JP6435888B2 - Item recommendation program, apparatus, and method - Google Patents

Description

  The present invention relates to an item recommendation program, an item recommendation device, and an item recommendation method.

  There is a technique for presenting an item such as a product estimated to be of interest to the user to the user. In such a technique, for example, an item to be recommended to the user is determined using a collaborative filtering or content-based filtering technique. In the collaborative filtering and content-based filtering methods, items to be recommended are determined based on similarity, and therefore, only items similar to previously recommended items tend to be recommended.

  Therefore, as a viewpoint different from the similarity, an information recommendation method for recommending an item that is unexpected to the user, that is, an item that is of interest to the user but cannot be assumed easily is proposed. In this method, an item feature space is constructed using item attributes, and item selection distribution of the user is obtained using item evaluation information of the user. Then, a recommended index having a small value is calculated for an item that is close to the obtained distribution center and is far from the distribution shape, and a non-selected item having a small index is determined as a recommended item.

  Also, the response content determined in response to the user's request is output, and the event-related potential of the user is measured starting from the timing when the response content is output, and the event-related potential is about 600 ms from the start point. Techniques for determining user disappointment have been proposed. In this technique, the correction of the response content is instructed using the determination result of determining the disappointment of the user.

JP 2011-096025 A International Publication No. 2005/001677 Pamphlet

  However, in the prior art that makes an unexpected recommendation, an item that is close to the distribution center of the item selected by the user in the past and is away from the distribution shape is selected. Contains items that the user has not selected in the past. Items that the user did not select are considered to be items that the user gave a negative evaluation such as bad or not interested in the item. Therefore, in the conventional method, there is a possibility of recommending an item which is negative for the user but is negatively evaluated.

  An object of the present invention is to recommend an unexpected item that is unlikely to be negatively evaluated.

  As one aspect, the present invention refers to a storage unit when recommending any item included in a plurality of items to a recommendation destination. In the storage unit, already recommended items included in the plurality of items and the evaluation information of the recommended destinations for the already recommended items are stored in association with each recommended destination. And the non-positive evaluation item matched with the said evaluation information which shows evaluation other than positive evaluation among the said already recommended items is extracted for every said recommendation destination. Next, the extracted non-correct evaluation in a space having a coordinate axis corresponding to each of the plurality of attributes of the item and in which the position of the item is determined corresponding to the combination of the attribute values of the item for the plurality of attributes Specify the direction that minimizes the dispersion of item positions. Further, an item to be recommended to the recommendation destination is determined based on the specified direction and the position of any one of the plurality of items in the space.

  As one aspect, there is an effect that an unexpected item that is difficult to be negatively evaluated can be recommended.

It is a block diagram which shows schematic structure of the item recommendation system which concerns on 1st and 2nd embodiment. It is a figure for demonstrating the feature vector which shows the position of the item in feature space. It is a figure which shows an example of an item feature table. It is a figure which shows an example of the item evaluation table in 1st and 2nd embodiment. It is a functional block diagram which shows schematic structure of an item recommendation apparatus. It is a figure for demonstrating the use which uses a non-positive evaluation orthogonal vector. It is a figure for demonstrating the production | generation of a non-positive evaluation orthogonal vector. It is a figure for demonstrating determination of a recommendation item. It is a figure which shows an example of the display of a recommendation item. It is a figure which shows schematic structure of the computer which functions as an item recommendation apparatus. It is a flowchart which shows an example of an item recommendation process. It is a flowchart which shows an example of the interest vector production | generation process in 1st Embodiment. It is a flowchart which shows an example of a recommendation item determination process. It is a flowchart which shows an example of the evaluation information acquisition process in 1st Embodiment. It is a figure showing an example of change of a recommendation item. It is a figure for demonstrating the production | generation of the interest vector in 2nd Embodiment. It is a flowchart which shows an example of the interest vector production | generation process in 2nd Embodiment. It is a block diagram which shows schematic structure of the item recommendation system which concerns on 3rd Embodiment. It is a functional block diagram of the control part regarding the production | generation of item selection information. It is a figure which shows an example of the item evaluation table in 3rd Embodiment. It is a flowchart which shows an example of the evaluation information acquisition process in 3rd Embodiment.

  Hereinafter, an example of an embodiment relating to the present invention will be described in detail with reference to the drawings. In the following embodiments, for example, a case where items such as recommended products are presented to the user when using a web service such as online shopping will be described as an example.

<First Embodiment>
As shown in FIG. 1, the item recommendation system 10 according to the first embodiment includes an item recommendation device 20, a storage device 30 that stores various types of information used by the item recommendation device 20, and a user terminal used by the user. 40. The item recommendation device 20 and the storage device 30, and the item recommendation device 20 and the user terminal 40 are each connected via a network such as the Internet.

  The user terminal 40 is an information processing apparatus such as a personal computer (PC), a mobile phone, a smartphone, or a tablet that includes an input device such as a keyboard, a mouse, and a touch panel, and a display device.

  The storage device 30 stores an item feature table 31, an evaluation information database (DB) 32, and an item presentation information DB 33.

  In the item feature table 31, feature vectors for each of a plurality of items handled by the item recommendation system 10 are recorded. The feature vector represents the position (coordinate) of each item in a space (hereinafter referred to as “feature space”) spanned by coordinate axes corresponding to each of a plurality of attributes of the item. The position of the item in the feature space is determined based on the feature of the item, and the feature of the item is determined based on attribute values for a plurality of attributes of the item. That is, the feature vector is represented by a vector having attribute values for a plurality of attributes of each item as elements. Specifically, the feature vector of each item is represented by a vector <Aj: vj> including each of the plurality of attributes Aj and the attribute value vk of each attribute as elements. In order to handle a plurality of items handled by the item recommendation system 10 in the same feature space, each item is set in advance so as to have the same attribute group. Note that the attributes of an item include, for example, a pattern, shape, size, material, price, manufacturer, date of manufacture, and the like when the item is “clothes”.

  With reference to FIG. 2, an example will be described in which each item having an item identifier (item ID) ID_1, ID_2, and ID_3 has attributes A1, A2, and A3. As shown in FIG. 2, the feature space spanned by the coordinate axes corresponding to each of the attributes A1, A2, and A3 includes the position of each item (circle in FIG. 2). The position of each item is represented by the above-described feature vector. For example, if an item with item ID = ID_1 has attribute value v11 for attribute A1, attribute value v12 for attribute A2, and attribute value v13 for attribute A3, the feature vector of the item with item ID = ID_1 is <A1: v11, A2: v12, A3: v13>.

  The item feature table 31 stores the feature vector as described above in association with the item ID of each item. FIG. 3 shows an example of the item feature table 31. In the example of FIG. 3, the item ID (ID_i, i = 1, 2,..., I) is associated with each row, and the attribute Aj (j = 1, 2,..., J) is associated with each column. The item feature table 31 is composed of a matrix. Here, I is the number of items included in the item feature table 31, and J is the number of attributes that each item has. In the example of FIG. 3, the attribute value vij of the attribute Aj for the item with the item ID = ID_i is stored in the cell where the row corresponding to the item ID = ID_i and the column corresponding to the attribute Aj intersect. Therefore, each row of the item feature table 31 represents a feature vector of each item with item ID = ID_i. Note that the item feature table 31 of FIG. 3 indicates that a feature vector for each of I items having J attributes is included.

  In the evaluation information DB 32, evaluation information that is determined as an item recommended to the user (hereinafter referred to as “recommended item”) and indicates the user's evaluation of the item presented to the user is stored as an item evaluation table for each user. . FIG. 4 shows an example of the item evaluation table 32A for a user whose user identifier (user ID) is “user_1”. In the example of FIG. 4, each row corresponds to one piece of evaluation information, and each piece of evaluation information includes “item ID”, “presentation time”, and “evaluation” information. “Item ID” is the item ID of the recommended item presented. “Presentation time” is information indicating the time when the recommended item identified by the item ID in the “Item ID” column is presented to the user. “Evaluation” is the user's evaluation of the recommended item identified by the item ID in the “Item ID” column. The user's evaluation is a positive evaluation (positive evaluation) when the recommended item presented by the user is selected, and a non-positive evaluation (negative evaluation) when the recommended item is not selected. For example, the positive evaluation can be “1” and the non-positive evaluation can be “0”. Note that selecting a recommended item means a case where the user confirms the details of the recommended item by clicking the presented recommended item or purchases the recommended item presented.

  In the example of FIG. 4, the items whose item IDs are ID'_1, ID'_2, ..., ID'_n, ..., ID'_N are recommended to the user in order from time t1 to tN. Represents. Further, the recommended items with the item IDs ID′_1 and ID′_N are positively evaluated, and the recommended items with the ID′_2 and ID′_n are not positively evaluated. ID'_n (n = 1, 2,..., N, N is the total number of evaluation information stored in the item evaluation table 32A) is ID_i (i = 1, 2,...) In FIG. , I).

  In the item presentation information DB 33, item information to be presented to the user is stored for each item in association with the item ID. The item information is, for example, item image data, text data explaining details of the item size, price, and the like.

  As illustrated in FIG. 5, the item recommendation device 20 includes an acquisition unit 21, an extraction unit 22, a specification unit 23, and a determination unit 24 as functional units.

  The acquisition unit 21 records user evaluation information for the presented recommended item in the item evaluation table 32 </ b> A of the corresponding user in the evaluation information DB 32. Specifically, the acquisition unit 21 records the item ID and the presentation time of the recommended item determined by the determination unit 24 described later and presented to the user in the item evaluation table 32A of the corresponding user in the evaluation information DB 32. In addition, the acquisition unit 21 sets the corresponding “evaluation” field to “0 (non-correct evaluation)” as an initial value. Moreover, the acquisition part 21 acquires the item selection information transmitted from the user terminal 40, when the recommended item shown to the user is selected by the user. The item selection information includes the user ID and the item ID of the selected item. The acquisition unit 21 specifies the item evaluation table 32A of the corresponding user based on the user ID included in the item selection information. Then, the acquisition unit 21 updates the “evaluation” column corresponding to the item ID included in the item selection information to “1 (correct evaluation)” in the identified item evaluation table 32A.

  In addition, the acquisition unit 21 acquires various types of information transmitted from the user terminal 40 such as information for instructing the start or end of item recommendation.

  When recommending an item to a user, the extraction unit 22 extracts a recommended item (hereinafter referred to as “non-correct evaluation item”) that has been evaluated incorrectly by the user from among recommended items already presented to the user. . Specifically, the extraction unit 22 refers to the item evaluation table 32A for the corresponding user and extracts all the item IDs of the evaluation information whose “evaluation” column is “0 (non-correct evaluation)”. The extraction unit 22 generates a non-correct evaluation item list in which the extracted item IDs are listed, and delivers the non-correct evaluation item list to the specifying unit 23.

  In the feature space, the specifying unit 23 is not surprising for the user when viewed from the position of the item that has already been presented. Identify the minimum direction. Here, the reason why the specifying unit 23 specifies the direction in which the dispersion of the positions of the non-positive evaluation items is minimized will be described.

  As shown in the upper part of FIG. 6, consider a case where an item group including a plurality of items having different shapes and colors is used as a recommended item candidate to be presented to the user. “Shape” and “color” are attributes of the item. In FIG. 6, the difference in the shape of the graphic represents a difference in the shape of the item, and the difference in the type of hatching applied to the graphic represents a difference in the color of the item.

  Assume that the recommended item history selected by the user is obtained as shown in the middle of FIG. 6 while the recommended items are presented to the user sequentially. From this history, it can be seen that among the recommended items already presented, the item selected by the user has a small difference in shape and a large difference in color. That is, it is considered that the user has already studied the shape, is interested in the color, and is considering another color. Therefore, it is appropriate to recommend an item with an unrecommended color that has a round shape and does not exist in the history.

  On the other hand, as shown in the lower part of FIG. 6, it is assumed that a history of recommended items not selected by the user is obtained. From this history, it can be seen that among the recommended items already presented, the items not selected by the user have a small shape difference and a large color difference. That is, it is considered that the user is not interested in square-shaped items. Therefore, it is appropriate to recommend items having shapes other than squares.

  The above contents will be described using the position of the item in the feature space. Here, in order to simplify the description, a case where there are two types of attributes, that is, a case where the feature space is a plane will be described.

  Of the recommended items already presented to the user, for the recommended items that have been positively evaluated by the user (hereinafter referred to as “correctly evaluated items”), the positions indicated by the feature vectors of the positively evaluated items are plotted in the feature space. And analyze the distribution of positive evaluation items. For example, assume that a distribution 81 of positive evaluation items is obtained as shown in FIG. In the distribution 81 of positive evaluation items, the direction in which the variance is small represents the direction in which the attribute value of the attribute that the user has already considered (the shape in the middle example in FIG. 6) changes. It represents the direction in which the attribute value of the attribute of interest (color in the middle example of FIG. 6) changes. Therefore, a case is considered in which an item located in a direction along an axis (hereinafter, referred to as “positive evaluation axis 82”) in which the distribution of the positive evaluation item distribution 81 is maximized is determined as a recommended item. The positive evaluation axis 82 can be obtained from the vector of the first principal component when the principal component analysis is applied to the distribution 81 of the positive evaluation items. When applied to the middle example of FIG. 6, items located in the direction along the positive evaluation axis 82 indicate that the difference in shape is small and the difference in color is large. By determining an item located in the direction along the positive evaluation axis 82 as a recommended item as seen from the position of the item already presented, the item is recommended that reflects the tendency of the recommended item selected by the user so far. be able to.

  However, the recommendation of the item along the direction of the positive evaluation axis 82 strongly reflects the tendency of the recommended item selected by the user so far, and thus it is not possible to make an unexpected recommendation. Therefore, as described in the example in the lower part of FIG. 6, it is considered to reflect the history of recommended items that the user has not selected, that is, the information of the non-correct evaluation items.

  The positions indicated by the feature vectors of the non-positive evaluation items are plotted in the feature space, and the distribution of the positions of the non-positive evaluation items is analyzed. For example, as shown in FIG. Is obtained. In this non-positive evaluation item distribution 83, the direction in which the variance is small represents the direction in which the attribute value of the attribute that the user is not interested in (the shape in the lower example of FIG. 6) changes.

  Therefore, an item located in the direction along the axis where the variance of the distribution 83 of the non-positive evaluation items is minimized as viewed from the position of the already presented item is determined as the recommended item. When applied to the example in the lower part of FIG. 6, an item located in a direction along the axis where the variance of the distribution 83 of the non-positive evaluation items is minimized is an item having a shape different from the shape of the item that has not been selected so far. It becomes. Thus, in order to reflect the information on the non-correct evaluation item in the recommended item, it is possible to make an unexpected recommendation as compared with the case where an item located in the direction along the positive evaluation axis 82 is recommended. Further, it can be said that the direction along the axis where the variance of the distribution 83 of the non-positive evaluation items is the minimum represents the direction that is positive evaluation as a complement of the non-positive evaluation. Therefore, it is possible to avoid determining an item that becomes a negative evaluation as a recommended item.

  For the above reason, the specifying unit 23 specifies the direction in which the dispersion of the positions of the non-positive evaluation items is minimized. In the first embodiment, the axis orthogonal to the axis (hereinafter referred to as “non-positive evaluation axis” 84) that maximizes the variance of the distribution 83 of the non-positive evaluation items is uncorrelated with the non-positive evaluation axis 84. I think that. Therefore, an axis orthogonal to the non-positive evaluation axis 84 (hereinafter referred to as “non-positive evaluation orthogonal axis 85”) is specified as the direction in which the dispersion of the positions of the non-positive evaluation items is minimized as shown in FIG.

  Specifically, the specifying unit 23 acquires, from the item feature table 31, each feature vector of the incorrect evaluation item indicated by the item ID included in the incorrect evaluation item list delivered from the extraction unit 22. Then, the specification unit 23 plots the position indicated by each feature vector of the acquired non-correct evaluation item in the feature space to obtain a distribution 83 of non-correct evaluation items. Then, the specifying unit 23 obtains the vector of the second principal component when the principal component analysis is applied to the obtained non-positive evaluation item distribution 83 as the non-positive evaluation orthogonal axis 85. Further, the specifying unit 23 weights the non-positive evaluation orthogonal axis 85 so that the norm is 1, thereby converting the unit vector into a unit vector to generate a non-positive evaluation orthogonal vector 86. The specifying unit 23 sets the generated non-positive evaluation orthogonal vector 86 as an interest vector 89 indicating a direction in which an unrecommended item is searched in order to determine a recommended item in which the user is interested, and passes the result to the determination unit 24. .

  The determining unit 24 determines a recommended item based on the interest vector 89 delivered from the specifying unit 23 and the positions of the plurality of items in the feature space. Specifically, as illustrated in FIG. 8, the determination unit 24 calculates a straight line parallel to the interest vector 89 passing through the position of the recommended item most recently presented (p1 in FIG. 8) in the feature space. Specify as 90. The deciding unit 24 determines that the position projected on the axis of interest 90 from the position of the unrecommended item (p2, p3, p4 in FIG. 8) is the position of the recommended item most recently presented. A near unrecommended item is determined as a recommended item.

  Note that the position of the recommended item presented most recently is specified from the item feature table 31 using the item ID of the latest evaluation information as the “presentation time” in the item evaluation table 32A of the corresponding user, and using this item ID as a key. It is obtained from the feature vector. Further, the position of the unrecommended item is obtained from the feature vector of the item ID that is not included in the item evaluation table 32A of the corresponding user among the item IDs included in the item feature table 31.

  Further, the determination unit 24 acquires information on the corresponding item from the item presentation information DB 33 using the determined item ID of the recommended item as a key, and presents it to the user. For example, as illustrated in FIG. 9, the determination unit 24 sets a recommended item display area 72 in addition to the main display area 71 in the display area 70 of a web service such as online shopping used by the user. And the determination part 24 can show a recommended item to a user by transmitting the page information which displays the information of the determined recommended item to this recommended item display area 72 to the user terminal 40. The main display area 71 displays a screen according to a user operation, such as a menu screen for web services such as online shopping, detailed product information, and a purchase screen.

  The determination unit 24 notifies the acquisition unit 21 of the item ID of the recommended item presented to the user, the presentation time, and the user ID of the recommended user. Thereby, as above-mentioned, item ID and presentation time are recorded on the item evaluation table 32A specified by user ID by the acquisition part 21. FIG.

  The item recommendation device 20 can be realized by, for example, a computer 50 shown in FIG. The computer 50 includes a CPU 51, a memory 52 as a temporary storage area, and a nonvolatile storage unit 53. The computer 50 also includes an input / output interface (I / F) 54 to which an input / output device 58 such as a display device and an input device is connected. The computer 50 also includes a read / write (R / W) unit 55 that controls reading and writing of data with respect to the recording medium 59, and a network I / F 56 connected to a network such as the Internet. The CPU 51, memory 52, storage unit 53, input / output I / F 54, R / W unit 55, and network I / F 56 are connected to each other via a bus 57.

  The storage unit 53 can be realized by a hard disk drive (HDD), a solid state drive (SSD), a flash memory, or the like. An item recommendation program 60 for causing the computer 50 to function as the item recommendation device 20 is stored in the storage unit 53 as a storage medium. The item recommendation program 60 includes an acquisition process 61, an extraction process 62, a specification process 63, and a determination process 64.

  CPU51 reads the item recommendation program 60 from the memory | storage part 53, expand | deploys to the memory 52, and executes the process which the item recommendation program 60 has sequentially. The CPU 51 operates as the acquisition unit 21 illustrated in FIG. 5 by executing the acquisition process 61. The CPU 51 operates as the extraction unit 22 illustrated in FIG. 5 by executing the extraction process 62. Further, the CPU 51 operates as the specifying unit 23 illustrated in FIG. 5 by executing the specifying process 63. Further, the CPU 51 operates as the determination unit 24 illustrated in FIG. 5 by executing the determination process 64. As a result, the computer 50 that has executed the item recommendation program 60 functions as the item recommendation device 20.

  The function realized by the item recommendation program 60 can be realized by, for example, a semiconductor integrated circuit, more specifically, an ASIC (Application Specific Integrated Circuit) or the like.

  Next, the operation of the item recommendation system 10 according to the first embodiment will be described. When the user terminal 40 starts using a web service such as online shopping, the item recommendation process shown in FIG. At this time, the user recommendation unit 20 is notified of the user ID logged into the web service such as online shopping from the user terminal 40.

  In step S10 of the item recommendation process shown in FIG. 11, an interest vector generation process described later is executed, whereby an interest vector 89 is generated. Next, in step S20, a recommended item determination process, which will be described later, is executed, whereby a recommended item is determined using the interest vector 89. Next, in step S30, a later-described evaluation information acquisition process is executed, so that a recommended item is presented to the user, and evaluation information from the user for the recommended item is acquired.

  Next, in step S <b> 40, the acquisition unit 21 instructs the user terminal 40 to end the use of a web service such as online shopping, or the instruction to end the display of the recommended item in the recommended item display area 72. Is received. If not received, the process returns to step S10, and if received, the item recommendation process ends.

  Next, with reference to FIG. 12, the interest vector generation process executed in step S10 of the item recommendation process shown in FIG. 11 will be described.

  In step S11 of the interest vector generation process shown in FIG. 12, the extraction unit 22 refers to the item evaluation table 32A for the user as the item recommendation destination based on the notified user ID, and the “evaluation” column is displayed. All item IDs of the evaluation information of “0 (non-correct evaluation)” are extracted. Then, the extraction unit 22 generates a non-correct evaluation item list in which the extracted item IDs are listed, and delivers the non-correct evaluation item list to the specifying unit 23.

  Next, in step S12, the specifying unit 23 determines whether the list length of the non-correct evaluation item list, that is, the number of item IDs included in the non-correct evaluation item list is two or more. If the list length is 2 or more, the process proceeds to step S13. If the list length is less than 2, the process returns to the item recommendation process shown in FIG.

  In step S <b> 13, the specifying unit 23 acquires, from the item feature table 31, the feature vector of each non-correct evaluation item indicated by the item ID included in the non-correct evaluation item list. Then, the specification unit 23 obtains the distribution 83 of the non-positive evaluation items in the feature space, and uses the second principal component vector when the principal component analysis is applied to the distribution 83 of the non-positive evaluation items as the non-positive evaluation. Obtained as the orthogonal axis 85. Further, the specifying unit 23 generates a non-positive evaluation orthogonal vector 86 by converting the non-positive evaluation orthogonal axis 85 into a unit vector.

  Next, in step S14, the specifying unit 23 sets the generated non-positive evaluation orthogonal vector 86 as the interest vector 89, and the process returns to the item recommendation process illustrated in FIG.

  Next, with reference to FIG. 13, the recommended item determination process executed in step S20 of the item recommendation process shown in FIG. 11 will be described.

  In step S21 of the recommended item determination process shown in FIG. 13, the determination unit 24 determines whether an interest vector 89 has been generated. If an interest vector is generated in steps S13 and S14 of the interest vector generation process (FIG. 12), an affirmative determination is made, and the process proceeds to step S22. If a negative determination is made in step S12 of the interest vector generation process (FIG. 12), this step is also a negative determination, and the process proceeds to step S24.

  In step S <b> 22, the determination unit 24 specifies, as the interest axis 90, a straight line parallel to the interest vector 89 that passes through the position of the recommended item presented most recently in the feature space. Then, the determination unit 24 projects each position of the unrecommended item on the interest axis 90. Next, in step S23, the determination unit 24 determines an unrecommended item whose position projected onto the axis of interest 90 is the closest to the position of the recommended item presented most recently as the recommended item, and the processing is shown in FIG. Return to the item recommendation process shown.

  On the other hand, in step S24, the determination unit 24 randomly determines a recommended item from unrecommended items, and the process returns to the item recommendation process shown in FIG.

  Next, the evaluation information acquisition process executed in step S30 of the item recommendation process shown in FIG. 11 will be described with reference to FIG.

  In step S31 of the evaluation information acquisition process illustrated in FIG. 14, the determination unit 24 acquires information on the corresponding item from the item presentation information DB 33 using the determined item ID of the recommended item as a key, and presents it to the user. In addition, the determination unit 24 notifies the acquisition unit 21 of the item ID of the recommended item presented to the user, the presentation time, and the user ID of the recommended user. Upon receiving the notification, the acquisition unit 21 records the item ID and the presentation time in the item evaluation table 32A of the corresponding user specified by the user ID, and sets the corresponding “evaluation” field as an initial value “0 (non-correct evaluation). ) ”.

  Next, in step S32, the acquisition unit 21 determines whether the user has selected a recommended item. This determination is affirmative when the item selection information transmitted from the user terminal 40 when the recommended item presented to the user is selected by the user is acquired by the acquisition unit 21, and is not acquired Is a negative determination. The case where the item selection information is not acquired by the acquisition unit 21 is a case where the recommended item is not selected by the user within a predetermined period after the recommended item is displayed in the recommended item display area 72. The predetermined period may be, for example, a predetermined period, or a period from when a recommended item is displayed in the recommended item display area 72 to when the next recommended item is displayed.

  If the determination in step S32 is affirmative, the process proceeds to step S33, and the acquisition unit 21 specifies the item evaluation table 32A of the corresponding user based on the user ID included in the item selection information. And the acquisition part 21 corrects the applicable item by updating the "evaluation" column corresponding to the item ID contained in the item selection information to "1 (correct evaluation)" in the identified item evaluation table 32A. Record as an item. Then, the process returns to the item recommendation process shown in FIG.

  On the other hand, if the determination in step S32 is negative, the process directly returns to the item recommendation process shown in FIG. Since the initial value in the “evaluation” column of the item evaluation table 32A is set to “0 (non-correct evaluation)”, the recommended item that is not selected by the user is recorded in the item evaluation table 32A as a non-correct evaluation item as it is. It will be.

  In addition, when a negative determination is made in step S40 of the item recommendation process shown in FIG. 11 and the process returns to step S10, a standby time T is provided, and the process returns to step S10 after T seconds have elapsed. Thereby, as shown in FIG. 15, the recommended item displayed in the recommended item display area 72 is switched every time T seconds elapse.

  As described above, in the item recommendation device according to the first embodiment, the direction in which the dispersion of the positions of the non-right evaluation items is minimized is specified in the feature space based on the item attributes. Then, an unrecommended item located in the specified direction is determined as a recommended item when viewed from the position of the already presented item. For this reason, compared with the case where the item along a positive evaluation axis | shaft is made into a recommendation item, recommendation with unexpectedness can be performed. In addition, the direction in which the dispersion of the positions of the non-positive evaluation items is the minimum is the direction of the positive evaluation as a complement to the non-positive evaluation, and thus an item that is difficult to be negatively evaluated can be recommended.

  In the first embodiment, the case where the unrecommended item that is the closest to the position of the item recommended most recently as the position projected on the axis of interest on the position of the unrecommended item is determined as the next recommended item will be described. However, it is not limited to this. The unrecommended item with the shortest distance between the position of the unrecommended item and the interest axis (the length of the vertical line drawn from the position of the unrecommended item to the interest axis) may be determined as the next recommended item. In this case, the target may be limited to an unrecommended item existing within a predetermined distance from the position of the item recommended most recently.

Second Embodiment
Next, a second embodiment will be described. In the item recommendation system according to the second embodiment, the same parts as those of the item recommendation system 10 according to the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted.

  As shown in FIG. 1, the item recommendation system 210 according to the second embodiment includes an item recommendation device 220, a storage device 30, and a user terminal 40.

  As shown in FIG. 5, the item recommendation device 220 includes an acquisition unit 21, an extraction unit 222, a specification unit 223, and a determination unit 24 as functional units.

  Similarly to the extraction unit 22 in the first embodiment, the extraction unit 222 refers to the item evaluation table 32A and extracts all item IDs of evaluation information in which the “evaluation” column is “0 (non-correct evaluation)”. Generate non-correct evaluation item list. Similarly, the extraction unit 222 refers to the item evaluation table 32A, extracts all item IDs of the evaluation information whose “evaluation” column is “1 (correct evaluation)”, and generates a correct evaluation item list. The extraction unit 222 passes the generated correct evaluation item list and non-correct evaluation item list to the specifying unit 223.

  The specifying unit 223 specifies a direction based on the direction in which the dispersion of the positions of the non-positive evaluation items is the minimum and the direction in which the dispersion of the positions of the positive evaluation items is the maximum, and based on the specified direction, the interest vector 89 Is generated.

  Specifically, the specifying unit 223 acquires, from the item feature table 31, each feature vector of the correct evaluation item indicated by the item ID included in the correct evaluation item list passed from the extraction unit 222. Then, as illustrated in FIG. 16, the specifying unit 223 plots the position indicated by each feature vector of the acquired correct evaluation item in the feature space to obtain the distribution 81 of the correct evaluation item. Then, the specifying unit 223 obtains the vector of the first principal component when the principal component analysis is applied to the obtained positive evaluation item distribution 81 as the positive evaluation axis 82. Similarly, the specifying unit 223 obtains the non-correct evaluation item distribution 83 based on the item IDs included in the non-correct evaluation item list received from the extraction unit 222. Then, the specifying unit 223 obtains the vector of the second principal component when the principal component analysis is applied to the obtained non-positive evaluation item distribution 83 as the non-positive evaluation orthogonal axis 85.

  Further, the specifying unit 223 generates a positive evaluation vector 87 and a non-positive evaluation orthogonal vector 86 by converting each of the positive evaluation axis 82 and the non-positive evaluation orthogonal axis 85 into unit vectors. Then, the specifying unit 223 generates an interest vector 89 by converting the sum vector 88 of the positive evaluation vector 87 and the non-positive evaluation orthogonal vector 86 into a unit vector. The identification unit 223 delivers the generated interest vector 89 to the determination unit 24.

  The item recommendation device 220 can be realized by, for example, the computer 50 shown in FIG. An item recommendation program 260 for causing the computer 50 to function as the item recommendation device 220 is stored in the storage unit 53 of the computer 50. The item recommendation program 260 includes an acquisition process 61, an extraction process 262, a specification process 263, and a determination process 64.

  CPU51 reads the item recommendation program 260 from the memory | storage part 53, expand | deploys to the memory 52, and performs the process which the item recommendation program 260 has one by one. The CPU 51 operates as the extraction unit 222 illustrated in FIG. 5 by executing the extraction process 262. Further, the CPU 51 operates as the specifying unit 223 illustrated in FIG. 5 by executing the specifying process 263. Other processes are the same as those of the item recommendation program 60 in the first embodiment. As a result, the computer 50 that has executed the item recommendation program 260 functions as the item recommendation device 220.

  The function realized by the item recommendation program 260 can be realized by, for example, a semiconductor integrated circuit, more specifically, an ASIC or the like.

  Next, the operation of the item recommendation system 210 according to the second embodiment will be described. In the second embodiment, the interest vector generation process executed in step S10 of the item recommendation process shown in FIG. 11 is only different from the item recommendation process in the first embodiment. The interest vector generation process in the embodiment will be described.

  In step S101 of the interest vector generation process shown in FIG. 17, the extraction unit 222 refers to the item evaluation table 32A for the user who is the item recommendation destination, and the evaluation in the “evaluation” column is “1 (correct evaluation)”. A correct evaluation item list in which all item IDs of information are extracted and listed is generated. Similarly, the extraction unit 222 generates a non-correct evaluation item list in which all item IDs of evaluation information whose “evaluation” column is “0 (non-correct evaluation)” are extracted and listed. The extraction unit 222 passes the generated correct evaluation item list and non-correct evaluation item list to the specifying unit 223.

  Next, in step S102, the specifying unit 223 determines whether the list length of the correct evaluation item list, that is, the number of item IDs included in the correct evaluation item list is 2 or more. If the list length is 2 or more, the process proceeds to step S103. If the list length is less than 2, the process proceeds to step S108.

  In step S <b> 103, the specifying unit 223 acquires the feature vector of each positive evaluation item indicated by the item ID included in the positive evaluation item list from the item feature table 31. Then, the specifying unit 223 obtains the positive evaluation item distribution 81 in the feature space, and uses the first principal component vector when the principal component analysis is applied to the positive evaluation item distribution 81 as the positive evaluation axis 82. Ask. Further, the specifying unit 223 generates a positive evaluation vector 87 by converting the positive evaluation axis 82 into a unit vector.

  Next, in step S104, the specifying unit 223 determines whether the list length of the non-correct evaluation item list is 2 or more. If the list length is 2 or more, the process proceeds to step S105. If the list length is less than 2, the process proceeds to step S107.

  In step S <b> 105, the specifying unit 223 obtains a non-correct evaluation item distribution 83 based on the item IDs included in the non-correct evaluation item list. Then, the specifying unit 223 obtains a vector of the second principal component when the principal component analysis is applied to the non-positive evaluation item distribution 83 as the non-positive evaluation orthogonal axis 85. Further, the specifying unit 223 generates a non-positive evaluation orthogonal vector 86 by converting the non-positive evaluation orthogonal axis 85 into a unit vector.

  Next, in step S106, the specifying unit 223 generates an interest vector 89 by converting the sum vector 88 of the positive evaluation vector 87 and the non-positive evaluation orthogonal vector 86 into a unit vector. Then, the specifying unit 223 passes the generated interest vector 89 to the determining unit 24, and the process returns to the item recommendation process illustrated in FIG.

  On the other hand, if a negative determination is made in step S104, the non-positive evaluation orthogonal vector 86 cannot be generated. Therefore, in the next step S107, the specifying unit 223 uses the positive evaluation vector 87 generated in step S103 as the interest vector 89. Set to. Then, the process returns to the item recommendation process shown in FIG.

  In step S108, the specifying unit 223 determines whether the list length of the non-correct evaluation item list is 2 or more. If the list length is 2 or more, the process proceeds to step S109. In step S109, as in step S105, the specifying unit 223 generates the non-positive evaluation orthogonal vector 86.

  When a negative determination is made in step S102, an affirmative determination is made in step S108, and the process proceeds to step S109, the positive evaluation vector 87 is not generated. Therefore, in the next step S110, the specifying unit 223 sets the non-positive evaluation orthogonal vector 86 generated in step S109 as the interest vector 89, and the process returns to the item recommendation process shown in FIG.

  On the other hand, if it is determined in step S108 that the list length is less than 2, neither the positive evaluation vector 87 nor the non-positive evaluation orthogonal vector 86 can be generated. Therefore, the interest vector 89 is not generated, and the processing is as shown in FIG. Return to the item recommendation process shown in FIG.

  As described above, according to the item recommendation device according to the second embodiment, not only the direction in which the dispersion of the positions of the non-correct evaluation items used in the first embodiment is minimized, but also the distribution of the positions of the positive evaluation items. The vector of interest is specified also using the direction in which the maximum is. Then, an unrecommended item located in the direction of the interest vector is determined as a recommended item. Thereby, it is possible to make a recommendation in consideration of whether the user is preferentially seeking a similar item or a phase preferentially seeking an unexpected item.

  More specifically, the user seeks unexpectedness for the recommended item when the recommended item does not meet the user's interest. In such a case, first, only items that are not interested from the beginning are recommended, and secondly, only similar items are recommended, so that the user initially selected with interest. There is a possibility that the interest will eventually fade and the recommended item will not be selected.

  The first case can be avoided because a non-positive evaluation orthogonal vector that generates unexpectedness is used in the construction of the interest vector. In the second case, the positive evaluation vector and the non-positive evaluation vector (the non-positive evaluation axis converted into a unit vector) become similar (the absolute value of the inner product approaches 0). For this reason, as a result, as in the first case, the influence of the non-positive evaluation orthogonal vector causing the unexpectedness is greatly reflected in the interest vector, and the second case can be avoided. If the user continues to be interested in the recommended item, the user continues to select the recommended item, so the influence of the positive evaluation vector is increased, the unexpectedness is suppressed, and the item similar to the already recommended item Continue to be recommended. In this way, it is possible to make a recommendation corresponding to each case where the user continues to be interested in the recommended item, does not have it from the beginning, or gradually becomes less interested.

  In the second embodiment, a case has been described in which a vector obtained by converting a sum vector of a non-positive evaluation orthogonal vector and a positive evaluation vector into a unit vector is generated as an interest vector. However, the present invention is not limited to this case. For example, the sum vector may be obtained by multiplying each of the non-positive evaluation orthogonal vector and the positive evaluation vector by a coefficient depending on whether the positive evaluation by the user is important or the non-positive evaluation is important. .

<Third Embodiment>
Next, a third embodiment will be described. In addition, in the item recommendation system which concerns on 3rd Embodiment, about the part similar to the item recommendation system 10 which concerns on 1st Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  As shown in FIG. 18, the item recommendation system 310 according to the third embodiment includes an item recommendation device 320, a storage device 30, and a user side device 340.

  The user side device 340 includes a control unit 341, a display unit 342, a remote controller 343, and an electroencephalograph 344. The display unit 342 has the same function as the display device of the user terminal 40 in the first embodiment, and displays a screen including a recommended item display area 72 as shown in FIG. 9, for example. The remote controller 343 has functions similar to those of the input device such as a mouse, a keyboard, and a touch panel of the user terminal 40 in the first embodiment, and transmits various signals to the control unit 341 according to the operation of the user 45. The electroencephalograph 344 is worn on the head of the user 45, measures the electroencephalogram of the user 45, and transmits measurement data to the control unit 341.

  The control unit 341 can be realized by a computer having a CPU, a storage unit, a memory, and the like. The control unit 341 performs control to display information on the recommended item transmitted from the item recommendation device 320 on the display unit 342. In addition, the control unit 341 performs control to generate item selection information and transmit it to the item recommendation device 320. The generation of item selection information will be described in more detail.

  As illustrated in FIG. 19, the control unit 341 includes an electroencephalogram acquisition unit 314A, a P600 detection unit 341B, a selection signal reception unit 341C, and a generation transmission unit 341D as functional units related to generation of item selection information.

  The electroencephalogram acquisition unit 341 </ b> A acquires measurement data indicating the electroencephalogram of the user 45 measured by the electroencephalograph 344. The P600 detection unit 341B uses an event-related potential P600 component that represents a positive voltage displacement 600 milliseconds after the reference timing, based on the measurement data acquired by the electroencephalogram acquisition unit 341A. To detect. When the P600 component is detected, it indicates that the user has a negative evaluation such as disappointment or no interest in the recommended item presented. As in the first and second embodiments, it is possible to obtain an explicit non-positive evaluation by detecting the P600 component as compared with the case where the user does not select the recommended item as a non-positive evaluation. . In addition, it is not limited to P600 component, You may use the other biometric information which can determine that the user had negative evaluation. By detecting biometric information and regarding it as a non-positive evaluation, the burden on the user is small compared to a method in which the user is directly inquired about the evaluation to obtain an answer.

  The selection signal receiving unit 341 </ b> C receives a selection signal transmitted from the remote controller 343 when the user 45 operates the remote controller 343 to select a recommended item displayed on the display unit 342.

  The generation transmission unit 341D generates item selection information based on the detection result of the P600 detection unit 341B and the reception result of the selection signal reception unit 341C. Specifically, when the P600 component is detected by the P600 detection unit 341B, the generation transmission unit 341D includes the item ID and information indicating that the P600 component is detected in the item selection information. The item ID is the item ID of the recommended item displayed on the display unit 342 at the reference timing of P600 component detection. In addition, when the selection signal is received by the selection signal reception unit 341C, the generation transmission unit 341D includes the item ID and information indicating that the selection signal has been received in the item selection information. The item ID is the item ID of the recommended item displayed on the display unit 342 when the selection signal is received by the selection signal receiving unit 341C. Further, the generation transmission unit 341D includes the user ID in the item selection information. The generation transmission unit 341D transmits the generated item selection information to the item recommendation device 320.

  Note that the recommended item display area 72 is recommended during the period for determining whether or not the P600 component is detected by the P600 detection unit 341B and the period for determining whether or not the selection signal is received by the selection signal receiving unit 341C. A predetermined period can be set after the item is displayed. The predetermined period may be, for example, a predetermined period, or a period from when a recommended item is displayed in the recommended item display area 72 to when the next recommended item is displayed.

  As illustrated in FIG. 5, the item recommendation device 320 includes an acquisition unit 321, an extraction unit 22, a specification unit 23, and a determination unit 24 as functional units.

  The acquisition unit 321 records user evaluation information on the presented recommended item in the item evaluation table 332A of the corresponding user in the evaluation information DB 332. FIG. 20 shows an example of the item evaluation table 332A. The user evaluation recorded in the “evaluation” column is different from the item evaluation table 32A in the first embodiment. In the item evaluation table 332A according to the third embodiment, when the P600 component is detected, the user's evaluation is regarded as a non-positive evaluation, and the recommended item that the P600 component is not detected and the user is presented is selected. In such cases, a positive evaluation is made. If no P600 component is detected and the recommended item presented by the user is not selected, no evaluation is given (no evaluation). For example, the positive evaluation can be “1”, the non-positive evaluation can be “−1”, and the no evaluation can be “0”.

  Specifically, the acquisition unit 321 records the item ID and the presentation time of the recommended item determined by the determination unit 24 and presented to the user in the item evaluation table 332A of the corresponding user in the evaluation information DB 332. In addition, the acquisition unit 321 sets the corresponding “evaluation” field to “0 (no evaluation)” as an initial value. In addition, the acquisition unit 321 acquires item selection information transmitted from the user terminal 40 when a recommended item presented to the user is selected by the user. As described above, the item selection information includes the user ID, information indicating whether or not the P600 component has been detected, and information indicating whether or not the selection signal has been received. The acquisition unit 321 specifies the item evaluation table 332A of the corresponding user based on the user ID included in the item selection information. When the item selection information includes information indicating that P600 is detected, the acquisition unit 321 displays an “evaluation” column corresponding to the item ID included in the item selection information in the specified item evaluation table 332A. Update to “0 (non-positive evaluation)”. If the item selection information does not include information indicating that P600 has been detected and includes information indicating that a selection signal has been received, the acquisition unit 321 sets “1 (correct) Update to "evaluation)".

  The item recommendation device 320 can be realized by, for example, the computer 50 shown in FIG. An item recommendation program 360 for causing the computer 50 to function as the item recommendation device 320 is stored in the storage unit 53 of the computer 50. The item recommendation program 360 includes an acquisition process 361, an extraction process 62, a specification process 63, and a determination process 64.

  CPU51 reads the item recommendation program 360 from the memory | storage part 53, expand | deploys to the memory 52, and executes the process which the item recommendation program 360 has sequentially. The CPU 51 operates as the acquisition unit 321 illustrated in FIG. 5 by executing the acquisition process 361. Other processes are the same as those of the item recommendation program 60 in the first embodiment. As a result, the computer 50 that has executed the item recommendation program 360 functions as the item recommendation device 320.

  The function realized by the item recommendation program 360 can be realized by, for example, a semiconductor integrated circuit, more specifically, an ASIC or the like.

  Next, the operation of the item recommendation system 310 according to the third embodiment will be described. In the third embodiment, since the evaluation information acquisition process executed in step S30 of the item recommendation process shown in FIG. 11 is only different from the item recommendation process according to the first embodiment, referring to FIG. An evaluation information acquisition process in the third embodiment will be described. In addition, about the process same as the evaluation information acquisition process in 1st Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  In step S31 of the evaluation information acquisition process illustrated in FIG. 21, the determination unit 24 presents the determined recommended item information to the user. In addition, the determination unit 24 notifies the acquisition unit 321 of the item ID of the recommended item presented to the user, the presentation time, and the user ID of the recommended user. Upon receiving the notification, the acquisition unit 321 records the item ID and the presentation time in the item evaluation table 332A of the corresponding user specified by the user ID, and sets the corresponding “evaluation” field as an initial value “0 (no evaluation)”. To "".

  Next, in step S331, the acquisition unit 332 determines whether or not a P600 component has occurred in the brain wave of the user 45 for the recommended item. If the acquisition unit 332 acquires item selection information from the user terminal 40 and the acquired item selection information includes information indicating that the P600 component has been detected, an affirmative determination is made, and the process proceeds to step S34. On the other hand, when the information indicating that the P600 component is detected is not included in the item selection information, a negative determination is made, and the process proceeds to step S332.

  In step S34, the acquisition unit 321 specifies the item evaluation table 332A of the corresponding user based on the user ID included in the item selection information. Then, the acquisition unit 321 updates the “evaluation” column corresponding to the item ID included in the item selection information to “0 (non-correct evaluation)” in the identified item evaluation table 332A, thereby removing the corresponding item. Record as a positive evaluation item. Then, the process returns to the item recommendation process shown in FIG.

  On the other hand, if the determination in step S331 is negative, the process proceeds to step S332. In step S332, the acquisition unit 321 determines whether the user has selected a recommended item. If the item selection information includes information indicating that the selection signal has been received, an affirmative determination is made and the process proceeds to step S33. In step S33, the acquisition unit 321 records the corresponding item as a positive evaluation item by updating the “evaluation” field corresponding to the item ID included in the item selection information to “1 (correct evaluation)”. Then, the process returns to the item recommendation process shown in FIG.

  On the other hand, if the determination in step S332 is negative, the process directly returns to the item recommendation process shown in FIG. Since the initial value in the “evaluation” column of the item evaluation table 332A is set to “0 (no evaluation)”, the recommended item that is not detected by the user and is not selected by the user is regarded as an item that is not evaluated as it is. It will be recorded in the evaluation table 332A.

  As described above, according to the item recommendation device according to the third embodiment, the user's non-positive evaluation for the recommended item is acquired from the biological information such as the P600 component of the brain wave. For this reason, an explicit non-positive evaluation can be obtained without imposing a burden of directly inquiring the user. Avoiding items that have a negative evaluation more clearly by determining recommended items that are located in the direction of the interest vector that contains the non-positive evaluation orthogonal vector generated using explicit non-positive evaluation items Can do.

  In the third embodiment, as in the first and second embodiments, a user terminal 40 such as a PC may be used instead of the user device 340. In this case, the user terminal 40 has a function of receiving measurement data transmitted from the electroencephalograph 344. Also in the first and second embodiments, instead of the user terminal 40, a user side device 340 that does not include the electroencephalograph 344 may be used.

  Moreover, although each said embodiment demonstrated the case where the recommendation item was shown one by one, you may make it show a some recommendation item at once. In this case, for example, the distance between the position of the unrecommended item projected on the axis of interest and the position of the latest recommended item, or the upper predetermined number in the order of the distance between the position of the unrecommended item and the axis of interest. An unrecommended item may be determined as a recommended item.

  Also, in each of the above embodiments, when using a web service such as online shopping, the item recommendation process illustrated in FIG. 11 is looped in the background, and the recommended item is switched every time T seconds elapse. It is not limited. For example, the process of steps S10 to S30 of the item recommendation process shown in FIG. 11 may be executed once every time the user is instructed to recommend an item.

  Moreover, although each said embodiment demonstrated the case where items, such as goods, were recommended in web services, such as online shopping, it is not limited to this case. For example, in a news site or the like, each article can be handled as an item and an article recommended to the user can be recommended.

  In the above description, the item recommendation programs 60, 260, and 360 are stored (installed) in the storage unit 53 in advance. However, the present invention is not limited to this. The program according to the disclosed technology can be provided in a form recorded on a recording medium such as a CD-ROM, a DVD-ROM, or a USB memory.

  Regarding the above embodiments, the following additional notes are disclosed.

(Appendix 1)
On the computer,
When recommending any item included in a plurality of items to a recommendation destination, the already recommended item included in the plurality of items, and the evaluation information of the recommendation destination for the already recommended item, Referring to the storage unit that stores the information in association with each recommendation destination, out of the already recommended items, the non-correct evaluation item associated with the evaluation information indicating evaluation other than positive evaluation is selected as the recommendation. Extract every destination,
The position of the extracted non-correct evaluation item in a space having a coordinate axis corresponding to each of the plurality of attributes of the item and in which the position of the item is determined corresponding to the combination of the attribute values of the item for the plurality of attributes Identify the direction in which the dispersion of
An item recommendation program for executing a process including determining an item to be recommended to the recommendation destination based on the identified direction and the position of any one of the plurality of items in the space.

(Appendix 2)
Determining the item to be recommended means that, in the space, among the unrecommended items included in the plurality of items, each position of the unrecommended item passes through the position of the item recommended most recently. The item recommendation program according to claim 1, further comprising: determining the unrecommended item closest to the position of the item recommended most recently as a position projected on a straight line parallel to the direction as an item recommended for the recommendation destination. .

(Appendix 3)
Determining the item to be recommended means that, in the space, among the unrecommended items included in the plurality of items, the distance from the straight line parallel to the direction passing through the position of the item recommended most recently is the shortest The item recommendation program according to supplementary note 1, including determining the unrecommended item as an item recommended to the recommendation destination.

(Appendix 4)
In addition to extracting the non-positive evaluation item, referring to the storage unit, out of the already recommended items, extract a positive evaluation item associated with the evaluation information indicating a positive evaluation,
When specifying the direction, in the space, based on a first direction that minimizes the dispersion of the positions of the non-positive evaluation items and a second direction that maximizes the dispersion of the positions of the positive evaluation items. Identify the third direction,
The item to be recommended is to determine an item to be recommended to the recommendation destination based on the identified third direction and the position of any one of the plurality of items in the space. Including Item recommendation program according to appendix 1.

(Appendix 5)
The item recommendation program according to supplementary note 4, wherein a direction of a vector obtained by combining the unit vector in the first direction and the unit vector in the second direction is specified as the third direction.

(Appendix 6)
Determining the item to be recommended means that, in the space, among the unrecommended items included in the plurality of items, each position of the unrecommended item passes through the position of the item recommended most recently. The appendix 4 or appendix 5, wherein the unrecommended item whose position projected onto a straight line parallel to the third direction is closest to the position of the item recommended most recently is determined as an item recommended for the recommendation destination. Item recommendation program.

(Appendix 7)
Determining the recommended item is a distance between a straight line parallel to the third direction passing through the position of the most recently recommended item among the unrecommended items included in the plurality of items in the space. The item recommendation program according to supplementary note 4 or supplementary note 5, wherein the unrecommended item closest to the item is determined as an item recommended to the recommendation destination.

(Appendix 8)
Any one of appendix 1 to appendix 7 for causing the computer to execute a process further including obtaining evaluation information indicating an evaluation other than the positive evaluation from a change in biometric information of the user who is the recommendation destination Item recommendation program of item 1.

(Appendix 9)
The item recommendation program according to appendix 8, wherein a P600 component generated in association with an event is acquired from an electroencephalogram measured by an electroencephalograph attached to a user as the change in the biological information.

(Appendix 10)
When recommending any item included in a plurality of items to a recommendation destination, the already recommended item included in the plurality of items, and the evaluation information of the recommendation destination for the already recommended item, Referring to the storage unit that stores the information in association with each recommendation destination, out of the already recommended items, the non-correct evaluation item associated with the evaluation information indicating evaluation other than positive evaluation is selected as the recommendation. An extraction unit for extracting each destination;
In the space that has coordinate axes corresponding to each of the plurality of attributes of the item and that determines the position of the item corresponding to the combination of the attribute values of the item for the plurality of attributes, A specific unit for identifying a direction in which the dispersion of the positions of the positive evaluation items is minimized,
A determining unit that determines an item to be recommended to the recommendation destination based on the direction specified by the specifying unit and a position of any one of the plurality of items in the space;
Item recommendation device.

(Appendix 11)
The determination unit includes a straight line parallel to the direction passing through the position of the item recommended most recently for each position of the unrecommended item among the unrecommended items included in the plurality of items in the space. The item recommendation device according to appendix 10, wherein the unrecommended item whose projected position is closest to the position of the item recommended most recently is determined as an item recommended for the recommendation destination.

(Appendix 12)
In the space, the determination unit includes the unrecommended item that is closest to a straight line parallel to the direction passing through the position of the most recently recommended item among unrecommended items included in the plurality of items. Item recommendation apparatus of Additional remark 10 which determines as an item recommended to the said recommendation destination.

(Appendix 13)
The extraction unit refers to the storage unit and extracts a positive evaluation item associated with the evaluation information indicating a positive evaluation among the already recommended items,
The specifying unit is based on a first direction in which the dispersion of the positions of the non-correct evaluation items is minimized and a second direction in which the dispersion of the positions of the positive evaluation items is maximum in the space. Identify the direction of
The determination unit determines an item to be recommended to the recommendation destination based on the third direction specified by the specifying unit and the position of any one of the plurality of items in the space. The item recommendation device described.

(Appendix 14)
14. The item recommendation device according to appendix 13, wherein the specifying unit specifies a direction of a vector obtained by combining the unit vector in the first direction and the unit vector in the second direction as the third direction.

(Appendix 15)
In the space, the determination unit includes the third direction passing through the position of the most recently recommended item for each position of the unrecommended item among the unrecommended items included in the plurality of items. 15. The item recommendation device according to supplementary note 13 or supplementary note 14, wherein the unrecommended item whose position projected onto a parallel straight line is closest to the position of the most recently recommended item is determined as an item recommended for the recommendation destination.

(Appendix 16)
In the space, the determination unit includes the unrecommended item included in the plurality of items, the unrecommended item having the closest distance to a straight line parallel to the third direction passing through the position of the item recommended most recently. Item recommendation apparatus of supplementary note 13 or supplementary note 14 which determines a recommendation item as an item recommended to the said recommendation destination.

(Appendix 17)
The item recommendation device according to any one of Supplementary Note 10 to Supplementary Note 16, further including an acquisition unit that acquires evaluation information indicating an evaluation other than the positive evaluation from a change in biometric information of the user who is the recommendation destination.

(Appendix 18)
The item recommendation device according to appendix 17, wherein the acquisition unit acquires a P600 component generated in association with an event from an electroencephalogram measured by an electroencephalograph attached to a user as the change in the biological information.

(Appendix 19)
On the computer,
When recommending any item included in a plurality of items to a recommendation destination, the already recommended item included in the plurality of items, and the evaluation information of the recommendation destination for the already recommended item, Referring to the storage unit that stores the information in association with each recommendation destination, out of the already recommended items, the non-correct evaluation item associated with the evaluation information indicating evaluation other than positive evaluation is selected as the recommendation. Extract every destination,
The position of the extracted non-correct evaluation item in a space having a coordinate axis corresponding to each of the plurality of attributes of the item and in which the position of the item is determined corresponding to the combination of the attribute values of the item for the plurality of attributes Identify the direction in which the dispersion of
An item recommendation method for executing a process including determining an item to be recommended to the recommendation destination based on the identified direction and the position of any one of the plurality of items in the space.

(Appendix 20)
Determining the item to be recommended means that, in the space, among the unrecommended items included in the plurality of items, each position of the unrecommended item passes through the position of the item recommended most recently. The item recommendation method according to claim 19, further comprising: determining the unrecommended item closest to the position of the item recommended most recently as a position projected on a straight line parallel to the direction as an item recommended for the recommendation destination. .

(Appendix 21)
Determining the item to be recommended means that, in the space, among the unrecommended items included in the plurality of items, the distance from the straight line parallel to the direction passing through the position of the item recommended most recently is the shortest The item recommendation method according to appendix 19, including determining the unrecommended item as an item recommended to the recommendation destination.

(Appendix 22)
In addition to extracting the non-positive evaluation item, referring to the storage unit, out of the already recommended items, extract a positive evaluation item associated with the evaluation information indicating a positive evaluation,
When specifying the direction, in the space, based on a first direction that minimizes the dispersion of the positions of the non-positive evaluation items and a second direction that maximizes the dispersion of the positions of the positive evaluation items. Identify the third direction,
The item to be recommended is to determine an item to be recommended to the recommendation destination based on the identified third direction and the position of any one of the plurality of items in the space. Including Item recommendation method according to appendix 19.

(Appendix 23)
23. The item recommendation method according to appendix 22, wherein a direction of a vector obtained by combining the unit vector in the first direction and the unit vector in the second direction is specified as the third direction.

(Appendix 24)
Determining the item to be recommended means that, in the space, among the unrecommended items included in the plurality of items, each position of the unrecommended item passes through the position of the item recommended most recently. 24. The supplementary note 22 or the supplementary note 23, wherein the unrecommended item whose position projected on a straight line parallel to the third direction is closest to the position of the item recommended most recently is determined as an item recommended for the recommendation destination. Item recommendation method.

(Appendix 25)
Determining the recommended item is a distance between a straight line parallel to the third direction passing through the position of the most recently recommended item among the unrecommended items included in the plurality of items in the space. 24. The item recommendation method according to supplementary note 22 or supplementary note 23, wherein the unrecommended item that is closest is determined as an item recommended to the recommendation destination.

(Appendix 26)
Any one of appendix 19 to appendix 25, which causes the computer to execute processing further including obtaining evaluation information indicating an evaluation other than the positive evaluation from a change in biometric information of the user who is the recommendation destination. The item recommendation method described.

(Appendix 27)
27. The item recommendation method according to appendix 26, wherein a P600 component generated in association with an event is acquired from an electroencephalogram measured by an electroencephalograph attached to a user as the change in the biological information.

(Appendix 28)
On the computer,
When recommending any item included in a plurality of items to a recommendation destination, the already recommended item included in the plurality of items, and the evaluation information of the recommendation destination for the already recommended item, Referring to the storage unit that stores the information in association with each recommendation destination, out of the already recommended items, the non-correct evaluation item associated with the evaluation information indicating evaluation other than positive evaluation is selected as the recommendation. Extract every destination,
The position of the extracted non-correct evaluation item in a space having a coordinate axis corresponding to each of the plurality of attributes of the item and in which the position of the item is determined corresponding to the combination of the attribute values of the item for the plurality of attributes Identify the direction in which the dispersion of
An item recommendation program for executing a process including determining an item recommended for the recommendation destination based on the identified direction and the position of any one of the plurality of items in the space is stored. Storage medium.

10, 210, 310 Item recommendation system 20, 220, 320 Item recommendation device 21, 321 Acquisition unit 22, 222 Extraction unit 23, 223 Identification unit 24 Determination unit 30 Storage device 31 Item feature table 32, 332 Evaluation information database (DB)
32A, 332A Item Evaluation Table 33 Item Presentation Information DB
40 user terminal 45 user 50 computer 51 CPU
52 Memory 53 Memory 60, 260, 360 Item recommendation program 72 Recommended item display area 81 Positive evaluation item distribution 82 Positive evaluation axis 83 Non-positive evaluation item distribution 84 Non-positive evaluation axis 85 Non-positive evaluation orthogonal axis 86 Non-positive evaluation Orthogonal vector 87 Positive evaluation vector 89 Interest vector 90 Interest axis 340 User side device 341 Control unit 314A EEG acquisition unit 341B P600 detection unit 341C Selection signal reception unit 341D Generation transmission unit 344 EEG

Claims (9)

On the computer,
When recommending any item included in a plurality of items to a recommendation destination, the already recommended item included in the plurality of items, and the evaluation information of the recommendation destination for the already recommended item, Referring to the storage unit that stores the information in association with each recommendation destination, out of the already recommended items, the non-correct evaluation item associated with the evaluation information indicating evaluation other than positive evaluation is selected as the recommendation. Extract every destination,
The position of the extracted non-correct evaluation item in a space having a coordinate axis corresponding to each of the plurality of attributes of the item and in which the position of the item is determined corresponding to the combination of the attribute values of the item for the plurality of attributes Identify the direction in which the dispersion of
An item recommendation program for executing a process including determining an item to be recommended to the recommendation destination based on the identified direction and the position of any one of the plurality of items in the space.   Determining the item to be recommended means that, in the space, among the unrecommended items included in the plurality of items, each position of the unrecommended item passes through the position of the item recommended most recently. The item recommendation according to claim 1, further comprising: determining the unrecommended item closest to the position of the item recommended most recently as a position projected on a straight line parallel to the direction as an item recommended for the recommendation destination. program. In addition to extracting the non-positive evaluation item, referring to the storage unit, out of the already recommended items, extract a positive evaluation item associated with the evaluation information indicating a positive evaluation,
When specifying the direction, in the space, based on a first direction that minimizes the dispersion of the positions of the non-positive evaluation items and a second direction that maximizes the dispersion of the positions of the positive evaluation items. Identify the third direction,
The item to be recommended is to determine an item to be recommended to the recommendation destination based on the identified third direction and the position of any one of the plurality of items in the space. The item recommendation program according to claim 1.   The item recommendation program according to claim 3, wherein a direction of a vector obtained by combining the unit vector in the first direction and the unit vector in the second direction is specified as the third direction.   Determining the item to be recommended means that, in the space, among the unrecommended items included in the plurality of items, each position of the unrecommended item passes through the position of the item recommended most recently. 5. The non-recommended item that is closest to the position of the item recommended most recently as a position projected on a straight line parallel to the third direction is determined as an item recommended for the recommendation destination. Item recommendation program.   6. The computer according to claim 1, further comprising: acquiring evaluation information indicating evaluation other than the positive evaluation from a change in biological information of the user who is the recommendation destination. Item recommendation program given in any 1 paragraph.   The item recommendation program according to claim 6, wherein a P600 component generated in association with an event is acquired from an electroencephalogram measured by an electroencephalograph attached to a user as the change in the biological information. When recommending any item included in a plurality of items to a recommendation destination, the already recommended item included in the plurality of items, and the evaluation information of the recommendation destination for the already recommended item, Referring to the storage unit that stores the information in association with each recommendation destination, out of the already recommended items, the non-correct evaluation item associated with the evaluation information indicating evaluation other than positive evaluation is selected as the recommendation. An extraction unit for extracting each destination;
In the space that has coordinate axes corresponding to each of the plurality of attributes of the item and that determines the position of the item corresponding to the combination of the attribute values of the item for the plurality of attributes, A specific unit for identifying a direction in which the dispersion of the positions of the positive evaluation items is minimized,
A determining unit that determines an item to be recommended to the recommendation destination based on the direction specified by the specifying unit and a position of any one of the plurality of items in the space;
Item recommendation device. On the computer,
When recommending any item included in a plurality of items to a recommendation destination, the already recommended item included in the plurality of items, and the evaluation information of the recommendation destination for the already recommended item, Referring to the storage unit that stores the information in association with each recommendation destination, out of the already recommended items, the non-correct evaluation item associated with the evaluation information indicating evaluation other than positive evaluation is selected as the recommendation. Extract every destination,
The position of the extracted non-correct evaluation item in a space having a coordinate axis corresponding to each of the plurality of attributes of the item and in which the position of the item is determined corresponding to the combination of the attribute values of the item for the plurality of attributes Identify the direction in which the dispersion of
An item recommendation method for executing a process including determining an item to be recommended to the recommendation destination based on the identified direction and the position of any one of the plurality of items in the space.