category belonging to the category type = “Creator” includes item producers, directors, producers, writers, composers, songwriters, performers, performers, and the like. That is, the names of directors and producers correspond to “category names”.

  Further, “genre” may be used as the category type. For example, when the item is music content, the category (category name) belonging to the category type = “genre” includes rock, jazz, classic, folk, blues, and the like. When the item is movie content, the category (category name) belonging to the category type = “genre” includes SF, action, comedy, animation, suspense, and the like.

  Further, “region” may be used as the category type. For example, category names belonging to category type = “region” include country or region names such as Japan, the United States, the United Kingdom, France, and Tokyo.

  Further, as the category type, “mood” representing an atmosphere of an item or an atmosphere suitable for using the item may be used. In this case, the category (category name) belonging to the category type = “mood” may be terms such as healing, exciting, dramatic, simple, and astringent.

A “situation” indicating a situation suitable for using an item may be a category type. For example, the category (category name) belonging to the category type = “situation” includes before sleeping, jogging, train, driving, studying, and the like. When the item is music content and “mood” or “situation” is used as a category type, the information selection device 10 or another device is disclosed in JP 2007-322598 (JP).
2007-322598 A) Music may be classified into categories using the technology disclosed in A).

  Further, the category group related to the price range and discount rate of the item may be classified into one category type. For example, the category (category name) belonging to the category type includes high-priced products, middle range price ranges, bargains, 20% OFF, 50% OFF, and the like. Further, a category group related to the size and color of the item may be classified into one category type. Of course, the item attribute information and the category attribute information given here are merely examples, and are not limited to these examples. For example, the item attribute may include item price information.

  FIG. 7D shows an example of the item-category correspondence table 101C. The item-category correspondence table 101C is a table for associating item identifiers with category identifiers, and an arbitrary number of category identifiers can be associated with one item identifier. The item-category correspondence table 101C can associate any number of item identifiers with one category identifier. In the example shown in FIG. 7D, three category identifiers correspond to “ItemID-1”, one “ItemID-2”, and two “ItemID-3”. When a plurality of categories correspond to one item, the category types may be the same or different. Although not shown in FIG. 7D, in addition to the item identifier and the category identifier, a heterogeneous degree of association that is the degree of association between the item and the category may be stored in the item-category correspondence table 101C.

  The heterogeneous degree of association may be supplied from an external device to the information selection device 10 as a table in which the item identifier, the category identifier, and the heterogeneous degree of association are associated (ie, a heterogeneous association degree table). In addition, in order to derive the degree of dissimilarity, it may be derived using a category dictionary describing keywords for each category as described in Patent Document 1. For example, a value obtained by dividing the number of appearances for each category of keywords (for example, product name, product description, program name, performer name) included in each item by the total number of appearances of the keyword included in each item (keyword for each category Relative appearance rate) may be used as the degree of dissimilarity.

  As another method, the dissimilarity degree is obtained by using a usage history corresponding to a user (user) and an item used by the user, and a degree of interest in each user (user) and each category (category relevance). ) May be calculated using category interest data collected by questionnaire survey or the like. Specifically, for a user (user group B) using a certain item (item A), the category relevance total of the user group B is calculated for each category (total by category), and the total by category is calculated. A value obtained by dividing the above by the total of the category relevance level in the user group B may be used as the heterogeneous relevance level. For example, the item A is used by three users (U1, U2, U3), the degree of association of U1 with respect to category 1 is “1.0”, the degree of association with category 2 is “0.5”, and U2 The degree of association with category 1 is “0.8”, the degree of association with category 2 is “0.3”, the degree of association with category 1 of U3 is “0.0”, and the degree of association with category 2 is “0.4”. In some cases, the total relevance of category 1 is “1.0 + 0.8 + 0.0 = 1.8”, and the total relevance of category 2 is “0.5 + 0.3 + 0.4 = 1.2”. Further, since the total sum of the relevance levels of the three users is “1.0 + 0.5 + 0.8 + 0.3 + 0.0 + 0.4 = 3.0”, the dissimilarity relevance level between the item A and the category 1 is “1.8 ÷ 3.0 = 0.6 ”and the dissimilarity between item A and category 2 is“ 1.2 ÷ 3.0 = 0.4 ”.

Instead of calculating the total value of category relevance, the number of users (supporting users) whose category relevance is a predetermined value (for example, “0.5”) or more is counted for each category, and the number of supporting users Is divided by the total number of users using the item A, and the degree of dissimilarity between the item A and each category may be used. For example, in the above example, if the number of users having a relevance of 0.5 or more is counted, the number of support users in category 1 is “2”, and the number of support users in category 2 is “1”. The dissimilarity between category 1 is “2 ÷ 3 = 0.667”, and the dissimilarity between item A and category 2 is “1 ÷ 3 = 0.333”. Further, the degree of heterogeneity may be calculated using the number of times the user has used the item.
When the dissimilarity degree is omitted, all combinations of items and categories stored in the item-category correspondence table 101C correspond to the dissimilarity degree having a constant value (for example, “1”).

  As described in the first embodiment, the information selection device 10 may acquire information (that is, metadata) related to an item or a category from an external device as necessary. In this case, the metadata storage unit 101 may be omitted. For example, the information selection device 10 may acquire metadata from the item providing server 20 (e.g., an item storage unit 202 described later).

  The transmission / reception unit 109 transmits / receives data to / from the item providing server 20 or the terminal device 30 via the network 40 (in the case of the configuration example of FIG. 2, the network 42).

  The control unit 110 performs various processes for performing overall control of the information selection device 10. For example, as described later, the control unit 110 receives a usage request transmitted from the item providing server 20 or the terminal device 30 via the transmission / reception unit 109, and associates a user identifier and an item identifier included in the usage request with a usage history. Information is stored in the usage history storage unit 102.

  The usage history storage unit 102 stores information indicating an item usage history by the user. The item usage history may be stored, for example, in a table format (referred to as item usage history table 102). In the present embodiment, item use is executed by the item providing server 20 providing an item in response to a use request from a user. In the present embodiment, an example in which a user is identified using a user identifier (user ID) as a user entity identifier will be mainly described. However, when a mobile phone is used as the terminal device 30, for example, a terminal identifier (terminal ID) ) May be used instead of the user identifier. The terminal identifier can be obtained, for example, when the information selection device 10 or the item providing server 20 and the terminal device 30 are connected. For example, when a personal computer or the like is used as the terminal device 30, the Web browser or the terminal device 30 may be identified using Cookie or similar identification information instead of the user identifier. As described above, the user or the terminal device used by the user is referred to as a “use subject”, and the user identifier and the terminal identifier are collectively referred to as a “use subject identifier”.

  The item usage history table 102A can adopt various storage forms for storing usage history information. For example, as shown in the item use history table 102A-1 in FIG. 8A, the use subject identifier and the item identifier can be stored in association with each other. In this example, one usage request corresponds to one row of the table 102A-1. As can be seen from the combination of “UserID-1” and “ItemID-3” in the first and fourth rows of the table, even if the combination of the use subject identifier and the item identifier is the same, The table row data is added to. For this reason, other processing units (for example, the relevance calculation unit 104) can easily count the number of uses for each item indicated by the item identifier and the number of users used for each item. When a plurality of item identifiers are included in one use request, as many table rows as the number of item identifiers may be allocated and stored.

  The item use history table 102A-2 illustrated in FIG. 8B is an example in which a use subject identifier, an item identifier, and an item use time are stored in association with each other. Similar to the item usage history table 102A-1 shown in FIG. 8A, one usage request corresponds to one row of the table 102A-2. When the use request indicates the item use time, the control unit 110 stores the item use time information extracted from the use request in the table 102A-2. When the item usage time information is not included in the usage request, the control unit 110 uses the time (time point) when the information selection device 10 receives the usage request, for example, using a clock built in the information selection device 10. What is necessary is just to use as time information.

  In the example of FIG. 8B, the date and time up to the second unit such as “January 1, 2010 10:15:20” is used as the expression format of the item use time. However, the item use time may be expressed in an arbitrary format such as a date and time up to a millisecond unit, a date up to a day unit, a month unit, and a year unit. Note that the usage request may include an evaluation value (for example, a numerical value indicating the degree of likes and dislikes such as like = 3, neither is = 2, and dislike = 1) by the user. In this case, the item use history table 102A-2 may store the use subject identifier, the item identifier, the item use time, and the item evaluation value in association with each other.

  An item use history table 102A-3 illustrated in FIG. 8C is an example in which a use subject identifier, an item identifier, and an item use count are stored in association with each other. As will be described later, when the relevance calculation unit 104 does not use the item usage time, the storage capacity of the usage history storage unit 102 can be reduced by using the item usage history table 102A-3. When the evaluation value (item evaluation value) for the item by the user is included in the usage request, the item usage history table 102A-3 associates the usage subject identifier, the item identifier, the item usage count, and the latest item evaluation value. May be stored.

  The category relevance level storage unit 103 stores the category relevance level for the combination of the use subject identifier and the category identifier. The category relevance may be stored in a table format (referred to as a category relevance table 103A), for example. FIG. 9A shows a table 103 A- 1 that is an example of the category relevance table 103. The table 103A-1 associates the use subject identifier, the category identifier, and their category relevance. The category relevance can be represented by a numerical value of 0 or more and 1 or less, for example. Here, when the value of category relevance is 1, it indicates that the relevance, similarity, preference or correlation between the user (user) and the category is the highest, and when it is 0, the lowest Represents. For example, FIG. 9A shows that the category relevance between “UserID-1” and “CategoryID-2” is “0.85”.

  FIG. 9B shows a table 103A-2 which is another example of the category relevance table 103A. The table 103A-2 is a matrix format table. One of the rows and columns of this matrix corresponds to the user identifier, and the other corresponds to the category identifier. The category relevance table 103A may be created by other methods. In the present embodiment, as will be described later, the relevance calculation unit 104 calculates the category relevance using the data stored in the usage history storage unit 102, and creates the category relevance table 103A. However, the category relevance may be calculated using other methods.

  The recommendation information storage unit 108 records recommendation information including the recommendation category and the recommended item selected by the information selection unit 107. The recommendation information may be stored, for example, in a table format (referred to as a recommendation information table 108A). The recommendation information table 108A associates a user identifier (use subject identifier), a recommendation category identifier, and a recommended item identifier. In addition to these identifiers, the table 108A may record at least one of a recommendation category recommendation rank and a recommendation item recommendation rank.

  FIG. 10 shows an example of the recommendation information table 108A. The use subject identifier corresponds to an item identifier included in a recommendation request that is a trigger for outputting recommendation information. The recommendation request will be described later. The recommended category identifier is an identifier of a category recommended corresponding to the use subject identifier. The recommended item identifier is an identifier of an item having a high degree of association with the user in the item group belonging to the recommended category. That is, the recommended category identifier and the recommended item identifier associated in the recommendation information table 108A are also associated in the item-category correspondence table 101C.

  As shown in the example of FIG. 10, the number of recommended category identifiers (recommended categories) may be different for each use subject identifier. Further, the number of recommended item identifiers (recommended items) corresponding to each recommended category identifier (recommended category) may be different. There may be only one recommended category identifier corresponding to a certain user entity identifier. However, in the present embodiment, it is desirable that the recommendation information table 108A includes one or more user entity identifiers having a plurality of recommended category identifiers.

  The category order in FIG. 10 indicates the recommendation order of the recommended category for each user entity identifier. In the example of FIG. 10, a category with a lower category ranking number has a higher priority and is preferentially presented to the user. The item ranking indicates the recommendation ranking of the recommended item for each combination of the use subject identifier and the recommended category identifier. In the example of FIG. 10, it is assumed that an item with a smaller item rank number has a higher priority rank and is preferentially presented to the user. In FIG. 10, recommended category identifiers are stored in descending order of category ranking for each user entity identifier. However, when storing the recommended category identifier in association with the category order, the recommended categories may be stored in an appropriate order. The same applies to the storage order of recommended item identifiers.

  Instead of the category ranking and item ranking, a recommendation level indicating that the higher the numerical value, the higher the priority ranking may be used. That is, a category or item having a higher recommendation level is preferentially presented to the user. Further, the recommendation information table 108A may be created according to a rule for storing recommendation category identifiers in descending order of recommendation ranking (or in descending order of recommendation ranking) for each use subject identifier. Thereby, the storage of the category recommendation order in the table 108A can be omitted. Further, the item recommendation order may be omitted by applying the same rule to the recommended item identifier. Alternatively, all recommended category identifiers (or all recommended item identifiers) recorded in the table 108A may be treated as the same recommendation order, or for the recommended categories (or recommended items) read from the table 108A. The recommendation order may be given at random.

  The degree-of-association calculation unit 104 uses the data stored in the usage history storage unit 102 to calculate the item degree of association for the combination of the use subject and the item, and stores them in the item association degree storage unit 105. As described in the first embodiment, the item relevance may be defined in advance by an administrator or an operator, or may be supplied to the information selection device 10 from an external device. In this case, the degree-of-association calculation unit 104 may be omitted.

  Further, the relevance calculation unit 104 may calculate a category relevance level for a combination of a user and a category, and store the calculated category relevance level in the category relevance storage unit 103.

  The item association degree storage unit 105 stores the item association degree for the combination of the use subject and the item calculated by the association degree calculation unit 104. The item association degree may be stored in a table format (referred to as an item association degree table 105A), for example. FIG. 11 shows an example of the item association degree table 105A. In the example of this figure, L1 related item identifiers corresponding to the user entity identifier “UserID-1” and L2 related item identifiers corresponding to the user entity identifier “UserID-2” are stored. Here, L1 and L2 may be the same number or different.

  That is, the table 105A may store the same number of related item identifiers for all the user entity identifiers, or may store an arbitrary number of related item identifiers for each user entity identifier. Further, the table 105A may store the item relevance level for all combinations of the use subject and the item calculated by the relevance level calculation unit 104, or a relationship having a relatively high item relevance level with a certain use subject identifier. Only the item identifier may be stored as a related item set. By storing only a part, the storage capacity of the item association degree storage unit 105 can be reduced. Further, as shown in FIG. 11, the table 105A may store related item identifiers in descending order of item relevance for each use subject identifier.

  The number of related item identifiers corresponding to one user entity identifier is basically plural, but may be one. However, it is desirable that the related item identifier is 2 or more in at least one user entity identifier. In addition, the item relevance level calculated by a device other than the information selection device 10 may be recorded in the item relevance level table 105A, and in this case, the relevance level calculation unit 104 may be omitted.

  The information selection unit 107 uses the data stored in the category relevance storage unit 103 and the item relevance storage unit 105 to select a recommended category and a recommended item to be presented to the user. Then, the information selection unit 107 creates recommendation information including a recommendation category and a recommended item, and stores the recommendation information in the recommendation information storage unit 108 (for example, the recommendation information table 108A).

<Item provision server>
The item providing server 20 provides an item or information related to the item in response to a request from the terminal device 30. FIG. 12 is a block diagram illustrating a configuration example of the item providing server 20. In the example of FIG. 12, the item providing server 20 includes a user management unit 201, an item storage unit 202, a data storage unit 203, a transmission / reception unit 204, and a control unit 205.

The item providing server 20 may be configured using a general computer system. For example, the item providing server 20 includes a CPU, a RAM, a ROM, a nonvolatile storage device (for example, an HDD (Hard
Disc drive), flash memory), network interface, etc., can be used for configuration. That is, a general computer can function as the item providing server 20 by executing a program for performing processing / algorithms related to provision of items or item information as described below.

  The transmission / reception unit 204 transmits / receives data to / from the information selection device 10 and the terminal device 30 via the network 40 (in the case of the configuration of FIG. 2, the network 42). The control unit 205 controls the entire item providing server 20.

  The user management unit 201 stores a user identifier for identifying a user who uses the terminal device 30 or a terminal identifier for identifying the terminal device 30, that is, a use subject identifier. The item providing server 20 performs, for example, an admission process before the user starts using the item, and stores a use subject identifier related to the user who has completed the admission process in a storage unit (not shown). In addition, the item providing server 20 may store user attribute information such as a login name, password, name, date of birth, contact information, and payment method in the user management unit 201 in association with the use subject identifier.

  The item storage unit 202 stores information related to items provided by the item providing server 20. The item storage unit 202 stores the same information as the metadata storage unit 101 of the information selection device 10. However, if the item is digital content and the item can be provided to the terminal device 30 via the network 40 (that is, content distribution), the item storage unit 202 stores the data in the metadata storage unit 101. In addition, the item identifier and the item main body (data such as digital contents) are stored in association with each other.

  Note that the control unit 205 transmits the data in the item storage unit 202 to the information selection device 10 via the transmission / reception unit 204 according to the update of the item storage unit 202 or according to a predetermined schedule, and the metadata. You may make it store in the storage part 101. FIG. On the contrary, the control unit 205 may receive the data in the metadata storage unit 101 from the information selection device 10 and store the data in the item storage unit 202. Alternatively, the information selection device 10 may transmit a message requesting item attribute information to the item providing server 20. Then, the control unit 205 may read data corresponding to the request message from the item storage unit 202 and transmit the data to the information selection device 10 via the transmission / reception unit 204.

  The data storage unit 203 can store various data. For example, the data stored in the recommended information storage unit 108 of the information selection device 10 can be copied and stored in the data storage unit 203. In this case, since the terminal device 30 can receive the recommendation information from the item providing server 20, the processing load on the information selection device 10 can be reduced. The data storage unit 203 may store data similar to the usage history storage unit 102 of the information selection device 10. In this case, it is possible to refer to the data storage unit 203 from the information selection device 10 and omit the usage history storage unit 102 of the information selection device 10.

<Terminal device>
The terminal device 30 is a device used by a user. FIG. 13 is a block diagram illustrating a configuration example of the terminal device 30. In the example of FIG. 13, the terminal device 30 includes a control unit 301, a transmission / reception unit 302, a browser unit 303, and an application unit 304. The terminal device 30 includes a CPU, a RAM, a ROM, a nonvolatile storage device (for example, an HDD (Hard
Disc drive), flash memory), network interface, etc., can be used for configuration. That is, a general computer can function as the terminal device 30 by executing a program for performing an algorithm and processing as described below. Further, the terminal device 30 as a computer system may be a mobile phone, a smartphone, a tablet computer or the like having a Web browser function.

  The control unit 301 performs overall control of the terminal device 30. The transmission / reception unit 302 transmits / receives data to / from the terminal device 30 or the information selection device 10 via the network 40 (in the case of the configuration of FIG. 2, the network 42 is further).

The browser unit 303 includes an HTTP (Hypertext Transfer Protocol) protocol and an HTTPS (HTTP over Secure Socket).
Layer (Hypertext Markup) from a server (for example, a WWW (World Wide Web) server) using a protocol such as Layer
Language) operates to acquire a Web page including a document, an image, and the like and output information based on the Web page to the display device 320. The browser unit 303 may further have a function of interpreting and executing a program described in a script language or the like. For example, the function of the browser unit 303 can be realized by the terminal device 30 as a computer system executing a Web browser program.

  The application unit 304 performs various processes in the application layer according to the usage of the terminal device 30 or a user instruction. The function of the application unit 304 can be realized by the terminal device 30 as a computer system executing various application programs.

  When the terminal device 30 is, for example, a personal computer, a display device 320 (for example, LCD, OELD, CRT display) and an input device 330 (for example, keyboard, mouse, touch panel, track) for receiving operation instructions from the user Ball, remote control) may be connected. When the terminal device 30 is a mobile phone terminal, a smartphone, a tablet computer, or the like, generally, the display device and the input device are integrally arranged on the terminal device 30. In the following description, it is assumed that the display device 320 and the input device 330 are connected to the terminal device 30 for convenience.

<System operation>
<Operation of the entire system>
With reference to the flowchart of FIG. 14, an example of operation | movement of the system 1 whole containing the information selection apparatus 10, the item provision server 20, and the terminal device 30 is demonstrated. First, in step S <b> 100, the terminal device 30 uses the browser unit 303 to use the URL (Uniform) of the item providing server 20.
Resource Locator). Specifically, a request (use start request) to a predetermined Web page provided by the item providing server 20 is transmitted to the item providing server 20.

When the terminal device 30 is, for example, a personal computer, the terminal device 30 may accept input of a login name (user ID) and a password by the user and transmit a use start request including these. In this case, before step S100, HTML (Hyper for accepting the input of the login name and password) is accepted.
Text Markup Language) data and the like may be transmitted from the item providing server 20 to the terminal device 30. Alternatively, in order to omit the transmission of the login name and the password, the terminal device 30 uses a cookie or a technique similar thereto to issue a use start request including data that can identify the user or browser using the terminal device 30. You may send it. Further, when the terminal device 30 is a mobile phone terminal, a smartphone, a tablet computer, or the like, a use start request including a terminal identifier may be transmitted in order to omit transmission of a login name and a password.

  In step S <b> 110, the control unit 205 of the item providing server 20 receives a use start request from the terminal device 30 via the transmission / reception unit 204, and determines whether the user is a registered user by referring to the user management unit 201. To do. Specifically, when the use start request includes a login name and a password, the control unit 205 may collate them with the login name and password stored in the user management unit 201. When the use start request includes a terminal identifier, the control unit 205 may determine whether or not the use start request matches the use subject identifier stored in the user management unit 201. If it is a registered user or terminal (Yes in S110), the process proceeds to step S130. If not (No in S110), the process proceeds to step S120.

  In step S <b> 120, the control unit 205 of the item providing server 20 transmits a Web page (HTML) for membership processing to the terminal device 30 via the transmission / reception unit 204. Although not shown in the figure, a user who uses the terminal device 30 uses the input device 330 to input necessary information on the Web page for the membership process, and transmits it to the item providing server. Thereby, the item providing server 20 performs a membership process including storing the information in the user management unit 201. The terminal device 30 may transmit a use start request again after the membership processing is completed. In each step after step S130, the information selection device 10 and the item providing server 20 acquire a user identifier of a user who uses the terminal device 30 or a use subject identifier which is a terminal identifier of the terminal device 30 as necessary. be able to.

  In step S <b> 130, the control unit 205 of the item providing server 20 creates Web page response data corresponding to the use start request while referring to the item storage unit 202, and transmits the response data to the terminal device 30 via the reception unit 204. . This response data includes information introducing at least one of a category and an item. The response data is composed of HTML data, image data, video data, audio data, and the like, and may be transmitted to the terminal device 30 in a plurality of times. The response data may include information for identifying the user or the terminal device 30 using a technique such as Cookie.

  In step S <b> 140, the terminal device 30 receives the response data from the item providing server 20 and displays the information on the display device 320. An example of the display screen is shown in FIG. The example of this figure is a display screen for introducing “new arrival items” that the item providing server 20 has recently started providing. Information for introducing an item as shown in the figure can be transmitted from the item providing server 20 to the terminal device 30 at various timings. Of course, the response data transmitted in step S130 is not limited to the page introducing the “new item” but may be various pages. It may be a page corresponding to the top page of a general Internet site.

  In FIG. 15, “Item ABC” is the title of the first item, “SF” is the category name to which the first item belongs, and the display “This item is a movie produced in 2001…” It is description (item description) regarding the first item. The example of FIG. 15 includes a display element 1401 that indicates a button, a link, or the like (hereinafter referred to as a recommended link) for providing recommendation information to a user who is using the terminal device 30. Further, a display element 1402 indicating a button or a link for using the first item (hereinafter referred to as a usage link) is included.

  The user can select a recommended link (display element 1401) or a usage link (display element 1402) by an operation such as clicking using the input device 330. Although not displayed on the display screen, the response data includes a use subject identifier, and the recommended link is associated with the use subject identifier. The response data includes an item identifier of each item, and each usage link is associated with the item identifier and the main subject identifier of the item to be selected.

  Returning to the flowchart of FIG. In step S150, the terminal device 30 determines whether the recommended link has been selected by the user via the input device 330. If the recommended link is designated (Yes), the process proceeds to step S160. If the recommended link is not designated (No), the process proceeds to step S190.

  In step S160, the terminal device 30 transmits a request (recommendation request) to the URL corresponding to the recommended link. In this embodiment, the case where the recommended link corresponds to a predetermined URL of the information selection device 10 will be described. However, the recommended link may correspond to a predetermined URL of the item providing server 20. The recommendation request includes a use subject identifier associated with the recommendation link. Hereinafter, the use subject identifier included in the recommendation request is referred to as a “request identifier”. Further, the recommendation request may include information regarding the number of necessary recommendation information (for example, the number of recommended categories, the number of recommended items of each recommended category, etc.).

  In step S170, the control unit 110 of the information selection device 10 receives a recommendation request via the transmission / reception unit 109, and displays display recommendation data (that is, display control information) corresponding to the request identifier included in the request. The creation unit 106 creates it. Then, the control unit 110 transmits the created display recommendation data to the terminal device 30.

  The display control information creation unit 106 identifies a user entity identifier that matches the request identifier while referring to the recommendation information table 108A of the recommendation information storage unit 108 illustrated in FIG. Is read. Further, the display control information creation unit 106 reads out a corresponding recommended item identifier and item order for each recommended category identifier. Furthermore, the display control information creation unit 106 reads out category attribute information (for example, category type, category name, category description) corresponding to the read recommended category identifier from the category information table 101B. Furthermore, the display control information creation unit 106 reads item attribute information (eg, title, item description, item time) corresponding to the read recommended item identifier from the item information table 101A of the metadata storage unit 101. Then, the display control information creating unit 106 creates display recommendation data using the read data.

  The display recommendation data (that is, display control information) is data described in HTML (Hyper Text Markup Language) that can be interpreted by the browser unit 303 of the terminal device 30. The display control information creation unit 106 may create the recommended display data so that the display position of the category attribute information of the recommended category changes according to the category order on the display device 320 connected to the terminal device 30. Since the category ranking is determined by the category priority, it can be said that the display position of the category attribute information is determined according to the category priority. In addition, the display control information creation unit 106 may create the recommended display data so that the display position of the item attribute information of the recommended item changes according to the item rank. The display recommendation data may include a request identifier.

  For reference, description will be made using the example shown in FIG. Consider a case where the request identifier is “UserID-10”. The display control information creation unit 106 includes three recommended category identifiers “CategoryID-100”, “CategoryID-102”, and “CategoryID-110” associated with the same user entity identifier as the request identifier (UserID-10). These category orders “1”, “2”, and “3” are read out. Furthermore, the display control information creation unit 106 includes three recommended item identifiers “ItemID-20”, “ItemID-25”, “ItemID-11” associated with “CategoryID-100”, and the item ranking “1”. ”,“ 2 ”,“ 3 ”are read out. The same applies to other recommended category identifiers. However, the display control information creation unit 106 may read all recommended category identifiers and recommended item identifiers corresponding to the specified user-identifying identifiers, or may read a predetermined number in descending order of category ranking and item ranking. When the recommendation request specifies the number of recommended information, the display control information creating unit 106 may read the recommended category identifier or the recommended item identifier in the order from the highest recommendation order.

  In step S170, the usefulness of the recommended item read from the recommendation information table 108A is determined again. If there is an item that has no usefulness, it is desirable to perform processing that does not include it in the display recommendation data. As will be described later, in the recommended information selection process, a recommended item is determined from the recommended item candidate set (useful item set) and stored in the recommended information table 108A, but when the recommended information selection process is executed. In general, there is a time difference from the time point until step S170 is executed, and the usefulness of the item may change during this period. For this reason, it is desirable to determine the usefulness of the item anew in step S170. For example, there is a possibility that there is an item that has been provided at the time when the recommendation information selection process is executed, but at the time when step S170 is executed. Further, when it is determined that an item used by a user more than a predetermined number of times is not useful according to the characteristics of the service, the number of times of use may change during the period. For example, when it is determined that an item used by the user once or more is not useful, the display control information creation unit 106 refers to the item usage history table 102A of the usage history storage unit 102 and corresponds to the request identifier. The item identifier (used item identifier) used by the user in the past may be specified, and the used item identifier (used item) may be excluded from the recommended information provided to the user. By performing such a process, highly accurate recommendation is possible in an item providing service in which a user often uses (for example, purchases and browses) the same item only once. For example, once purchased digital content is suitable for a service that can be repeatedly used (reproduced) by the terminal device 30.

  Returning to the flowchart of FIG. In step S <b> 180, the terminal device 30 receives display recommendation data from the information selection device 10. The browser unit 303 interprets the display recommendation data and displays the recommendation list on the display device 320 in the format shown in FIG. 16, for example. As shown in the figure, by using an expression such as “Specially recommended information recommended only for you”, it is clear that it is not personally provided information to all users but personalized information that is different for each user. It is effective in raising the user's interest. In addition, a simple explanation such as a background on which recommendation information such as “a bargain found by the system analyzing the trend of items used in the past” is created may be added. The numbers including “1)”, “2)”, and “3)” in FIG. 16 indicate the category rank (that is, the recommendation rank). The text display including “suspense”, “SF”, and “healing” in FIG. 16 indicates the category name of the recommended category. The display order (or display position) of the recommended categories in FIG. 16 is determined according to the category order. Specifically, the category with the higher category ranking may be displayed at a position that is more easily noticed by the user. For example, as shown in FIG. 16, when information on a plurality of categories is arranged in the vertical direction, a category having a higher category order may be displayed on the upper side of the display screen. In addition, when information on a plurality of categories is arranged in the left-right direction, a category having a higher category order may be displayed on the left side of the display screen. The text display including “genre” and “mood” in FIG. 16 indicates the category type name. As shown in the example of FIG. 16, the display screen based on the display recommendation data may include a plurality of category types.

  The recommended item information may be displayed so that the correspondence with the recommended category can be understood. In the example of FIG. 16, information on three items is displayed as recommended items corresponding to the recommended category “suspense”. The numbers 1 to 3 surrounded by circles in FIG. 16 indicate item ranks (that is, recommended ranks). The text display “Item OPQ” indicates the title of the item having the first recommendation order. The text display “This item ca n’t take your eyes off” indicates the explanation of the first item. In the example of FIG. 16, as in the example of FIG. 15, an “item use” button associated with the use link is displayed for each item. The display order (display position) of the attribute information of each item is determined according to the item order. Since the item ranking is determined according to the strength of the item relevance, it can be said that the display order of the item attribute information is determined according to the item relevance.

  Although not shown in FIG. 16, display elements (for example, buttons) for displaying more recommended items (related items) for each recommended category may be arranged. For example, a display element including text such as “See more recommended items of this genre (category)” may be arranged near the display “if you like the suspense (genre)” in FIG. And when a user operates the display element, you may make it display more recommended items of "suspense" corresponding to a request identifier. That is, the recommended items in the category specified by the user may be displayed. The recommended item of “suspense” displayed as the first recommendation list may be excluded and the second recommendation list may be displayed, or may be displayed in a newly added form without excluding them.

Further, information related to the degree of association indicating the strength of the relationship between categories is stored in advance in the storage unit inside the control unit 110, and a display element for displaying an item of a category having a high degree of association with the recommended category is displayed. May be. For example, when the degree of association between the “suspense” category and the “horror” category is high, “see horror recommended items” or “horror if you like suspense” are displayed near the display of “if you like suspense (genre)” in FIG. A display element including text such as “Please check it out” may be arranged.
Then, when the user operates the display element, a “horror” recommended item corresponding to the request identifier may be displayed.

  Returning to the flowchart of FIG. In step S <b> 190, the terminal device 30 determines whether a use link has been selected by the user using the input device 330. This usage link can typically be an item purchase request, but various items such as item playback, item preview, item detailed information display, or registration of evaluation information (evaluation value) for the item, etc. May be requested. If the usage link is selected (Yes), the process proceeds to step S200, and if not (No), the process proceeds to step S250.

  In step S200, the terminal device 30 transmits a request (use request) to the URL corresponding to the use link. In the present embodiment, a case where the usage link corresponds to a predetermined URL of the item providing server 20 will be described. Note that the terminal device 30 may directly transmit a use request to the information selection device 10 in addition to the item providing server 20.

  The use request includes an item identifier of an item selected by the user and a use subject identifier that identifies the user or the terminal device 30. When the user uses a plurality of items at a time, the terminal device 30 may transmit one usage request including item identifiers of the plurality of items, or a plurality of usages each including one item identifier. You may send a request.

  In step S <b> 210, the transmission / reception unit 204 of the item providing server 20 relays (transfers) the usage request received from the terminal device 30 to the information selection device 10. At this time, the control unit 205 of the item providing server 20 may extract information such as an item identifier and a usage subject identifier from the usage request and store them in the data storage unit 203 as usage information.

  In step S220, the control unit 110 of the information selection apparatus 10 receives the usage request via the transmission / reception unit 109, extracts information such as an item identifier and a usage subject identifier from the usage request, and uses the usage history storage unit as the usage history. 102 is stored. Then, the control unit 110 transmits a message indicating that the use history information has been stored to the item providing server 20 via the transmission / reception unit 109.

  In step S <b> 230, after receiving the storage end message from the information selection device 10, the control unit 205 of the item providing server 20 performs processing for providing the item to the terminal device 30. For example, when the item to be provided is digital content, the control unit 205 reads the item body corresponding to the item identifier indicated in the use request from the item storage unit 202 and transmits it to the terminal device 30 via the transmission / reception unit 204. To do. When the item is an article as a tangible object, a delivery process including sending delivery request information to the delivery company system is performed. At this time, the control unit 205 may perform billing processing. Further, when detailed item information is requested, the control unit 205 reads additional information such as “item description” from the item storage unit 202 and transmits the additional information to the terminal device 30.

  In step S <b> 240, the terminal device 30 performs a process related to the use of the item provided from the item providing server 20. For example, when the item is digital content, the item is reproduced and displayed. If the item is an article, a message indicating that the delivery process has been accepted is displayed on the screen.

  In step S250, the terminal device 30 determines whether or not there is an operation end instruction such as the user closing the browser. When there is an operation end instruction (Yes), the process of the terminal device 30 is ended, and when there is no operation end instruction (No), the process returns to Step S150 and the process is continued.

  The above is the description of the operation of the entire system. In step S160 in FIG. 14, the terminal device 30 transmits a recommendation request to the information selection device 10, but other methods may be used. For example, the terminal device 30 may transmit a recommendation request to the item providing server 20, and the item providing server 20 may relay (transfer) the recommendation request to the information selection device 10. In addition, the control unit 110 of the information selection device 10 may transmit the data stored in the recommendation information storage unit 108 to the item providing server 20 via the transmission / reception unit 109 at a periodic or aperiodic timing. . Then, the control unit 205 of the item providing server 20 may receive the data via the transmission / reception unit 204 and store the data in the data storage unit 203 in advance. And the control part 205 of the item provision server 20 reads recommendation data from the data storage part 203 as a process equivalent to step S170 for responding to a recommendation request, creates recommendation data for display, and sends it to the terminal device 30. You may make it transmit. In this case, it is possible to reduce the processing load of the information selection device 10 that accompanies creation and transmission of display recommendation data.

  Further, in step S210 of FIG. 14, the item providing server 20 relays the usage request from the terminal device 30 to the information selecting device 10, but other methods may be used. For example, the usage request may be transmitted directly from the terminal device 30 to the information selection device 10 at the same time as the usage request transmission in step S200 or at an appropriate timing.

  In addition, in step S220 of FIG. 14, in addition to storing the usage history information, the information selection device 10 creates display recommendation data corresponding to the usage subject identifier included in the usage request in the same manner as in step S170. Then, the display recommendation data may be transmitted to the item providing server 20. In step S230, the item providing server 20 may transmit the display recommendation data to the terminal device 30 in addition to performing the item providing process. That is, in this case, the terminal device 30 can receive recommendation information corresponding to the item identifier included in the usage request every time the usage request is transmitted. For example, it is suitable when the user wants to recommend another item immediately after purchasing one item.

  In addition, in the item providing service that does not require special user registration processing because a terminal identifier such as a mobile phone terminal can be used, the use request transmitted in step S200 may include a terminal identifier as a use subject identifier. Good. Thereby, it is possible to omit the registered user confirmation process in step S110 and the transmission of data necessary for the membership process in step S120.

<Operation of information selection device>
A specific example of the processing operation of the information selection device 10 in the third embodiment will be described. First, a specific example of the recommendation information creation operation will be described with reference to the flowchart of FIG.

  The control unit 110 of the information selection device 10 issues an instruction to each processing unit of the information selection device 10 at a predetermined timing, and starts a process of creating recommendation information. The timing of creating the recommendation information may be periodic or aperiodic. For example, at least one of the following three types can be used as the recommendation information creation timing.

  The first timing for creating the recommendation information is a predetermined date and time or a predetermined time interval. For example, “every day at 6 am and 6 pm”, “every Monday at 10:30 am”, “every 12 hours”, “every 24 hours”, and the like. At this time, the time interval fluctuates, such as “6 am on weekdays, 6 am and 6 pm on Saturdays and Sundays”, “every 3 hours on weekdays, every 6 hours on Saturdays, and every 12 hours on Sundays”. Also good. Also, the time interval may be changed according to the season, such as shortening the time interval in summer and increasing the time interval in winter. When this first timing is used, the processing load of the information selection device 10 can be reduced as compared with the case where other timings are used. In particular, if setting is made so that recommendation information is created in a time zone where the number of recommendation requests is small, it is effective in reducing the processing load of the information selection device 10.

  The second timing for creating the recommendation information is every time a recommendation request based on the recommendation request transmission process (step S160 in FIG. 14) of the terminal device 30 is received a predetermined number of times. In this case, the information selection device 10 may create recommendation information first, and then perform display recommendation data creation / transmission processing (step S170). By adjusting the predetermined number of times (that is, the number of receptions of recommendation requests until the recommendation information is updated), the balance between the processing load of the information selection device 10 and the newness of the recommendation information can be adjusted. . For example, if recommendation information is created each time a recommendation request is received with a predetermined number of times, the latest recommendation information can be provided, although the processing load on the information selection device 10 increases. In the following, for the sake of simplicity, when the second timing is used, it is assumed that recommendation information is created every time a recommendation request is received (a predetermined number of times is 1).

  The third timing for creating the recommendation information is every time a usage request based on the usage request transmission (step 200) of the terminal device 30 is received a predetermined number of times. When the usage request is generated less frequently than the recommendation request, the use of the third timing is more suitable than the second timing. By adjusting the predetermined number of times, it is possible to adjust the balance between the processing load of the information selection device 10 and the new recommendation information. If the predetermined number of times is set to one and recommendation information is created each time a use request is received, the processing load of the information selection device 10 increases, but the latest recommendation information can be provided. Hereinafter, in order to simplify the description, when the third timing is used, it is assumed that recommendation information is created every time a use request is received (a predetermined number of times is 1).

  In the following description, a set of usage subject identifiers for which recommendation information is created is referred to as a “recommended target set”. When the recommendation information is created at the first timing, the number of elements (the number of use subjects) of the recommended target set is basically large. When creating recommendation information at the second and third timings, if the predetermined number of times described above is 1, the number of elements in the recommended target set is one.

  First, in step S400, the degree-of-association calculation unit 104 that has received an instruction from the control unit 110 calculates an item degree of association between each element of the recommended target set and each item, Are stored in the item relevance storage unit 105.

  In step S410, the degree-of-association calculation unit 104 that has received an instruction from the control unit 110 calculates the degree of association between each element of the recommended target set and each category, and associates the user with the category and the calculated category degree of association. And stored in the category relevance storage unit 103.

  In step S420, the information selection unit 107 that has received an instruction from the control unit 110 creates recommendation information. Specifically, the information selection unit 107 selects a recommended category and a recommended item for each user belonging to the recommended target set while referring to the category relevance storage unit 103 and the item relevance storage unit 105. Recommendation information including information is stored in the recommendation information storage unit. Then, the controller 110 is notified that the recommendation information creation operation has been completed.

<Item relevance calculation (step S400)>
Next, a specific example of the item relevance calculation process (step S400) will be described with reference to the flowchart of FIG.

  In step S <b> 500, the relevance calculation unit 104 reads item usage histories by a plurality of users from the item usage history table 102 </ b> A of the usage history storage unit 102. Here, all item usage histories may be read, or usage histories that satisfy a predetermined condition may be read. For example, the association degree calculation unit 104 uses the usage history that satisfies the condition that “the item usage time is within a predetermined range” based on the usage history table 102A including the item usage time information as in the example of FIG. May be read out. The condition regarding the item use time can be specified as, for example, “the use time is within the past 4 months” or “the difference between the use time and the current time is 3 days or more and less than 30 days”.

  The degree-of-association calculation unit 104 may read out usage histories within a predetermined number in order of new usage time for each user. For example, when the predetermined number is set to 20, for an item having a usage count of 20 times or more, a usage history of 20 items is read out in order from the newest usage period, and the usage subject having a usage count of less than 20 is read. Reads all usage histories related to the user. In this way, it is possible to efficiently create a related item set even for a user who has not used the item recently and uses the item less frequently. Further, when the recommendation information is created at the second timing, the usage history is read so that the usage history corresponding to the request identifier included in the recommendation request is included. Further, when the recommendation information is created at the third timing, the usage history is read so that the usage history corresponding to the usage subject identifier included in the usage request is included.

  Then, a set σ of usage subjects (that is, usage subject identifiers) included in the usage history read out in step S500 is created. Hereinafter, the number of user entities (user entity identifiers) included in the set σ is assumed to be Us.

  In step S510, the relevance calculation unit 104 creates a recommended target set K1. As described above, when the recommendation information is created at the second timing, the request identifier (the use subject identifier included in the recommendation request) may be included in the recommended target set K1.

  When the recommendation information is created at the third timing, the relevance calculation unit 104 may enter the use subject identifier included in the use request into the recommended target set K1.

  When the recommendation information is created at the first timing, the set of use subjects created in step S500 may be set as the recommended target set K1. In this case, a related item set is created for each user entity identifier included in the set σ. The recommended target set K1 created here may be referred to by other processing units such as the information selection unit 107 and the control unit 110.

  In step S520, the relevance calculation unit 104 selects one unprocessed user from the recommended target set K1 created in step S510. The user subject to be processed is defined as a user subject x.

  In step S530, the relevance calculation unit 104 uses the usage history read in step S500, and the usage subject x and another usage subject x2 (x2εσ, x ≠ x2) belonging to the usage subject set σ. The similarity is calculated.

Specifically, the number | I [x] ｘI [x2] | of the items that the usage subject x and the other usage subject x2 have used together is calculated, and this is calculated for the usage subject x and the usage subject x2. The user similarity may be Su [x] [x2]. Further, it may be calculated using a Jacquard coefficient shown in the equation (1). In equation (1), I [x] indicates a set of items used by the user subject x, I [x2] indicates a set of items used by the user subject x2, and | I [x] ∩I [x2] | Indicates the number of items used by both the use subject x and the use subject x2, and | I [x] ∪I [x2] | indicates the number of items used by at least one of the use subject x and the use subject x2. Show.

Further, when additional information such as information on the number of item usage from the item usage history or information on evaluation performed by the user on the item (evaluation value) is included, the similarity between users is a cosine measure or a Pearson product-moment correlation coefficient. It may be calculated using For example, the inter-user similarity Su [x] [x2] may be calculated using a cosine measure as shown in Equation (2). Here, E [x] [y] indicates the number of use or evaluation value of the item y by the user (user) x, and E [x2] [y] indicates the number of use of the item y by the user (user) x2. An evaluation value is shown. Ms is the number of types of items (number of types of item identifiers) included in the item usage history read in step S500.

Further, the similarity between users Su [x] [x2] may be calculated using a Pearson product-moment correlation coefficient as shown in the equation (3). Here, Ic [x] [x2] is a set of items used or evaluated by the users x and x2. Ea [x] is an average value of the number of times that the use subject x has used an item belonging to Ic [x] [x2] or an average value of the evaluated evaluation values. Ea [x2] is the average value of the number of times that the user x2 has used the item belonging to Ic [x] [x2] or the average value of the evaluation values.

  Further, the inter-user similarity Su [x] [x2] may be calculated using the Euclidean distance between E [x] [y] and E [x2] [y] or other distances.

  In addition, when item usage time information is stored in the usage history storage unit 102, when calculating E [x] [y] and the like, the weight of the new usage history is set higher than the usage history with the old usage time. May be calculated. Specifically, first, the difference between the time (such as date and time) when the item relevance calculation process is executed and the item use time is calculated. This difference becomes a small value in the new item usage history, and this difference becomes a large value in the old usage history. Then, the output value when this difference is input to the monotonically decreasing function is used as a weighting coefficient, and a value obtained by multiplying E [x] [y] and E [x2] [y] by the weighting coefficient is used to Alternatively, the user-to-user similarity Su [x] [x2] may be calculated according to equation (3). That is, the value of E [x] [y] is increased as the usage time when the usage subject x uses the item y is newer.

  Further, the item time information (creation or provision start time) of the item y is used by using the information of the item creation time or the item provision start time stored in the item information table 101A (for example, 101A-1 or 101A-2). The user-to-user similarity Su [x] [x2] may be calculated using a weight that becomes larger as the value is newer. Specifically, first, the difference between the time (such as date and time) when the item relevance calculation process is executed and the item time information is calculated. This difference is a small value when the production or registration of the item y is new, and a large value when the item y is old. Then, the output value when this difference is input to the monotonically decreasing function is used as a weighting coefficient, and a value obtained by multiplying E [x] [y] and E [x2] [y] by the weighting coefficient is used to Alternatively, the user-to-user similarity Su [x] [x2] may be calculated according to the equation (3).

  Furthermore, the principal component analysis or the like is performed on a matrix whose matrix elements are E [x] [y] (x = 1 to Us, y = 1 to Ms), which is the number of uses or evaluation value for the item y of the user x. Multivariate analysis such as quantification type 3 may be applied to calculate the similarity between users. By applying multivariate analysis, it is possible to generate a vector corresponding to each user constituting the matrix and having a reduced number of dimensions, so that a cosine measure in the vector space of two vectors corresponding to any two users Or the Euclidean distance may be used to calculate the inter-user similarity Su [x] [x2]. In addition to the above, any method may be used as long as it is an index representing the similarity between two users.

In step S532, the relevance calculating unit 104 creates a similar user set Λ [x] corresponding to the user x based on the similarity between users calculated in step S530. The similar user set Λ [x] includes a use subject (use subject identifier) other than the use subject x.
The first method for creating a similar user set is a method in which a use subject x2 (x2εσ, x ≠ x2) having an inter-user similarity Su [x] [x2] of a predetermined value or more is placed in the similar user set.
The second method of creating a similar user set is a method of selecting the use subject x2 in a descending order of the similarity between users Su [x] [x2] and putting it in the similar user set. When the number of usage subjects x2 for which the similarity between users Su [x] [x2] is calculated is greater than a predetermined number, a predetermined number of usage subjects may be selected in descending order of similarity between users. Otherwise, all the users x2 for which the similarity between users is calculated may be included in the similar user set.
A third method for creating a similar user set is a method in which all the use subjects x2 for which the inter-user similarity Su [x] [x2] is calculated are put into the similar user set. Although this third method has the largest amount of processing, it can increase the types (variations) of items included in the related item set by combining with a third method for calculating the degree of item relevance described later.

  In step S534, the relevance calculation unit 104 calculates the item relevance W [x] [y] between the user x and the item y using the similar user set Λ [x] created in step S532. The first method for calculating the degree of item relevance is a method that uses the number of times the item y is used. Specifically, from the usage history data read out in step S500, data corresponding to a set of similar users is extracted from the usage subject, and the number of usages for each item (registration of usage history) is performed on the extracted data. Count). Then, regarding the item y used by the similar user set, the usage count is set as the item relevance W [x] [y]. For the item y that is not used by the similar user set, the item relevance may be set to “0”.

  The second method of calculating the item relevance is a method using the number of users (number of users) using the item y. Specifically, from the usage history data read out in step S500, data whose usage subject identifier corresponds to a similar user set is extracted, and the number of types of usage subject identifiers for each item for the extracted data. (Unique number of users) is counted. Then, regarding the item (item y) used by the similar user set, the number of types of the use subject identifier is set as the item relevance W [x] [y]. For the item y that is not used by the similar user set, the item relevance may be set to “0”. To explain a specific example, in a similar user set, only one user (user B) uses an item (item y) and user B uses item y three times. The item relevance level according to the first method is “3”, and the item relevance level according to the second method is “1”.

The third method for calculating the item relevance is a method using the similarity between users. Specifically, the item relevance W [x] [y] is calculated based on the equation (4). Here, Su [x] [r] is the similarity between users of the user subject x and the user subject r included in the similar user set Λ [x], and E [r] [y] is determined in step S500. In the read usage history data, the usage count r is the number of times the user y has used the item y. For the item y that is not used by the similar user set, the item relevance may be set to “0”. When the third method for creating a similar user set is used, it is preferable to use the third method for calculating the item relevance.

  In the first to third methods described above, the item relevance may be calculated using information on item creation time or item provision start time (time information) stored in the item information table 101A. . For example, a value obtained by multiplying the item relevance W [x] [y] obtained by the above-described method by a weighting factor that becomes a larger value as the time information of the item y is new is calculated, and the item relevance W [x] ] [Y] may be used in the following processing.

  In step S540, the relevance calculation unit 104 creates a related item set Ω [x] related to the user x based on the item relevance calculated in step S534, and stores it in the item relevance storage 105. The related item set Ω [x] is a set including item identifiers of item groups that are highly related to the user x.

  The first method for creating a related item set is a method in which all items whose degrees of association with the user x are calculated in the related item set Ω [x] in step S534. This method increases the amount of processing related to category priority calculation, which will be described later, but increases the number of items used for category priority calculation, so if you want to emphasize the diversity and recommendation accuracy of items included in recommendation information. Is suitable.

  The second method for creating a related item set is a method of selecting an item having a relatively high degree of relevance W [x] [y] with the user x and putting it in the related item set Ω [x]. Specifically, an item whose degree of association W [x] [y] with the user x is equal to or greater than a predetermined threshold is selected. In addition, items may be selected within a range not exceeding a predetermined value in descending order of the degree of association W [x] [y] with the user x. For example, when the number of items for which the degree of association W [x] [y] with the user x is less than a predetermined number, all items for which the degree of association W [x] [y] is calculated are selected. If not, a predetermined number of items may be selected in descending order of relevance W [x] [y]. According to the second method, the storage capacity required for the item relevance storage unit 105 can be reduced. In addition, it is possible to relatively reduce the amount of processing related to category priority calculation described later.

  Then, the relevance calculation unit 104 associates the use subject identifier of the use subject x with each item identifier included in the related item set Ω [x], and the item relevance level W [x] [y]. It records in the item association degree table 105 </ b> A of the item association degree storage unit 105. Specifically, the item identifier of the use subject x corresponds to the use subject identifier (use subject ID, user ID) in the item association degree table 105A shown in FIG. Each item identifier included in the related item set Ω [x] corresponds to a related item identifier (item ID) in the item relevance degree table 105A. In addition, when the use subject identifier which matches one of the identifiers included in the recommended target set K1 is already stored in the item relevance level table 105A, the relevance level calculation unit 104 may update (overwrite) them. On the other hand, when user entity identifiers that do not match any identifier included in the recommended target set K1 are stored in the relevance level table 105, the relevance level calculation unit 104 does not change them.

  In step S550, the relevance calculation unit 104 determines whether another user can be selected. The degree-of-relevance calculation unit 104 determines “Yes” when there is an unprocessed user in the recommended target set K1 created in step S510, and determines “No” when there is no unprocessed item. To do. The relevance calculation unit 104 returns to step S520 when it determines “Yes” and repeats the process, and ends the item relevance calculation process when it determines “No”. In the item relevance calculation process described above, the item relevance is calculated without using information indicating the relationship between items and categories.

  In step S400 (after execution of step S520, before execution of step S530, etc.), processing similar to step S730 described later is further performed to determine the usefulness of the item for the user x, and recommendation A candidate item set may be created. In step S534, the item relevance may be calculated only for items belonging to the recommended candidate item set. In step S534, the related item set may be created for items belonging to the recommended candidate item set in step S540, without being limited to the items belonging to the recommended candidate item set. Further, when the recommended candidate item set is created without excluding the items used by the user (when items that have been canceled are excluded), in step S530, the items are limited to items belonging to the recommended candidate item set. The similarity between users may be calculated using the information. Thus, by considering the recommendation candidate item set in the item relevance calculation process, it is possible to reduce the calculation amount and the storage capacity necessary for creating the recommendation information.

<Category relevance calculation (step S410)>
Next, the first method of the category relevance calculation process (step S410) will be described with reference to the flowchart of FIG.

  Steps S600 to S620 are the same processes as steps S500 to S520 in item relevance calculation. The conditions for reading the usage history in step S600 may be the same as the conditions for reading the usage history in step S500, but it is better to read a usage history for a longer period than the usage history read in step S500. User preferences for items change in a relatively short period of time, while user preferences for categories tend not to change much in a short period of time. To calculate the item relevance and category relevance accurately, a usage history of a certain level or more is required. Item relevance may be too high. On the other hand, the category relevance is less likely to have such problems, so it is better to use a relatively long-term usage history. Note that by referring to the item-category correspondence table 101C, the item identifier included in the usage history read in step S600 can be associated with the category identifier. Let Cs be the number of types of category identifiers corresponding to the usage history read in step S600.

  In step S630, the relevance calculation unit 104 uses the usage history read in step S600, and the usage subject x and another usage subject x2 (x2εσ, x ≠ x2) belonging to the usage subject set σ. The user similarity Su [x] [x2] is calculated. As a first method for calculating the similarity between users, the same method as in step S530 can be used.

As a second method of calculating the similarity between users, the number of uses regarding the category of the user can be used. Specifically, first, the usage history read in step S600 and the item-category correspondence table 101C are combined using the item identifier as a key, the category used by the user x is identified, and the usage count for each category is determined. Is calculated. The usage count for each category is also calculated for the usage subject x2. Specifically, G [x] [p] indicates the number of uses of the item corresponding to the category p by the use subject x, and G [x2] [p] indicates the use count of the item corresponding to the category p by the use subject x2. p], and the number of types of category identifiers corresponding to the usage history read in step S600 is Cs, and the inter-user similarity Su [x] [x2] is calculated using the cosine measure shown in equation (5). Or you may calculate using the Pearson product-moment correlation coefficient shown to (6) Formula. However, Gc [x] [x2] is a set of categories used by the users x and x2, and Ga [x] is the number of times the user x used a category belonging to Gc [x] [x2]. Average value of. Ga [x2] is an average value of the number of times the user x2 has used the category belonging to Gc [x] [x2].

In step S632, the relevance calculation unit 104 creates a similar user set Λ [x] corresponding to the user x using the same method as in step S532.
In step S634, the relevance calculation unit 104 calculates the category relevance H [x] [p] between the user x and the category p using the similar user set Λ [x] created in step S632. The first method for calculating the category relevance is a method using the number of times the item corresponding to the category p is used. Specifically, from the usage history data read out in step S600, data corresponding to the set of similar users is extracted, and the item-category correspondence table 101C is referenced for the extracted data. However, the number of items used for each category (the number of registered usage histories) is counted. Then, the usage count of the item corresponding to the category p used by the similar user set is set as the category relevance H [x] [p]. For the category p that is not used by the similar user set, the category relevance may be set to “0”.

  The second method of calculating the category relevance is a method using the number of users (number of users) using items corresponding to the category p. Specifically, from the usage history data read out in step S600, data corresponding to the set of similar users is extracted, and the item-category correspondence table 101C is referenced for the extracted data. However, the number of types of user entity identifiers (a unique number of user entities) is counted for each category. Then, the category relevance H [x] [p] is defined as the number of types of use subject identifiers using any item corresponding to the category p. For the category p that is not used by the similar user set, the category relevance may be set to “0”. To explain a specific example, in a similar user set, a certain user U1 uses item I1 three times, another user U2 uses item I2 twice, and item I1 and item I2 are both in category p. Suppose that it corresponds. In this case, the category relevance by the first method is “3 + 2 = 5”, and the category relevance by the second method is “2” because the two users U1 and U2 are using it.

A third method for calculating the category relevance is a method using similarity between users. Specifically, the category relevance H [x] [p] is calculated based on the equation (7).

  Here, Su [x] [r] is the similarity between users of the user subject x and the user subject r included in the similar user set Λ [x], and G [r] [p] is determined in step S600. In the read usage history data, the usage number r is the number of times that the usage subject r has used the category p. For the category p that is not used by the similar user set, the item relevance may be set to “0”. When the third method for creating a similar user set is used, it is preferable to use the third method for calculating the category relevance.

  In step S <b> 640, the relevance calculation unit 104 creates a related category set に 関 す る [x] related to the user x and stores it in the category relevance storage unit 103. The column of the use subject identifier in the category relevance table 103A shown in FIG. 9 is used as the use subject identifier of the use subject x, and each category identifier included in the related category set Π [x] corresponds to the category identifier in the table 103A. Good. In order to create the related category set Π [x], the relevance calculation unit 104 may perform the same processing by replacing the item in step S540 with a category. Specifically, the relevance calculation unit 104 may create the related category set Π [x] using the first or second method described in step S540.

  In step S650, the relevance calculation unit 104 determines whether an unprocessed usage subject can be selected. The degree-of-association calculation unit 104 determines “Yes” when there is an unprocessed user in the recommended target set K1 created in step S610, and “No” when there is no unprocessed user. Is determined. When determining “Yes”, the relevance calculation unit 104 returns to step 620 to repeat the process, and when determining “No”, the relevance calculation unit ends the category relevance calculation process. The above is the description of the first method of the category relevance calculation process.

  The second method of the category relevance calculation process uses only the usage history of the user x when calculating the category relevance. In the second method of category relevance calculation processing, steps S630 and S632 are omitted after executing the processing of S600 to S620 described above, and processing of step S636 (not shown) is executed instead of step S634.

  In step S636, the relevance calculation unit 104 extracts the usage history corresponding to the usage subject x from the usage history read out in step S600, and uses it to extract the category between the usage subject x and the category p. Relevance H [x] [p] is calculated. Specifically, the usage history corresponding to the usage subject x and the item-category correspondence table 101C are combined using the item identifier as a key, the item usage count J [p] for each category is counted, and J [p] May be set as the category relevance H [x] [p] (H [x] [p] = J [p]). The number of usage histories corresponding to the usage subject x may be J2, and a value obtained by dividing J [p] by J2 may be the category relevance (H [x] [p] = J [p] / J2). That is, the relative ratio of the number of uses (category use count) in which the use subject x uses an item corresponding to the category p may be used as the category relevance. For categories that are not used by the user x, the category relevance may be set to a predetermined small value (eg, “0”). After executing the process of step S636, steps S640 to S650 are executed as in the first method. The above is the description of the second method of the category relevance calculation process.

  In step S410 (after the execution of step S620 and before the execution of step S630, etc.), the processing similar to step S730 described later is further performed to determine the usefulness of the item for the user x and recommended. A candidate item set may be created. Then, when calculating the category relevance in step S634, the number of use of the item, the number of users using the item, and the like may be counted only for items belonging to the recommended candidate item set. Furthermore, when the recommended candidate item set is created without excluding the items used by the user (when items that have been canceled are excluded), the items are limited to items belonging to the recommended candidate item set in step S630. The similarity between users may be calculated using the information. Thus, by considering the recommendation candidate item set in the category relevance calculation process, it is possible to reduce the calculation amount and the storage capacity necessary for creating the recommendation information.

  Note that in the item relevance calculation step described above, normalization processing may be performed so that the maximum value or total value of the item relevance becomes a predetermined value. For example, the normalization process may be performed so that the maximum value of the item relevance level is “1” and the minimum value is “0”. The same applies to the category relevance.

<Recommendation information creation (step S420)>
Next, details of the selection process of recommendation information included in the recommendation information creation (step S420) will be described below. FIG. 20 is a flowchart showing a first method of recommendation information selection processing.

  In step S710, the information selection unit 107 selects one unprocessed user from the recommended target set K1 created in step S510. The item to be processed is defined as a use subject x.

  In step S720, the information selection unit 107 initializes the buffer B for calculating the category priority. The category priority is an index for determining a recommended category corresponding to the user x. Specifically, the information selection unit 107 refers to the category relevance table 103A and specifies the total number Nc of category types stored therein. Then, Nc buffers B [k] corresponding to the category k (category identifier k) are prepared, and their values are initialized to “0”.

  In step S730, the information selection unit 107 creates a recommended candidate item set for the use subject x. In this embodiment, a non-recommended item set η [x] for the user subject x, which is a complement of the recommended candidate item set, is created. Specifically, first, an item used by the user x for a predetermined number of times η1 or more is extracted while referring to the item usage history table 102A. Note that items that have been used a predetermined number of times η1 or more in a predetermined period (for example, the last one month) may be extracted. Then, the item identifier of the extracted item is put into the non-recommended item set η [x]. This predetermined number of times η1 may be determined according to the type of item and the characteristics of the service.

  Specifically, η1 is set to a large value for an item (for example, food, consumables, printer paper, etc.) having a property that the user frequently uses (purchases) the same item repeatedly. In addition, for an item having such a property, a predetermined period for counting the number of times of use may be set relatively short (for example, when a predetermined period of a normal item is one month, it is set to one week) etc). On the other hand, η1 is set to a small value for items (for example, automobiles, computers, home appliances, etc.) that do not frequently use (purchase) the same items. Moreover, you may set the predetermined period which counts the frequency | count of utilization relatively long with respect to the item of such a property (For example, when the predetermined period of a normal item is one month, it is set to one year. etc). Further, in principle, an item having a property that the user uses (for example, purchases and browses) the same item only once (for example, digital content that can be repeatedly reproduced by the terminal device 30 after purchase) is set to η1 = 1. Depending on the type of item and the characteristics of the service, by setting η1 to a very large value, items that are frequently used may not be included in the non-recommended item set η [x] (is included in the recommended candidate item set). It may be.

  Furthermore, an item is added to the non-recommended item set η [x] using the information of the item provision period. First, referring to the item information table 101A, the date and time obtained by adding a predetermined time η2 (for example, 1 hour, 24 hours, etc.) to the current date and time are compared with the item provision end time (date and time) to provide the item If the end time is earlier, the item is put into the non-recommended item set. Such an item is determined to be unsuitable as a recommended item because the provision period has already ended or is about to end soon. Next, the date and time obtained by adding a predetermined time η3 (for example, one week and one month) to the current date and time are compared with the item provision start time (date and time). If the item provision start time is later , Put the item into the non-recommended item set. Usually, η2 <η3. Such an item is determined to be unsuitable as a recommended item at the present time since the start of provision is quite long in the future.

  In addition, in a service in which a favorite item, a wish list (Wish list), or the like can be registered by the user's own operation, it can be determined that the user is sufficiently aware of the existence of the item. Even if such an item is recommended to the user anew, there is a low possibility that the recommendation information will lead to the item usage behavior. Therefore, such an item may be included in the non-recommended item set. In addition, in a recording reservation or viewing reservation service for a TV program or a net program (a program that can be auditioned at a predetermined date and time), a program (item) for which the user has made a recording / viewing reservation is also the presence of the item. Can be judged to be fully recognized. Therefore, the program (item) for which recording / viewing reservation is being made may be included in the non-recommended item set.

  Furthermore, as another method of creating the non-recommended item set η [x], the number of times the item is used, the consistency with the item provision period, the possibility that the user has already recognized the item, etc. are digitized, and those numbers The total non-usefulness (or usefulness) is calculated using, and a non-recommended item set can be created using the non-usefulness. For example, first, a value Da [x] [y] (Da [x] [y] / η1) obtained by counting the number of times E [x] [y] used by the user x by the user x and dividing it by η1 described above. Is calculated. Next, a value T1 is calculated by subtracting the item provision end time (date and time) Te of the item y from the date and time obtained by adding the predetermined time η2 to the current date and time T0. T1 = (T0 + η2) −Te. Then, an output value Db [y] = Z (T1) when T1 is input to the function Z (t) is calculated.

  Here, the function Z (t) is a monotonically increasing function that outputs a value of 0 or more, and has a characteristic that the output is small when the input t is negative and the output is large when the input t is positive. . For example, the function Z (t) can be obtained using a sigmoid function. Next, a value T2 is calculated by subtracting the date and time obtained by adding a predetermined time η3 to the current date and time T0 from the item provision start time (date and time) Ts of the item y. T2 = Ts− (T0 + η3). Then, an output value Dc [y] = Z (T2) when T2 is input to the above-described function Z (t) is calculated. Next, a numerical value Dd [x] [y] indicating the possibility that the user x recognizes the item y is calculated. For example, if user x registers item y in a favorite item or wish list, Dd [x] [y] = 1, otherwise Dd [x] [y] = 0. Good.

  Then, using the weighted addition value of Da [x] [y], Db [y], Dc [y], and Dd [x] [y], the non-use degree Dz [x] [y] is calculated. If each weighting factor is Wa, Wb, Wc, Wd, Dz [x] [y] = Wa × Da [x] [y] + Wb × Db [y] + Wc × Dc [y] + Wd × Dd [x] [Y]. This Dz [x] [y] has a larger value as the usefulness of the item y to the user x is lower. Then, when the non-use degree Dz [x] [y] is equal to or greater than a predetermined threshold value θz, the item y may be included in the non-recommended item set η [x]. Alternatively, the non-usefulness Dz [x] [y] is obtained by using a product of Da [x] [y], Db [y], Dc [y], and Dd [x] [y] instead of the weighted addition value. May be calculated. That is, Dz [x] [y] = Da [x] [y] × Db [y] × Dc [y] × Dd [x] [y] may be used.

  Further, a power value with Da [x] [y] as a base and a predetermined value γa as an exponent, a power value with Db [y] as a base and a predetermined value γb as an exponent, and a predetermined value γc as a base with Dc [y] as a base. It is also possible to calculate a power value that is an exponent, a power value that uses Dd [x] as a radix and a predetermined value γd as an index, and a product of each power value may be Dz [x] [y]. Of course, it is possible to calculate the usefulness of the item y with respect to the user x by the same method as described above, and to add items whose usefulness is greater than or equal to a predetermined value to the recommended candidate item set. In this embodiment, since the information selection unit 107 creates a non-recommended item set, it can be said that the information selection unit 107 has the function of the recommended usability determination unit 1073 of the first embodiment. Further, the information selection apparatus 1 may be configured such that an independent recommended usability determination unit is provided and the recommended usability determination unit creates a recommended candidate item set or a non-recommended item set.

  In step S740, the information selection unit 107 refers to the item relevance table 105A, and from among items that are related items corresponding to the user x and are not registered in the non-recommended item set η [x]. An unprocessed related item y is selected, and its item relevance W [x] [y] is acquired.

  In step S750, the information selection unit 107 acquires a candidate category (candidate category identifier) that is a category corresponding to the related item y while referring to the item-category correspondence table 101C. When there are a plurality of candidate category identifiers, the information selection unit 107 may acquire all of them. A set of the obtained candidate category identifiers is set as a candidate category set β, and the number of elements of the set β is set as Nβ.

  In step S760, the information selection unit 107 initializes a control variable j for designating elements of the candidate category set β to “1”. Hereinafter, the category that is the j-th element of β is denoted as β [j]. A category identifier may be used as β [j], or a number that can identify the category (for example, a number such as “1”, “2”, “3”, etc.) may be used.

  In step S770, the information selection unit 107 updates the category priority B [β [j]] corresponding to the category β [j]. Specifically, any one of the following first to fourth methods may be used.

The first method for updating category priority is a method using equation (8). γ1 and γ2 are each a positive constant. Although not shown in the equation (8), a value obtained by multiplying H [x] [β [j]] by a constant, a value obtained by adding a constant, or the like is used instead of H [x] [β [j]]. May be. A value obtained by multiplying W [x] [y] by a constant, a value obtained by adding a constant, or the like may be used instead of W [x] [y]. This first method is a power in which a value using the category relevance H [x] [β [j]] between the user x and the candidate category β [j] is a radix and the first predetermined value γ1 is an exponent. It can be said that the category priority is calculated using a multiplication value of the value and a power value having a value using the item relevance W [x] [y] as a base and a second predetermined value γ2 as an index. . By changing the values of γ1 and γ2, the balance between the item relevance and the category relevance in the category priority calculation can be changed.

The second method for updating category priority is a method using equation (9). Here, ω1 and ω2 are positive constants. As is clear from the equation (9), ω1 is a weighting coefficient for a value based on the category relevance H [x] [β [j]] between the user x and the candidate category β [j]. Further, ω2 is a weighting factor for the item relevance W [x] [y]. Still further, weighted addition may be performed using a value obtained by adding or subtracting a predetermined value to H [x] [β [j]] and W [x] [y]. That is, it can be said that the second method calculates the category priority using a weighted addition value of a value using the category relevance and a value using the item relevance. By changing the values of ω1 and ω2, the balance between the item relevance and the category relevance in the category priority calculation can be changed.

A third method for updating the category priority is a method using the equation (10). The difference from the first method is that the category relevance is converted into the extended relevance using the function F (X). The output when the category relevance is input to the function F (X) is the extended relevance. An example of the characteristics of the function F (X) is shown in FIGS. 22 (a) and 22 (b). In the example shown in FIG. 22A, the function F (X) has an S-shaped non-linear characteristic, and therefore the magnitude of the expanded relevance is almost the same in a region where the category relevance is high. A function with such characteristics is suitable when there is a category with a slightly high degree of category relevance and a category with a very high degree of category relevance, and when the influence of the category relevance on the category priority is desired to be the same in both cases. . In the example shown in FIG. 22B, the output Y increases with respect to the input when the input X is from 0 to X1, but when the input X is X1, the output Y becomes the maximum value Y1, and the input X is greater than X1. The output Y decreases in a large section, and the output value becomes Y2 when the input X is the maximum value X2. That is, X1 <X2 and Y1> Y2. When a function of characteristics as shown in FIG. 22B is used, the degree of expansion relevance can be lowered when the degree of category relevance is very large.

  Although it may be effective to recommend a category with a very high degree of association with the user x to the user x, the category with a very high degree of association is a category that the user x is familiar with, so the user x is unexpected. The probability of feeling sex is low. Depending on the nature of the item or service, the user may not be surprised by the recommendation information, and there may be a problem that the item usage is not promoted. In contrast, by calculating the category priority using the extended relevance based on the conversion function F (x) as shown in FIG. 22B, it becomes difficult to recommend a category that is obvious to the user x. Such a problem can be solved. For example, if you recommend a category that has a medium category relevance and a large number of items with a high degree of item relevance, you may feel that there are many interesting items among the categories that you are not conscious about. Since it is easy to lead to a new discovery of the user, the item use is easily promoted.

  Note that Y = 0 may be set when X = 0, but as shown in FIGS. 22A and 22B, the output value Y0 when X = 0 is a relatively small positive value. It may be. By doing so, it becomes easy to put a category having a very low degree of association with the user x in the recommendation information to some extent. Of course, the characteristics of the function F (X) shown in FIGS. 22A and 22B are merely examples, and functions having various characteristics can be used according to the properties of the items and services. Further, if a monotonically increasing linear function passing through the origin is used as the conversion function F (x), the same result as in the first method can be obtained.

A fourth method for updating the category priority is a method using the equation (11). The difference from the second method is that the category relevance is converted into the extended relevance using the function F (X) described in the third method. Even when the fourth method is used, the same effect as that described in the third method can be obtained.

  If the range of the category relevance value and the range of the item relevance value are significantly different, adjust the range of the category relevance level or the item relevance level or both using a logarithmic function. After that, the equations (8) to (11) may be applied.

  In step S780, the information selection unit 107 determines whether or not the value of the control variable j is smaller than Nβ. If it is smaller, the process proceeds to step S790, and if not, the process proceeds to step S800. In step S790, the information selection unit 107 increases the value of the control variable j by “1” and updates it. Then, the process returns to step S770.

  In step S800, the information selection unit 107 refers to the item relevance table 105A, and among the items that are related items corresponding to the user x and are not registered in the non-recommended item set η [x], It is determined whether other unprocessed related items can be selected. If it can be selected, the process returns to step S740; otherwise, the process proceeds to step S810.

  In step S810, the information selection unit 107 determines a recommended category for the user based on the value of the category priority (category priority buffer B). Specifically, the information selection unit 107 may select a number of categories that do not exceed a predetermined number in the order of higher category priority and set it as a recommended category. For example, when the number of categories for which the category priority is calculated is equal to or less than a predetermined number, the information selection unit 107 may set all the categories for which the category priority is calculated as recommended categories. On the other hand, if the number is larger than the predetermined number, the information selection unit 107 may select the predetermined number in descending order of the category priority and set these as recommended categories. Further, the information selection unit 107 may select a category having a category priority level equal to or higher than a predetermined value and set these as recommended categories. Further, the information selection unit 107 may select a number of categories that do not exceed a predetermined number in descending order of category priority from categories having a category priority of a predetermined value or higher.

  Furthermore, the information selection unit 107 assigns category rankings to the recommended categories in descending order of category priority. Then, the information selection unit 107 stores the use subject identifier, the category order, and the recommended category identifier in association with each other in the recommended information table 108A. In one process of step S810, the recommended category identifier corresponding to one user and the category order are stored in the recommended information table 108A.

  In step S820, the information selection unit 107 determines a recommended item for each recommended category determined in step S810. Specifically, the information selection unit 107 may combine the two tables by associating the related item identifier of the item relevance degree table 105A with the item identifier of the item-category correspondence table 101C. Then, the information selection unit 107 selects a row in which the “use subject identifier” column (item) matches the use subject x in the combined table, and sets the item relevance level for each category identifier of the recommended category. The items not registered in the non-recommended item set η [x] in the descending order may be selected so as not to exceed a predetermined number and set as recommended items.

  For example, when the number of items corresponding to the recommended category and not registered in the non-recommended item set η [x] is equal to or less than a predetermined number, all those items may be used as recommended items. On the other hand, if there are more items than the predetermined number, the predetermined number may be selected in descending order of the degree of item relevance and set as recommended items. Note that this predetermined number (that is, the upper limit number of recommended categories) may be set according to the category priority. For example, in a category having a high category priority, this predetermined number may be increased so that many items are selected as recommended items. On the other hand, in a category with a low category priority, this predetermined number may be reduced and a small number of items may be selected as recommended items.

  Further, the information selection unit 107 selects items that have an item relevance of a predetermined value or more and are not registered in the non-recommended item set η [x] for each category identifier of the recommended category, and selects them. It may be a recommended item. In this case, the predetermined value may be set according to the category priority. For example, in a category with a high category priority, this predetermined value may be reduced and many items may be selected as recommended items. On the other hand, in a category with low category priority, this predetermined number may be increased and a small number of items may be selected as recommended items.

  Further, the information selection unit 107 may select, for each category identifier of the recommended category, a number of items that do not exceed a predetermined number in descending order of the item relevance from items having an item relevance of a predetermined value or higher. Good.

  Then, the information selection unit 107 assigns item rankings to the combinations of the recommended category and the recommended item in descending order of the item relevance. Then, the information selection unit 107 associates the use subject identifier, the item order, and the recommended item identifier, and stores them in the recommended information table 108A. In one process of step S820, the recommended item identifiers corresponding to one user and their item ranks are stored in the recommended information table 108A.

  A specific example of the process in step S820 will be described with reference to FIG. FIG. 23 shows a state in which non-recommended items are excluded (internal table) after combining the two tables by associating the related item identifier of the item relevance table 105A with the item identifier of the item-category correspondence table 101C. Is schematically shown. For example, as can be seen from the fact that there are a total of three rows corresponding to the related item “ItemID-52”, when a plurality of categories correspond to one related item, there are as many rows in the internal table as the number of corresponding categories. Created. In the example of this figure, the item is music content, and “genre” and “mood” are used as category types. In step S810, “rock (genre)”, “jazz (genre)”, and “astringent” It is assumed that three categories of (mood) are determined as recommended categories. Further, a selection condition is used in which the user x is “UserID-10” and two items are selected in descending order of item relevance for each recommended category.

  There are three items corresponding to the recommendation category “lock”, “ItemID-50”, “ItemID-52”, and “ItemID-54”, and these are recommended candidate items. The item relevance levels are “0.90”, “0.80”, and “0.60”, respectively. For this reason, “ItemID-50” is selected as the item ranking first, and “ItemID-52” is selected as the item ranking second.

  Since there is only one item corresponding to the recommended category “Jazz” of “ItemID-58”, this is selected as the item ranking first place.

  There are three categories corresponding to the recommended category “Astringent”: “ItemID-52”, “ItemID-55”, and “ItemID-25”, and the item relevance is “0.80” and “0.72”, respectively. And “0.40”. For this reason, “ItemID-52” is selected as the first item ranking, and “ItemID-55” is selected as the second item ranking.

  In this example, “ItemID-52” is selected redundantly in the two categories of “lock” and “astringent”. In this way, the same item may be selected redundantly among a plurality of categories. Or you may select an item so that it may not overlap between several categories. For example, when selecting an item corresponding to “Astringent”, it is checked that “ItemID-52” is already selected as “Lock”, and this is excluded, and “ItemID-55” is set to item rank 1 It is also possible to select “ItemID-25” as the second item rank. When a recommendation candidate item corresponds to a plurality of recommendation categories, a recommendation category to be finally associated may be determined according to the category priority or the category relevance. For example, in the example shown in FIG. 23, when the category priority of “lock” is “0.9” and the category priority of “slow” is “0.7”, “ItemID-52” is relatively set. For example, processing corresponding to “lock” having a high category priority and not corresponding to “buoy” having a relatively low category priority may be performed. In addition, when the category priorities of the two categories are approximately the same, a category with a low category relevance may be prioritized to be a recommended category. Such processing can provide unexpected recommendation information to the user.

  In step S830, the information selection unit 107 determines whether another unprocessed user can be selected with reference to the item association degree table 105A. If it can be selected, the process returns to step S710, and if not, the recommendation information selection process ends.

  Next, the second method of the recommended information selection process in step S420 will be described with reference to the flowchart of FIG.

  In step S910, the information selection unit 107 selects one unprocessed user from the recommended target set K1 created in step S510. The user subject to be processed is defined as a user subject x.

  In step S920, the information selection unit 107 creates a non-recommended item set η [x] by the same method as in step S730.

  In step S930, the information selection unit 107 collects related items corresponding to the user x for each category, and creates a temporary table sorted in descending order of item relevance. Note that items registered in the non-recommended item set η [x] are not registered in the temporary table. If a temporary table corresponding to another user other than x already exists, the information selection unit 107 deletes it and then creates a new temporary table. Specifically, similar to the description in step S820, the related table identifier is associated with the related item identifier in the item relevance table 105A and the item identifier in the item-category correspondence table 101C, and then the two tables are not recommended. You can exclude items. Then, the information selection unit 107 selects a row in which the “use subject identifier” column (item) matches the use subject x in the combined table, and sets the table in descending order of the item relevance with the use subject for each category. Sort lines. By creating a temporary table excluding the non-recommended item set, the category priority can be calculated using the item relevance of the item to which the recommended candidate item set (complement of the non-recommended item set) belongs.

  An example of this temporary table is shown in FIG. As shown in this figure, in the temporary table, related items are sorted in the descending order of item relevance by category such as “Rock”, “Jazz”, “Blues”, “Astringent”, etc. Has been granted. Note that the item rank column may be omitted because the item rank is known from the storage position of the item identifier. Further, when providing an item ranking column, it is not always necessary to sort the rows of the table. In the example shown in this figure, “ItemID-101” is registered in a total of three categories of “Lock”, “Blues”, and “Astringent”, and “ItemID-110” is “Lock” and “Blues”. A total of two categories are registered. As can be seen from this, items corresponding to a plurality of categories are registered in a plurality of table rows for each category. In FIG. 24, “C1,” “C2,” “C3,” and “C4” items corresponding to “Rock”, “Jazz”, “Blues”, and “Astringent” are registered, respectively.

  When creating a temporary table, not all the related items registered in the item relevance table 105A are entered in the temporary table, but registered so that the number of related items in each category is the same. An upper limit may be set for the number of registered related items. For example, for one category, the maximum number of related items may be set to 20. In addition, only related items whose item relevance is equal to or greater than a predetermined value may be registered in the temporary table.

  In step S940, the information selection unit 107 calculates a category priority using the temporary table created in step S930. In the following, the category priority of category k registered in the temporary table corresponding to user x is represented by P [x] [k].

The first method for calculating category priority is a method using equation (12). Here, H [x] [k] is the category relevance between the user x and the category k. The category relevance level is obtained by referring to the category relevance level table 103A. γ3 and γ4 are positive constants, respectively. The score Q [x] [k] is a value related to the category k related to the related item registered in the temporary table, and reflects the degree of item relevance with the user x. A specific example of a method for calculating the score Q [x] [k] will be described later. Although not shown in equation (12), a value obtained by multiplying H [x] [k] by a constant, a value obtained by adding a constant, or the like may be used instead of H [x] [k]. Further, a value obtained by multiplying Q [k] by a constant, a value obtained by adding a constant, or the like may be used instead of Q [x] [k]. The first method is a power value having a value using the category relevance H [x] [k] as a radix and a third predetermined value γ3 as an index, and a value using the item relevance (Q [x] It can be said that the category priority is calculated using a value obtained by multiplying [k]) by a power and a fourth predetermined value γ4 as an exponent.

The second method for calculating category priority is a method using equation (13). Here, ω3 and ω4 are positive constants. Note that weighted addition may be performed using a value obtained by adding or subtracting a predetermined value to H [x] [k] and Q [x] [k], respectively. The second method uses a value obtained by weighted addition of a value using the category relevance H [x] [k] and a value (Q [x] [k]) using the item relevance, It can be said that the priority is calculated.

The third method for calculating the category priority is a method using the equation (14). The difference from the first method is that the extended relevance using the function F (X) is used as in the description of step S770. For example, when the degree of association between the user x (usage subject x) and the category k is very high, the category is a familiar category for the user x, but is likely to lack freshness. Therefore, an unexpected category can be recommended to the user x by deliberately lowering the category priority of a category having a very high degree of association using the function F (X).

A fourth method for calculating category priority is a method using equation (15). The difference from the second method is that the extended relevance degree using the function F (X) is used as in the description of step S770.

  Here, a method of calculating the score Q [x] [k] will be described. The first method for calculating the score Q [x] [k] is a method using the number of items whose item relevance with the user x is a predetermined value or more among the items in the temporary table corresponding to the category k. For example, in the temporary table, the number of related items whose item relevance is “0.5” or more may be counted for each category. In addition, when an item whose degree of item relevance with the user x is greater than or equal to a threshold value is entered in the related item set using the second method of related item set creation in step S540, the category registered in the temporary table The number of items “C1”, “C2”, “C3],“ C4], etc. for each category can be used as the score Q [x] [k] for each category.

The second method for calculating the score Q [x] [k] is when a predetermined number (NQ) items are selected from items in the temporary table corresponding to the category k in descending order of the degree of item relevance with the user x, or items This is a method that uses the sum of the item relevance levels when selecting items with a relevance level equal to or greater than a predetermined value, or a representative value of the item relevance levels. As the representative value, an average value, a median value, a mode value, or the like can be used. In the example shown in FIG. 24, when NQ = 3 and the average value is used, the score Q [x] [k] corresponding to the category “lock” is “(0.95 + 0.92 + 0.86) ÷ 3 = 0. 91 ".
The third method for calculating the score Q [x] [k] is a method for calculating the maximum value of the item relevance for the items in the temporary table corresponding to the category k. For example, in the example shown in FIG. 24, the score Q [x] [k] corresponding to the category “Rock” is “0.95”, and the score Q [x] [k] corresponding to the category “Jazz” is “0. 78 ". This method corresponds to the case where NQ = 1 in the second method.

  Even when any of the above-described first to third methods is used, the score Q [x] [k] is a value calculated using the item relevance with the user x of the related item corresponding to the category k. It can be said. In each method of calculating the score Q [x] [k], normalization processing such as setting the maximum value to “1” may be performed. The score Q [x] [k] may be calculated by combining the first and second methods described above. For example, when the value obtained by the first method is Q1 [x] [k] and the value obtained by the second method is Q2 [x] [k], Q1 [x] [k] and Q2 [ a product Qa [x] [k] with x] [k] or a weighted addition value Qb [x] [k] between Q1 [x] [k] and Q2 [x] [k] ( 12) The category priority may be calculated by substituting Qa [x] [k] or Qb [x] [k] into Q [x] [k] in the equations (15) to (15).

  In step S950, the information selection unit 107 determines a recommended category by the same method as in step S810, and stores the recommended category and the category ranking in the recommended information table 108A. Similar processing may be performed using the category priority P instead of the category priority buffer B in step S810.

  In step S960, the information selection unit 107 determines a recommended item for the user x. Specifically, using the temporary table created in step S930, for each recommended category determined in step S950, select a number of items that does not exceed a predetermined number in descending order of item relevance with the user x. It can be used as a recommended item. The predetermined number used here may be the same as the predetermined number NQ used in the second method for calculating the score Q in step S940. Alternatively, an item having a degree of item relevance with the user x may be selected as a recommended item. Furthermore, you may select the item of the number which does not exceed a predetermined number in order with high item relevance with the user subject x from the items whose item relevance rate with the user subject x is a predetermined value or more. Since the item is selected from the temporary table, the non-recommended item is excluded, that is, the recommended item is selected from the recommended candidate item set.

  In step S970, the information selection unit 107 determines whether another unprocessed user can be selected with reference to the item association degree table 105A. If it can be selected, the process returns to step S910. If not, the recommendation information selection process ends. The above is the description regarding the specific example of the recommendation information selection process included in the recommendation information creation process (step S420). In the present embodiment, the information selection unit 107 performs the recommended information selection process. However, the information selection unit 107 further includes a recommended utility determination unit, a category priority calculation unit, a category selection unit, and an item selection unit. The information selection apparatus 10 may be configured to include the above. For example, the recommended usability determination unit performs the processes of steps S730 and S920, the category priority calculation unit performs the processes of steps S770 and S940, and the category selection unit performs the processes of steps S810 and S950. Also good.

  As described above, the technique disclosed in Patent Document 1 determines the user preference for each program category calculated using a program included in the recording reservation history of the TV program by the user in order to determine the recommended category. use. However, the technique disclosed in Patent Document 1 does not sufficiently consider information of television programs (that is, corresponding to items in the present embodiment) belonging to each program category when determining a recommended category. For example, the recording reservation history is used to calculate the user preference level. However, since the user has already sufficiently recognized the existence of the recording reservation program, it is considered enough to exclude it from the recommended program. It is done. In addition, the program used to determine the recommended category may include a program that has already been broadcast, but such a program is a recommended program intended for a program to be broadcast in the future. Does not enter. That is, the difference (gap) between the program used for determining the recommended category and the recommended program is large. For this reason, in the technique disclosed in Patent Document 1, there is not always a recommended program associated with the recommended category (that is, equivalent to the recommended item of the present embodiment). Therefore, even if the technique disclosed in Patent Document 1 can present a recommended category based on a program included in a recording reservation history of a TV program by a user, the recommended program corresponding to the category and broadcast in the future (that is, , Corresponding to the recommended item of this embodiment) may not be sufficiently presented.

  The main purpose of information recommendation is to increase the use of items by the user by presenting the user with information regarding items that the user is interested in. By presenting information on recommended items divided into recommended categories, it is expected that the ease of understanding the recommended information is improved and the degree of interest of the user is increased. However, when recommendation categories that do not include an item having high relevance with the user are mixed in the recommendation information, the degree of satisfaction and interest of the user with respect to the recommendation information is lowered, which is rather counterproductive. For example, when a category is recommended but there are few recommended items corresponding to the category, there is a possibility that the user may have an impression that the assortment of the item providing service is not sufficient.

  In addition, in order to increase the number of recommended items corresponding to the recommended category, if many items having low item relevance are included and recommended, the user's trust in the information recommendation function may be impaired. In particular, after selecting a recommended category, when selecting a recommended item corresponding to it, exclude items that can be assumed that the user is fully aware of the existence of the item, or exclude items whose provision period has ended If the process is performed, there is a possibility that a recommended item corresponding to the recommended category cannot be sufficiently secured. On the other hand, unless the process of excluding items that are not useful to the user from the recommended items is not performed, the user's trust for the information recommendation function cannot be obtained. In addition, a display space or a display page increases in order to display a recommendation category that is not useful to the user, and an operation for browsing recommended information increases. Especially for devices such as mobile phones and smartphones that have limited operability, such as a small display screen, limited information that can be displayed at one time, and small operation parts such as buttons, etc., it is recommended to be compact with as few display pages and display areas as possible. Since it is necessary to present information, such a problem is likely to become apparent.

  On the other hand, in the first method of selecting recommended information according to the present embodiment, a non-recommended item set that is a set of items that can be determined to be unusable for the user is created in step S730, and non-recommended item set is determined in step S770. The category priority is calculated (updated) using the item relevance of items (recommended candidate items) other than the recommended item set, and the recommended category is determined based on the category priority in step S810. In the second method of selecting recommended information, the category priority is calculated using the item relevance in step S940, and the recommended category is determined based on the category priority in step S950.

  That is, the category priority used in this embodiment reflects the magnitude of the degree of item relevance between each item belonging to each candidate category and the user, similar to that in the first embodiment. In particular, because the recommendation category is determined using the item relevance of items that are likely to become recommended items (recommended candidate items), the difference between the item used to determine the recommended category and the recommended item (gap) ) Is small. For this reason, this embodiment can reduce possibility that the item which has high relationship with a user main body does not exist in a recommendation category. As a result, it is possible to prevent an unnecessary increase in the space and page necessary for displaying the recommendation information and the operation necessary for browsing the recommendation information. For this reason, this embodiment can create recommended information that is easy to use and understandable for the user, and can present it to the user.

  Furthermore, in this embodiment, the example which calculates a category priority using both a category relevance degree and an item relevance degree like (8) Formula-(15) was shown. As a result, a recommendation category that has a strong relationship, similarity, preference, or correlation with the user and that is highly likely to be associated with an item that has a high relationship with the user is determined. be able to.

<Fourth Embodiment>
In the present embodiment, a modified example of the recommendation list displayed in step S180 of the third embodiment will be described. FIG. 25 shows a modification of the recommendation list display according to the present embodiment. A difference from the example shown in FIG. 16 is that an “adjustment of extended relevance” button and a “re-obtain recommendation information” button are displayed at the upper right of the screen. The user presses the “adjustment of extended relevance” button, adjusts the characteristics of the conversion function F (x) for calculating the expansion relevance, and then presses the “re-require recommendation information” button to obtain new function characteristics. It is possible to obtain recommendation information created by using. Note that the “adjustment of extended relevance” button may be displayed on another screen such as the new item introduction screen shown in FIG. When the user does not press the “adjustment of expansion relevance” button, the expansion relevance is calculated using the standard function characteristic or the function characteristic set immediately before.

  The terminal device 30 may display an adjustment screen for the expansion relevance as shown in FIG. 26A or 26B in response to the user pressing the “adjustment of expansion relevance” button. The adjustment screen is generated by the terminal device 30 according to the display control information transmitted from the information selection device 10 or the item providing server 20. For example, the display control information creation unit 106 of the information selection device 10 generates display control information for outputting a screen (for example, FIG. 26A or 26B) that accepts adjustment of the characteristics of the conversion function F (x). What is necessary is just to produce | generate and transmit this to the terminal device 30.

  FIG. 26A is an example of a setting screen for the user to select one of predetermined options and set the characteristics of the function F (x). An example of the function F (x) corresponding to each option in the example of FIG. Options 1) to 5) in FIG. 26A correspond to the characteristics 1) to 5) in FIG. 27, respectively. Further, as shown in FIG. 26B, the user may directly specify the characteristics of the function F (x). In the example shown in this figure, the user can directly input a function curve or correct a function curve prepared in advance using the input device 330 (for example, a mouse, a trackball, or a touch panel).

  Note that the examples illustrated in FIGS. 26A and 26B and FIG. 27 are merely examples, and the characteristics of the function F (X) may be adjusted by the user using other methods. For example, as in the example shown in FIG. 26A, options relating to a category having a low degree of association or options relating to a category having an intermediate degree of association may be prepared and selected by the user. In addition, based on the characteristics of the conversion function used in the recommendation list presented to the user, relative changes to the characteristics (for example, “more highly relevant categories”, “less relevant categories are given higher priority”). Etc.) may be designated by the user to set a new conversion function characteristic. Information regarding the characteristics of the function F (X) set by the user is transmitted from the terminal device 30 to the information selection device 10 and stored in a storage area inside the information selection unit 107.

  When the user presses the “Re-obtain recommendation information” button, a recommendation request is transmitted from the terminal device 30 to the information selection device 10 as in step S160. In this embodiment, information regarding the characteristics of the function F (X) may be included in the recommendation request. The information selection device 10 creates recommendation information at the timing of receiving the recommendation request (corresponding to the above-described second timing of creation of recommendation information), and provides the terminal device 30 with recommendation information corresponding to the characteristics of the designated function. To do. Note that the information selection unit 107 may store characteristic information of a function used for calculating the extended relevance for each user. When creating recommendation information, the characteristics of the function most recently specified by the processing subject user are used. In this way, since the user does not need to frequently specify the characteristics of his / her favorite function, the convenience for the user is improved. Also, when the recommendation information is created at the first timing described above, the recommendation information can be created based on the characteristics of the extended relevance degree that suits each user's preference.

  In the present embodiment, after the user adjusts the characteristics of the function F (X), the recommendation information can be created using the function F (X), so that the user can easily obtain the recommendation information that suits his / her preference. it can. Also, by changing the function characteristics as appropriate, it is possible to obtain a variety of recommended information.

<Fifth Embodiment>
In the present embodiment, another modification of the recommendation list displayed in step S180 of the third embodiment will be described. 28A to 28C show modified examples of the recommendation list display according to the present embodiment. The display screen shown in FIGS. 28A to 28C includes a plurality of graphic objects corresponding to a plurality of recommended categories. Furthermore, the arrangement of each graphic object in the display screen is determined according to the category priority level of the recommended category represented by each graphic object.

  More specifically, in FIGS. 28A to 28C, the recommended category is displayed as a circular object. And the coordinate position of the vertical direction of the screen of each circular object is determined based on the category priority of the recommendation category which each circular object represents. For example, in FIG. 28A, the upper position in the screen corresponds to the higher category priority. That is, since “category A”, “category C”, and “category B” are arranged in this order from the top of the screen, “category A” has the highest category priority, and “category B” has the highest category priority. It will be low. In this way, the user can intuitively grasp the category priority of the recommended category by using the graphic object associated with the recommended category and determining the display position according to the category priority.

  The numbers shown in the circular objects in FIGS. 28A to 28C are the number of recommended items corresponding to the recommended category. In FIG. 28B, the size of the circular object is set according to the number of recommended items. Since the number of recommended items in “Category C” is “20”, the diameter of the circle is the largest. By displaying in this way, the user can intuitively grasp the number of recommended items corresponding to the recommended category.

  FIG. 28C shows an example in which the display color and the display pattern (pattern) are changed for each category. For example, when the category type is “mood”, a low-saturation color such as pastel color may be used for a category such as “healing”, and a primary color system is used for a category such as “violent”. Use highly saturated colors. Colors, patterns, icons, and the like that match each category type such as “genre”, “mood”, and “situation” may be set using knowledge such as color psychology.

  The display methods shown in FIGS. 28A to 28C are merely examples, and other display methods may be used. For example, an object shape other than a circle may be used, the object shape may be changed for each category, and a combination of the object shape, size, color, etc. may be changed for each category. In the two-dimensional graphs shown in FIGS. 28A to 28C, the horizontal coordinate value may be determined based on some index. For example, a category that is frequently used by the user is displayed on the left side of the screen, a category that has a new item creation time or a new item start time is displayed on the left side, The horizontal coordinate value may be determined based on the representative value or total value of the item price. The user may be able to specify display characteristics such as the display position, shape, size, and color of the graphic object associated with the recommended category according to his / her preference.

  Specifically, the control unit 301 of the terminal device 30 causes the user to input desired display characteristics using the input device 330, and transmits / receives data in which categories (category IDs, category names) and display characteristics are associated with each other. The information is transmitted to the information selection apparatus 10 via 302. The control unit 110 of the information selection device 10 receives the data via the transmission / reception unit 109 and stores the data in the internal memory of the display control information creation unit 106 in association with the use subject identifier. The display control information creation unit 106 may create display control information for the use subject identifier stored in the internal memory according to the display characteristics corresponding thereto.

  When the user selects a desired category from the categories shown in FIGS. 28A to 28C using the input device 330, recommended items corresponding to the selected category are displayed as shown in FIG. It may be. In this figure, only the name (title) of the recommended item is displayed, but of course, appropriate item attribute information may be displayed together. The user can use the item by selecting the desired item.

  The information selection device 10 may operate as follows in order to display the display screen shown in FIGS. 28 and 29 on the display device 320 of the terminal device 30. The display control information creation unit 106 of the information selection device 10 creates display control information that enables the display screen shown in FIGS. 28 and 29 to be displayed on the display device 320 in step S170 of FIG. It transmits to the terminal device 30. The display control information may include a program written in a script language or the like that can be executed by the browser unit 303 of the terminal device 30 in order to efficiently realize a user's interactive operation. The display control information creation unit 106 associates a graphic object with each of the plurality of recommended categories, and based on information on each recommended category and information on recommended items corresponding to each recommended category, the position, size, and shape of the graphic object Display control information is created by determining display characteristics such as colors.

  Note that the display control information creation unit 106 does not necessarily have to create display control information. For example, in step S170 of FIG. 14, the display control information creation unit 106 obtains information (such as category priority) related to each recommended category, which is the original information of the display control information, and information on recommended items corresponding to each recommended category. You may transmit to the terminal device 30 instead of display control information. The terminal device 30 may determine display characteristics such as the display position, shape, size, and color of the graphic object based on these pieces of information. Further, if necessary, after transmitting data such as item time information and item price information from the information selection device 10 to the terminal device 30, the terminal device 30 may determine display characteristics using the information. That is, it is arbitrary how the information selection device 10 and the terminal device 30 share the processing related to the display of the recommendation list.

  In the above description, an example is shown in which the arrangement of graphic objects corresponding to the recommended category is determined according to the category priority level. However, other display forms (for example, color, size, or shape) of the graphic object corresponding to the recommended category may be determined according to the category priority. Further, it may be possible to display the recommended category as text. In this case, the color, font size, or font type of the text indicating the recommended category may be determined according to the category priority. That is, the display form (for example, arrangement, color, size, shape, font, or font type) of the display object (for example, graphic or text, or a combination thereof) corresponding to the recommended category is determined according to the category priority. May be. Also in these modified examples, the user can intuitively grasp the category priority of the recommended category.

<Other embodiments>
In the third to fifth embodiments described above, the information selection device 10 performs the recommended information creation processing in step S420 (FIG. 17), but this processing may be performed on the terminal device 30 side. In this case, the application unit 304 of the terminal device 30 may perform an operation corresponding to the information selection unit 107. For example, at an appropriate timing prior to step S160 (FIG. 14), the application unit 304 can send the item information table 101A, the category information table 101B, the item-category via the item providing server 20 or directly from the information selection device 10. Data of the correspondence table 101C, the category relevance level table 103A, and the item relevance level table 105A is acquired. At this time, all data in each table may be acquired, or only data necessary for creating recommendation information for one request identifier in step S420 may be acquired. For the category association degree table 103A and the item association degree table 105A, only the data corresponding to the user identifier of the user who uses the terminal device 30 that performs the process or the terminal identifier of the terminal device may be acquired. In step S160, the application unit 304 may execute a process corresponding to step S420 instead of transmitting a recommendation request. In this case, the application unit 304 functions as the information selection unit 107.

  The technical idea of converting the category relevance to the extended relevance using the conversion function F (x) described in the third to fifth embodiments is an item for calculating the category priority for determining the recommended category. It can be applied even when the relevance is not used. In other words, the technical idea relating to the expanded relevance can be implemented independently of the “technical idea using the item relevance for the category priority calculation” described in the first to fifth embodiments. The technical idea related to the expanded relevance can be widely applied when the category relevance is used for determining the recommended category. In particular, as shown in FIG. 22B, by using a conversion function F (x) having a section in which the output value decreases as the input value increases, it is difficult to recommend a category that is obvious to the user. Therefore, it is possible to provide a recommendation category that is surprising to the user.

  Furthermore, the above-described embodiment is merely an example relating to application of the technical idea obtained by the present inventors. That is, the technical idea is not limited to the above-described embodiment, and various changes can be made.

DESCRIPTION OF SYMBOLS 1 Network system 10 Information selection apparatus 20 Item provision server 30 Terminal apparatus 40, 42 Network 101 Metadata storage part 102 Usage history storage part 103 Category relevance degree storage part 104 Relevance degree calculation part 105 Related set storage part 106 Display control information creation part 107 information selection unit 108 recommended information storage unit 109 transmission / reception unit 110 control unit 120 display device 130 input device 201 user management unit 202 item storage unit 203 data storage unit 204 transmission / reception unit 205 control unit 301 control unit 302 transmission / reception unit 303 browser unit 304 application Unit 320 display device 330 input device 1071 category selection unit 1072 item selection unit 1073 recommended usefulness determination unit

Claims (17)

Usefulness of recommending an item to the user is determined by using information about the period during which the item can be provided to the user or usage history in which the user has used the item. A recommendation usability determination unit that generates a recommendation candidate item set that is a set of items determined to be,
A category priority calculation unit that calculates a category priority by using an item relevance indicating an intensity of relevance between an item belonging to the recommended candidate item set and belonging to a candidate category and the user,
A category selection unit that selects a recommended category to be provided to the user based on a plurality of category priorities calculated for each of the plurality of candidate categories, from among the plurality of candidate categories. To select information. The category priority calculation unit
Among items belonging to the recommended candidate item set and belonging to the candidate category, the item group having the item relevance level equal to or higher than the first predetermined value, or the second predetermined value is not exceeded in descending order of the item relevance level Select a number of items as the top item set,
Calculate the total value or representative value or maximum value of the item relevance of a plurality of items included in the upper item set,
The information selection apparatus according to claim 1, wherein the category priority is calculated using the total value, the representative value, or the maximum value. The category priority calculation unit calculates the category priority by using the number of items that belong to the recommended candidate item set and belong to the candidate category and the item relevance is equal to or greater than a third predetermined value. The information selection device according to claim 1. Using a usage history regarding a plurality of usage subjects including the usage subject, calculating a similarity between users indicating the strength of similarity between the usage subject and other usage subjects, and based on the similarity between the users, The information selection apparatus according to any one of claims 1 to 3, further comprising an item relevance calculation unit that calculates the item relevance. The category priority calculation unit
The category priority is calculated using the item relevance and a category relevance indicating the strength of relevance between the user and the candidate category. The information selection device according to item. Using a usage history regarding a plurality of usage subjects including the usage subject, calculating a similarity between users indicating the strength of similarity between the usage subject and other usage subjects, and based on the similarity between the users, The information selection apparatus according to claim 5, further comprising a category relevance calculation unit that calculates the category relevance. The category priority calculation unit
(A) a power value in which a value using the category relevance is a radix and a fourth predetermined value is an exponent; and a power value in which a value using the item relevance is a radix and a fifth predetermined value is an index; Multiplication value of
(B) a weighted addition value between a value using the category relevance and a value using the item relevance;
The information selection apparatus according to claim 5, wherein the category priority is calculated using any one of the following. The category priority calculation unit
6. The extended relevance obtained by inputting the category relevance into a predetermined conversion rule is calculated, and the category priority is calculated using the item relevance and the extended relevance. 8. The information selection device according to any one of 7 to 7. The category priority calculation unit
The information selection apparatus according to claim 8, wherein a conversion function having a section in which an output value decreases as the input value increases is used as the conversion rule. The category priority calculation unit
(A) A power value having a value using the expanded relevance as a radix and a sixth predetermined value as an exponent; and a power value having a value using the item relevance as a radix and a seventh predetermined value as an index; Multiplication value of
(B) a weighted addition value between the value using the expanded relevance and the value using the item relevance;
The information selection apparatus according to claim 8, wherein the category priority is calculated using any one of the following. The information selection apparatus according to any one of claims 8 to 10, wherein the category priority calculation unit stores characteristic information of the conversion rule for each user. The recommended usability determination unit is a numerical value indicating the degree of usability by using at least one of information regarding a period during which an item can be provided to the user and the number of times the user has used the item. The information selection device according to any one of claims 1 to 11, wherein a useful degree is calculated, and an item having the useful degree equal to or greater than an eighth predetermined value is used as an element of the recommended candidate item set. The item selection part which selects a recommended item from the items which belong to the recommendation candidate item set and belong to the recommendation category based on the item relevance level. The information selection device according to item 1. The category priority calculating unit, the item association degree, and from the claims 1 and calculates the category priority with a heterologous relevance showing the Strength of association between the item and the candidate category 4. The information selection device according to any one of claims 6 to 13. A control information creating unit for creating display control information for outputting a display screen showing information on the recommended category;
The category selection unit selects a plurality of categories as recommended categories,
The display screen includes a plurality of display objects corresponding to the plurality of recommended categories,
The display state of each display object in the display screen is determined according to the category priority level of the recommended category represented by each display object. The information selection device described in 1. Usefulness of recommending an item to the user is determined by using information about the period during which the item can be provided to the user or usage history in which the user has used the item. Generating a recommended candidate item set that is a set of items determined to be;
Calculating a category priority using an item relevance indicating the strength of relevance between an item belonging to the recommended candidate item set and belonging to a candidate category and the user, and calculating each of a plurality of candidate categories Selecting a recommended category to be provided to the user based on the plurality of category priorities, from among the plurality of candidate categories. Usefulness of recommending an item to the user is determined by using information about the period during which the item can be provided to the user or usage history in which the user has used the item. A process for generating a recommended candidate item set that is a set of items determined to be,
A process of calculating a category priority using an item relevance indicating the strength of relevance between an item belonging to the recommended candidate item set and belonging to a candidate category and the user;
And causing the computer to execute a process of selecting, from the plurality of candidate categories, a recommended category to be provided to the user based on a plurality of category priorities calculated for each of the plurality of candidate categories. Information selection program.