Similarity calculation system, similarity calculation method, and program

Abstract

To calculate similarity between evaluation objects reflecting the subjectivity of a user.SOLUTION: For each of the plurality of evaluation targets, a plurality of evaluation items for evaluating similarity between the evaluation targets are set. For each evaluation item of each evaluation target, an attribute value indicating the degree of the attribute of the evaluation target for each evaluation item is set. A similarity calculation system includes evaluation information acquisition means for acquiring information on an attribute value of each evaluation item of each evaluation object, and similarity calculation means for calculating similarity between evaluation objects on the basis of an attribute value of an evaluation item selected by a user or corresponding to user's preference.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a similarity calculation system, a similarity calculation method, and a program. [Background technology]

[0002] A system for calculating the similarity between a plurality of evaluation targets is known (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2020-086692 Summary of the Invention [Problem to be solved by the invention]

[0004] However, when calculating the similarity between multiple evaluation targets, the similarity may vary depending on the subjective opinion of the user who performs the evaluation. However, the above system does not fully reflect the subjective opinion of the user when calculating the similarity between the evaluation targets.

[0005] The present disclosure has been made to solve such problems, and its main purpose is to provide a similarity calculation system, a similarity calculation method, and a program that can calculate the similarity between evaluation objects that reflects the user's subjectivity. [Means for solving the problem]

[0006] In order to achieve the above object, one aspect of the present disclosure is to A plurality of evaluation items are set for each of the plurality of evaluation objects in order to evaluate similarity between the evaluation objects, and an attribute value indicating the degree of an attribute of the evaluation object for each evaluation item is set for each evaluation item of the plurality of evaluation objects, evaluation information acquisition means for acquiring information on attribute values ​​of each evaluation item of each of the evaluation targets; a similarity calculation means for calculating a similarity between the evaluation targets based on attribute values ​​of the evaluation items selected by a user or corresponding to the user's preferences; Equipped with Similarity calculation system is. In this aspect, The device further comprises a preference information acquisition means for acquiring preference information of a user, The similarity calculation means may select the evaluation items based on the user's preference information acquired by the preference information acquisition means, and calculate the similarity between the evaluation targets based on the attribute values ​​of the selected evaluation items. In this aspect, The system further includes a selection item acquisition means for acquiring information on the evaluation items selected by the user, The similarity calculation means may calculate the similarity between the evaluation targets based on the attribute values ​​of the evaluation items acquired by the selected item acquisition means. In order to achieve the above object, one aspect of the present disclosure is to A plurality of evaluation items are set for each of the plurality of evaluation objects in order to evaluate similarity between the evaluation objects, and an attribute value indicating the degree of an attribute of the evaluation object for each evaluation item is set for each evaluation item of the plurality of evaluation objects, acquiring information on attribute values ​​of each evaluation item of each of the evaluation targets; calculating a similarity between the evaluation targets based on attribute values ​​of the evaluation items selected by a user or corresponding to a user's preference; Including, Similarity calculation method is. In order to achieve the above object, one aspect of the present disclosure is to A plurality of evaluation items are set for each of the plurality of evaluation objects in order to evaluate similarity between the evaluation objects, and an attribute value indicating the degree of an attribute of the evaluation object for each evaluation item is set for each evaluation item of the plurality of evaluation objects, A process of acquiring information on attribute values ​​of each evaluation item of each evaluation target; A process of calculating a similarity between the evaluation targets based on attribute values ​​of the evaluation items selected by a user or corresponding to the user's preferences; to the computer, program is. [Effects of the Invention]

[0007] According to the present disclosure, it is possible to provide a similarity calculation system, a similarity calculation method, and a program that can calculate the similarity between evaluation targets that reflects the user's subjectivity. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a block diagram showing a schematic system configuration of a similarity calculation system according to an embodiment of the present invention. [Figure 2] FIG. 10 is a diagram illustrating an example of attribute value information. [Figure 3] FIG. 10 is a diagram for explaining a similarity calculation method according to the present embodiment. [Figure 4] 10 is a flowchart showing an example of the flow of a similarity calculation method according to the present embodiment. [Figure 5] 1 is a block diagram showing a schematic system configuration of a similarity calculation system according to an embodiment of the present invention. [Figure 6] 1 is a block diagram showing a schematic system configuration of a recommendation system according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0009] Embodiment 1 The present embodiment will be described below with reference to the drawings. For example, when calculating the similarity between evaluation targets such as multiple items, the similarity may be significantly influenced by the subjectivity (preference) of the user who performs the evaluation, and may vary.

[0010] In contrast, the similarity calculation system according to this embodiment calculates the similarity between multiple evaluation targets, fully reflecting the user's subjectivity. The evaluation targets include, for example, products, sentences, words, images including still images and videos, sounds, tourist spots, etc.

[0011] The similarity calculation system according to this embodiment has the hardware configuration of a typical computer, including, for example, a processor such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), internal memory such as a RAM (Random Access Memory) or a ROM (Read Only Memory), storage devices such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive), an input / output I / F for connecting peripheral devices such as a display, and a communication I / F 15 for communicating with devices external to the device.

[0012] 1 is a block diagram showing a schematic system configuration of a similarity calculation system according to this embodiment. The similarity calculation system 1 according to this embodiment includes an evaluation information acquisition unit 2 that acquires information on attribute values ​​of evaluation items, a similarity calculation unit 3 that calculates similarities between evaluation targets, and a preference information acquisition unit 4 that acquires user preference information.

[0013] For each of the multiple evaluation targets, multiple evaluation items are set to evaluate similarities between the evaluation targets. For each evaluation item of each evaluation target, an attribute value is set that indicates the degree of the attribute of the evaluation target for each evaluation item. If the degree of the attribute of the evaluation target for the evaluation item is high, the attribute value will be large. Conversely, if the degree of the attribute of the evaluation target for the evaluation item is low, the attribute value will be small.

[0014] The evaluation information acquisition unit 2 is a specific example of evaluation information acquisition means. The evaluation information acquisition unit 2 acquires information on the attribute values ​​of each evaluation item of each of the evaluation targets (hereinafter referred to as attribute value information).

[0015] Fig. 2 is a diagram showing an example of attribute value information. The attribute value information may be configured as table information as shown in Fig. 2. For example, as shown in Fig. 2, the evaluation targets are items 1 and 2, and the evaluation items are A (active), B (park), C (relaxed), D (nature), E (city), F (beach), and G (cafe). The attribute values ​​are flag values ​​expressed as 0 or 1.

[0016] For example, if the item is an image and the evaluation item is B (park), if the image includes a park, the attribute value is set to 1, and if the image does not include a park, the attribute value is set to 0. Note that the attribute value does not have to be limited to the two values ​​of 0 and 1; for example, a decimal value such as 0.1 to 0.9 or a gradational value such as 0 to 10 indicating the likelihood or degree of attribute may be set.

[0017] More specifically, when the evaluation item is F (coast), if the image contains an ocean, its attribute value is set to 1, while if the image contains a river, its attribute value is set to 0.5, and if the image contains a mountain, its attribute value is set to 0.1. This is because the ocean has the highest attribute level for the evaluation item F (coast), while rivers do not have a very high attribute level for the ocean, and mountains have a low attribute level.

[0018] The attribute value information may be stored in advance in a storage unit such as a storage device. The evaluation information acquisition unit 2 may acquire the attribute value information from a storage device or the like, or may acquire information on the Web (World Wide Web) via the Internet or the like.

[0019] For example, the evaluation information acquisition unit 2 may acquire hashtags used in web services or tags attached to products on mail-order sites via the Internet, and set the acquired tags as evaluation items. Furthermore, a user may input and set the evaluation information to the evaluation information acquisition unit 2 via an input device or the like.

[0020] The evaluation information acquisition unit 2 may generate attribute value information by generating evaluation items and attribute values ​​using a machine learning device such as a neural network. For example, when the evaluation target is an image, the evaluation information acquisition unit 2 may input the image to the machine learning device, automatically set evaluation items such as "relaxing" and "natural," and automatically assign the attribute values.

[0021] The similarity calculation unit 3 is a specific example of a similarity calculation means. The similarity calculation unit 3 calculates the similarity between evaluation targets based on the attribute value information acquired by the evaluation information acquisition unit 2.

[0022] Incidentally, when calculating the similarity between a plurality of evaluation targets, the similarity is greatly influenced by the subjective opinion of the user who performs the evaluation, and may vary.

[0023] In contrast to this, the similarity calculation unit 3 according to this embodiment calculates the similarity between the evaluation targets by reflecting the user's subjectivity in the evaluation of the similarity between the evaluation targets. This makes it possible to calculate the similarity between the evaluation targets that reflects the user's subjectivity.

[0024] The preference information acquisition unit 4 is a specific example of a preference information acquisition means. The preference information acquisition unit 4 acquires user preference information. The preference information is information indicating the user's preferences, such as hobbies, favorite colors, favorite shapes, and favorite places.

[0025] In addition, the preference information may include browsing history information when the user browses the web, behavior history information when the user acts on the web, search history information when the user searches on the web, click history information when the user clicks on the web, and purchase history information when the user purchases products on the web.

[0026] The user's preference information may be stored in advance in a storage unit such as a storage device. The preference information acquisition unit 4 can acquire the preference information from a storage device or the like. Alternatively, the user may input the preference information to the preference information acquisition unit 4 via an input device or the like.

[0027] The similarity calculation unit 3 selects at least one evaluation item corresponding to the user's preference from among the multiple evaluation items set in the attribute value information, based on the preference information acquired by the preference information acquisition unit 4. The similarity calculation unit 3 may select, for example, from among the multiple evaluation items in the attribute value information, an evaluation item that has a high degree of correlation or relevance with the preference information.

[0028] Furthermore, the similarity calculation unit 3 may select, from among a plurality of evaluation items of the attribute value information, an evaluation item corresponding to an object (such as a product) included in, for example, browsing history information, behavior history information, search history information, purchase history information, etc. of the preference information. Note that "corresponding to an object" refers to, for example, an evaluation item that is the same as the object or an evaluation item that has a high relevance to the object.

[0029] The similarity calculation unit 3 may select, from among the multiple evaluation items of the attribute value information, an evaluation item corresponding to an item whose number of views, number of searches, or number of purchases is greater than or equal to a predetermined value, for example, based on the browsing history information or search history information of the preference information.

[0030] The similarity calculation unit 3 may calculate the similarity between the evaluation targets based only on the attribute values ​​of the selected evaluation items. This allows the similarity between the evaluation targets to be calculated using only the attribute values ​​of the evaluation items related to the user's preferences, making it possible to calculate the similarity between the evaluation targets that reflects the user's subjectivity.

[0031] The similarity calculation unit 3 may calculate the similarity between the evaluation targets by increasing the weighting of only the attribute values ​​of the selected evaluation items. The similarity calculation unit 3 may increase the weighting by, for example, setting a weighting coefficient greater than 1 (weighting coefficient > 1). This allows weighting and emphasizing only the attribute values ​​of the evaluation items related to the user's preferences, thereby making it possible to calculate the similarity between the evaluation targets that reflects the user's subjectivity.

[0032] The similarity calculation unit 3 may calculate the similarity between the evaluation targets by lowering the weighting of attribute values ​​other than the selected evaluation items. Furthermore, the similarity calculation unit 3 may calculate the similarity between the evaluation targets by increasing the weighting of the attribute value of an evaluation item having a higher correlation or relevance with preference information. Also, for example, the user may prioritize the evaluation items, and the similarity calculation unit 3 may increase the weighting of the attribute value of the evaluation item as the priority of the evaluation item increases.

[0033] Note that user preferences are influenced by current trends. Therefore, the similarity calculation unit 3 may select at least one evaluation item from among the multiple evaluation items set in the attribute value information based on not only the preference information acquired by the preference information acquisition unit 4 but also information on currently popular items (trend information).

[0034] This makes it possible to narrow down the evaluation items to those that more appropriately reflect the user's preferences. The similarity calculation unit 3 may acquire the trend information from, for example, the Web. For example, the similarity calculation unit 3 may select, from among the multiple evaluation items of the attribute value information, an evaluation item that has a high correlation or relevance with the preference information and is included in the trend information.

[0035] Furthermore, the similarity calculation unit 3 may select at least one evaluation item from a plurality of evaluation items set in the attribute value information based on preference information corresponding to the attribute to which the user belongs (male, female, age group, hobby, occupation, etc.). In this case, a database in which preference information is linked to each attribute may be constructed in advance.

[0036] The similarity calculation unit 3 calculates the similarity between the evaluation targets by calculating, for example, the cosine similarity or Euclidean distance between the evaluation targets based on the attribute values ​​of the evaluation items selected as described above.

[0037] A specific example of the similarity calculation method according to this embodiment will now be described with reference to Fig. 2 and Fig. 3. For example, the similarity calculation system 1 calculates the similarity between items 1 and 2 as follows. First, the evaluation information acquisition unit 2 acquires attribute value information of items 1 and 2 as shown in Fig. 2. Furthermore, the preference information acquisition unit 4 acquires preference information of users (1) and (2).

[0038] The similarity calculation unit 3 selects evaluation items B, C, F, and G that have a high correlation with the preferences of the user (1) from among the multiple evaluation items A to G set in the attribute value information, based on the preference information of the user (1) acquired by the preference information acquisition unit 4. The similarity calculation unit 3 calculates the similarity (e.g., 0.75) between items 1 and 2 based on the attribute values ​​of the evaluation items B, C, F, and G, as shown in Fig. 3(a).

[0039] In this case, if the subjective opinion of user (1) is reflected in the evaluation of the similarity, the similarity between items 1 and 2 will be high, and items 1 and 2 will be evaluated as similar.

[0040] On the other hand, the similarity calculation unit 3 selects evaluation items A, B, D, and E that correspond to the preferences of the user (2) from among the multiple evaluation items A to G set in the attribute value information, based on the preference information of the user (2) acquired by the preference information acquisition unit 4. As shown in FIG. 3(b), the similarity calculation unit 3 calculates the similarity (for example, 0) between items 1 and 2 based on the attribute values ​​of the evaluation items A, B, D, and E. In this case, if the subjectivity of the user (2) is reflected in the evaluation of the similarity, the similarity between items 1 and 2 will be low, and the items 1 and 2 will be evaluated as not being similar.

[0041] That is, the evaluation of the similarity differs depending on the subjective opinions of users (1) and (2), and according to this embodiment, an appropriate similarity can be calculated according to the subjective opinions of users (1) and (2).

[0042] Next, a similarity calculation method according to this embodiment will be described below. Fig. 4 is a flowchart showing an example of the flow of the similarity calculation method according to this embodiment.

[0043] The evaluation information acquisition unit 2 acquires attribute value information (step S101). The preference information acquisition unit 4 acquires user preference information (step S102).

[0044] Based on the preference information acquired by the preference information acquisition unit 4, the similarity calculation unit 3 selects at least one evaluation item that corresponds to the user's preference from among the multiple evaluation items set in the attribute value information (step S103).

[0045] The similarity calculation unit 3 calculates the similarity between the evaluation targets based on the attribute values ​​of the selected evaluation items (step S104).

[0046] As described above, according to the similarity calculation method according to this embodiment, it is possible to calculate the similarity between evaluation targets that reflects the user's preferences.

[0047] Embodiment 2 5 is a block diagram showing a schematic system configuration of a similarity calculation system according to this embodiment. The similarity calculation system 20 according to this embodiment may include a selected item acquisition unit 5 that acquires information on evaluation items selected by a user (hereinafter, selected evaluation item information) instead of the preference information acquisition unit 4. The selected item acquisition unit 5 is a specific example of a selected item acquisition means.

[0048] This selected evaluation item information includes evaluation items selected by the user himself / herself, and therefore reflects the user's subjectivity. Therefore, as will be described later, by calculating the similarity between evaluation targets using the selected evaluation item information, it is possible to calculate the similarity between evaluation targets that reflects the user's subjectivity.

[0049] The selected evaluation item information may be stored in advance in a storage unit such as a storage device. The evaluation information acquisition unit 2 can acquire the selected evaluation item information from a storage device or the like. Alternatively, the user may input the selected evaluation item information to the evaluation information acquisition unit 2 via an input device or the like.

[0050] The similarity calculation unit 3 calculates the similarity between the evaluation targets based on the attribute values ​​of the evaluation items included in the selected evaluation item information acquired by the selected item acquisition unit 5.

[0051] The similarity calculation unit 3 may calculate the similarity between evaluation targets by weighting only the attribute values ​​of the evaluation items included in the selected evaluation item information acquired by the selected item acquisition unit 5. In this way, by weighting and emphasizing only the attribute values ​​of the evaluation items selected by the user himself / herself, it is possible to calculate the similarity between evaluation targets that reflects the user's subjectivity.

[0052] Embodiment 3 6 is a block diagram showing a schematic system configuration of a recommendation system according to this embodiment. The recommendation system 10 according to this embodiment recommends items that match the preferences of the user to the user.

[0053] The recommendation system 10 according to this embodiment includes the above-described similarity calculation systems 1 and 20, and a recommendation unit 11 that recommends items. The recommendation unit 11 recommends to the user items whose similarity calculated by the similarity calculation system is equal to or greater than a predetermined value.

[0054] Users tend to seek items similar to items they have viewed in the past. For example, let item 1 in Figure 2 above be an item the user has viewed in the past, and item 2 above be a candidate item for recommendation.

[0055] As described above, the similarity calculation unit 3 selects evaluation items of attribute value information based on user preference information, etc., and calculates the similarity between items 1 and 2 based on the attribute values ​​of the selected evaluation items. If the recommendation unit 11 determines that the similarity between items 1 and 2 is equal to or greater than a predetermined value and items 1 and 2 are similar, it recommends item 2 to the user.

[0056] If the similarity between items 1 and 2 is less than a predetermined value and items 1 and 2 are not similar, the similarity calculation unit 3 may repeat the calculation of the similarity until an item similar to item 1 is found. In this case, the attribute value information may include information on the attribute values ​​of the evaluation items of items 1, 2, 3, ..., n.

[0057] The recommendation unit 11 may recommend, for example, item 2 determined to be similar to item 1 to the user by displaying it as an advertisement or recommended product on the screen of a mobile terminal such as a smartphone.

[0058] Although several embodiments of the present disclosure have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be embodied in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, and are also included in the scope of the invention and its equivalents as defined in the claims.

[0059] The present disclosure can also be implemented by causing a processor to execute a computer program to perform the processing shown in FIG. 4, for example.

[0060] The program can be stored and supplied to a computer using various types of non-transitory computer readable media. Non-transitory computer readable media include various types of tangible storage media. Examples of non-transitory computer readable media include magnetic recording media (e.g., flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (e.g., magneto-optical disks), CD-ROMs (Read Only Memory), CD-Rs, CD-R / Ws, and semiconductor memories (e.g., mask ROMs, PROMs (Programmable ROMs), EPROMs (Erasable PROMs), flash ROMs, and RAMs (Random Access Memory)).

[0061] The program may be provided to the computer by various types of transitory computer-readable media. Examples of transitory computer-readable media include electrical signals, optical signals, and electromagnetic waves. The transitory computer-readable media can provide the program to the computer via a wired communication path such as an electrical wire or optical fiber, or via a wireless communication path.

[0062] Each component of the similarity calculation system 1, 20 according to the above-described embodiments can be realized not only by a program, but also in part or in whole by dedicated hardware such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field-Programmable Gate Array). [Explanation of symbols]

[0063] 1 Similarity calculation system, 2 Rating information acquisition unit, 3 Similarity calculation unit, 4 Preference information acquisition unit, 5 Selection item acquisition unit, 10 Recommendation system, 11 Recommendation unit, 20 Similarity calculation system

Claims

1. A plurality of evaluation items are set for each of the plurality of evaluation objects in order to evaluate similarity between the evaluation objects, and an attribute value indicating the degree of an attribute of the evaluation object for each evaluation item is set for each evaluation item of the plurality of evaluation objects, evaluation information acquisition means for acquiring information on attribute values ​​of each evaluation item of each of the evaluation targets; a similarity calculation means for calculating a similarity between the evaluation targets based on attribute values ​​of the evaluation items selected by a user or corresponding to the user's preferences; Equipped with Similarity calculation system.

2. 2. The similarity calculation system according to claim 1, The device further comprises a preference information acquisition means for acquiring preference information of a user, the similarity calculation means selects the evaluation items based on the user's preference information acquired by the preference information acquisition means, and calculates the similarity between the evaluation targets based on the attribute values ​​of the selected evaluation items; Similarity calculation system.

3. 2. The similarity calculation system according to claim 1, The system further includes a selection item acquisition means for acquiring information on the evaluation items selected by the user, the similarity calculation means calculates the similarity between the evaluation targets based on the attribute values ​​of the evaluation items acquired by the selection item acquisition means; Similarity calculation system.

4. A plurality of evaluation items are set for each of the plurality of evaluation objects in order to evaluate similarity between the evaluation objects, and an attribute value indicating the degree of an attribute of the evaluation object for each evaluation item is set for each evaluation item of the plurality of evaluation objects, acquiring information on attribute values ​​of each evaluation item of each of the evaluation targets; calculating a similarity between the evaluation targets based on attribute values ​​of the evaluation items selected by a user or corresponding to a user's preference; Including, Similarity calculation method.

5. A plurality of evaluation items are set for each of the plurality of evaluation objects in order to evaluate similarity between the evaluation objects, and an attribute value indicating the degree of an attribute of the evaluation object for each evaluation item is set for each evaluation item of the plurality of evaluation objects, A process of acquiring information on attribute values ​​of each evaluation item of each evaluation target; A process of calculating a similarity between the evaluation targets based on attribute values ​​of the evaluation items selected by a user or corresponding to the user's preferences; to the computer, program.

Images