Search device, search method, and program

Abstract

The present invention provides a search device comprising: an input accepting unit for accepting a search query in which an attribute value and a confidence level of the attribute value are set for at least one attribute; a modifying unit for modifying the set search query on the basis of the attribute value and the confidence level to generate a modified search query; and a search unit for searching for registered information matching the modified search query from among registered information registered in a database.

Description

Search device, search method, and recording medium

[0001] The present invention relates to a search device, a search method, and a program.

[0002] Techniques related to the present invention are disclosed in Patent Documents 1 to 5. The techniques disclosed in Patent Documents 1 to 5 search for data that matches a search query set by a user.

[0003] International Publication No. 2022 / 201541 International Publication No. 2020 / 255307 Japanese Patent Application Laid-Open No. 2020-198053 Japanese Patent Application Laid-Open No. 2018-142160 Japanese Patent Application Laid-Open No. 2011-170778

[0004] A user sets a search query that specifies attribute values ​​(features) of a desired search target, and searches for the desired search target. Examples of searches include, but are not limited to, searching for images containing a desired person, searching for images taken at a desired location, searching for images containing a desired object, etc.

[0005] In such a search, if the content of the search query is inappropriate, the desired search target may not be found. "Inappropriate content of the search query" refers to a state in which the attribute values ​​set in the search query differ from the attribute values ​​of the desired search target. For example, the content of the search query may be inappropriate when the attribute values ​​of the desired search target are unclear at the time of the search and the user sets the search query based on vague memory or information heard from another person. None of Patent Documents 1 to 5 discloses such a problem or a solution.

[0006] In view of the above-mentioned problems, an example of an object of the present invention is to provide a search device, a search method, and a program that solve the problem of enabling a desired search target to be searched for with high accuracy.

[0007] According to one aspect of the present invention, there is provided a search device having: an input receiving means for receiving a search query in which an attribute value and a certainty of the attribute value are set for at least one attribute; a modifying means for modifying the set search query based on the attribute value and the certainty to generate a modified search query; and a search means for searching for registered information that matches the modified search query from registered information registered in a database.

[0008] According to one aspect of the present invention, a search method is provided in which one or more computers receive a search query in which an attribute value and a certainty of the attribute value are set for at least one attribute, modify the set search query based on the attribute value and the certainty, generate a modified search query, and search for registered information that matches the modified search query from registered information registered in a database.

[0009] According to one aspect of the present invention, a program is provided that causes a computer to function as: an input receiving means that receives a search query in which an attribute value and a certainty of the attribute value are set for at least one attribute; a modifying means that modifies the set search query based on the attribute value and the certainty and generates a modified search query; and a search means that searches for registered information that matches the modified search query from registered information registered in a database.

[0010] According to one aspect of the present invention, a search device, a search method, and a program are realized that solve the problem of enabling a desired search target to be searched for with high accuracy.

[0011] The above-mentioned objects and other objects, features and advantages will become more apparent from the following description of the preferred embodiments and the accompanying drawings.

[0012] FIG. 1 is a diagram showing an example of a functional block diagram of a search device. FIG. 2 is a diagram showing an example of a hardware configuration of a search device. FIG. 3 is a diagram showing an example of reference information. FIG. 4 is a diagram showing an example of index information. FIG. 5 is a flowchart showing an example of a processing flow of the search device. FIG. 6 is a flowchart showing another example of a processing flow of the search device. FIG. 7 is a diagram showing another example of reference information. FIG. 8 is a diagram showing another example of reference information. FIG. 9 is a diagram showing another example of reference information.

[0013] Hereinafter, embodiments of the present invention will be described with reference to the drawings. In all the drawings, like components are designated by like reference numerals, and the description thereof will be omitted as appropriate.

[0014] 1 is a functional block diagram showing an overview of a search device 10 according to a first embodiment. The search device 10 includes an input receiving unit 11, a correction unit 12, and a search unit 13.

[0015] The input accepting unit 11 accepts a search query in which an attribute value and a confidence level of the attribute value are set for at least one attribute. The correcting unit 12 corrects the set search query based on the attribute value and the confidence level to generate a corrected search query. The searching unit 13 searches for registered information that matches the corrected search query from registered information registered in a database.

[0016] According to the search device 10 that performs a search using a modified search query obtained by modifying a set search query, the desired search target may be found even if a set search query contains inappropriate content or is potentially inappropriate. With this search device 10, the desired search target can be found with high accuracy.

[0017] Second Embodiment Overview The search device 10 of the second embodiment is a specific implementation of the search device 10 of the first embodiment. The search device 10 of the second embodiment includes a means for correcting a search query by an appropriate method and generating a corrected search query.

[0018] When a user sets a search query based on vague memory or information heard from another person, the attribute values ​​of the search target may be confused within a certain range, resulting in an incorrect search query. The range of confusion is limited and rarely exceeds that range. For example, a person in their 30s may be confused with someone in their 20s, 40s, or 50s, but is rarely confused with someone in their teens or younger or someone in their 60s or older. A means for correcting a search query set by a user is determined in advance based on such regularities. The search device 10 then uses the means to appropriately correct the search query set by the user and generate a corrected search query. This is described in detail below.

[0019] "Hardware Configuration" An example of the hardware configuration of the search device 10 will be described. Each functional unit of the search device 10 is realized by any combination of hardware and software. Those skilled in the art will understand that there are various variations in the realization method and device. Software includes programs that are pre-loaded when the device is shipped, and programs downloaded from recording media such as CDs (Compact Discs) or servers on the Internet.

[0020] FIG. 2 is a block diagram illustrating an example of the hardware configuration of the search device 10. As shown in FIG. 2, the search device 10 has a processor 1A, a memory 2A, an input / output interface 3A, a peripheral circuit 4A, and a bus 5A. The peripheral circuit 4A includes various modules. The search device 10 does not necessarily have to have the peripheral circuit 4A. Note that the search device 10 may be composed of multiple devices that are physically and / or logically separated. In this case, each of the multiple devices can have the above hardware configuration.

[0021] The bus 5A is a data transmission path for the processor 1A, memory 2A, peripheral circuit 4A, and input / output interface 3A to mutually transmit and receive data. The processor 1A is, for example, a processing unit such as a CPU or a graphics processing unit (GPU). The memory 2A is, for example, a random access memory (RAM) or a read-only memory (ROM). The input / output interface 3A includes interfaces for acquiring information from input devices, external devices, external servers, external sensors, cameras, etc., and interfaces for outputting information to output devices, external devices, external servers, etc. The input / output interface 3A also includes an interface for connecting to a communication network such as the Internet. Examples of input devices include a keyboard, mouse, microphone, physical buttons, touch panel, etc. Examples of output devices include a display, speaker, printer, mailer, etc. The processor 1A can issue commands to each module and perform calculations based on the results of those calculations.

[0022] "Functional Configuration" Next, the functional configuration of the search device 10 of this embodiment will be described in detail. Fig. 1 shows an example of a functional block diagram of the search device 10. As shown in the figure, the search device 10 has an input receiving unit 11, a correction unit 12, and a search unit 13.

[0023] The input receiving unit 11 receives a search query in which an attribute value and a certainty of the attribute value are set for at least one attribute. An example of the search query is shown below: (Search query) "Male: 100%, and in his 30s: 70%, and wearing red pants: 100%"

[0024] "Attribute" is an attribute of the search target. Multiple attributes that can be used to search for the search target are set in advance. A user setting a search query determines the attribute to be used in the search query within the range of the preset attributes and sets the attribute value of the determined attribute. The attribute that is set differs for each search target. The search target can be, for example, a person, a place, an object, etc.

[0025] When the search target is a person, examples of attributes include, but are not limited to, gender, age, nationality, physique, height, weight, hairstyle, clothing characteristics, possessions characteristics, posture, movement, etc. These attributes may be further subdivided. For example, clothing characteristics may be subdivided into hat characteristics, glasses characteristics, jacket characteristics, shoe characteristics, etc.

[0026] When the search target is a location, examples of attributes include, but are not limited to, the type of object present at the location (e.g., tree, building, sign, etc.) and the external characteristics of the object (e.g., color, shape, size, etc.).

[0027] When the search target is an object, examples of attributes include the type of object (e.g., desk, chair, tree, building, sign, etc.) and the external characteristics of the object (e.g., color, shape, size, etc.), but are not limited to these.

[0028] In the above example of the search query, the search target is a person, and attribute values ​​of three attributes, namely, gender, age, and clothing characteristics, are set.

[0029] The "attribute value" indicates the specific content of the attribute as described above. For example, the attribute value of "attribute: gender" is male or female. A plurality of selectable attribute values ​​may be set in advance for each attribute. Then, a user setting a search query may determine the attribute value to be used in the search query within the range of selectable attribute values ​​set in advance. Note that the method of receiving input of attribute values ​​illustrated here is merely an example and is not limited to this.

[0030] In the above example search query, the attribute value for "attribute: gender" is set to male. Also, the attribute value for "attribute: age" is set to 30s. Also, the attribute value for "attribute: clothing characteristics" is set to wearing red pants.

[0031] The "certainty" indicates the degree of certainty that each attribute value set as a search query matches the attribute value of the search target, i.e., is correct. The user sets a certainty for each attribute value set as a search query. The certainty is set, for example, between 0 and 100%, but is not limited to this. For example, the user can set a low certainty for attribute values ​​of the search target that the user heard from an unreliable other person or attribute values ​​of the search target that the user recalled from a distant memory. Furthermore, for example, the user can set a high certainty for attribute values ​​of the search target that the user heard from a reliable other person or attribute values ​​of the search target that the user recently witnessed.

[0032] In the above example search query, the numbers displayed after each attribute value indicate the confidence level of each attribute value.

[0033] There are no particular limitations on the means by which the input accepting unit 11 accepts the input of a search query. For example, the input accepting unit 11 may accept the input of a search query via an input device included in the search device 10. Alternatively, the search device 10 may be a server. The input accepting unit 11 may then acquire a search query transmitted from a client terminal.

[0034] The correction unit 12 corrects the set search query based on the attribute value and the confidence factor, and generates a corrected search query.

[0035] As described above, when a search query is set based on a user's vague memory or on information heard from another person, the attribute values ​​of the search target may be confused within a certain range, resulting in an incorrect search query. The range of confusion falls within a certain range and rarely exceeds that range. For example, a person in their 30s may be confused with someone in their 20s, 40s, or 50s, but is rarely confused with someone in their teens or younger or someone in their 60s or older. The correction unit 12 is configured to correct the search query set by the user based on such regularities.

[0036] For example, reference information such as that shown in FIG. 3 may be generated in advance and registered in the search device 10. The reference information shown in FIG. 3 indicates "other attribute values ​​that may be confused with each attribute value set in the search query" and "the possibility of confusing each attribute value with each other attribute value." One attribute value is identified by the values ​​in the "Attribute" column and the "Attribute Value" column. The "Confusing Attribute Value" column indicates other attribute values ​​that may be confused with the one attribute value identified by the values ​​in the "Attribute" and "Attribute Value" columns. Note that the numerical value associated with each attribute value shown in the "Confusing Attribute Value" column indicates the possibility of confusing that one attribute value with each other attribute value.

[0037] The reference information shown in Figure 3 indicates that when "30s" is set in the search query, the person being searched for may actually be in their 20s, 40s, or 50s. Furthermore, the reference information shown in Figure 3 indicates that the person being searched for may actually be in their 20s (20%), 40s (14%), or 50s (7%).

[0038] Such reference information can be generated by any means. For example, an administrator of the search device 10 (such as a manufacturer's representative) may ask multiple subjects to look at people of different ages and respond to how old they think the people are. The administrator of the search device 10 may then aggregate the responses to generate reference information such as that shown in FIG. 3. The administrator of the search device 10 may also conduct the experiment under different conditions. For example, the administrator may ask multiple subjects to look at people of different ages without providing them with details, and then, after a certain period of time (e.g., several days), ask them to respond to how old they think the people are. As another example of how to generate reference information, the administrator of the search device 10 may generate reference information such as that shown in FIG. 3 based on their own experience and intuition. The method of generating reference information illustrated here is merely an example and is not limited to this.

[0039] Here, a specific example of the process of correcting a search query based on the reference information shown in FIG. 3 will be described.

[0040] The correction unit 12 can generate the following corrected search query from the search query shown below, for example, based on the reference information shown in FIG. 3. In the following correction, the search query is assigned the AND conditions of 20s: 14%, 40s: 10%, and 50s: 6%. The characteristics of this correction process will be explained below. (Search query) "Male: 100%, and 30s: 70%, and wearing red pants: 100%" (Corrected search query) "Male: 100%, and 30s: 70%, and 20s: 14%, and 40s: 10%, and 50s: 6%, and wearing red pants: 100%"

[0041] - Feature 1 - As in the above correction example, when the search query includes an attribute value (first attribute value) of a certain attribute (first attribute), the correction unit 12 generates a corrected search query that includes the attribute value (first attribute value) of the attribute (first attribute) and another attribute value (second attribute value) of the attribute (first attribute).

[0042] The search query in the above modified example includes attribute values ​​for three attributes: gender, age, and clothing characteristics. These three attributes become the first attribute. The attribute values ​​for each attribute, namely, male, age in their 30s, and wearing red pants, become the first attribute value.

[0043] In the above correction example, correction is made to the 30s among the three first attribute values. The correction unit 12 generates a corrected search query including the 30s (first attribute value) and other attribute values ​​of the same attribute, namely the 20s, 40s, and 50s (second attribute values). The second attribute value is a confusing attribute value associated with the first attribute value in the reference information shown in FIG. 3. That is, the second attribute value is an attribute value indicated in the reference information shown in FIG. 3 as possibly being confused with the first attribute value.

[0044] The correction unit 12 can also perform similar corrections on other first attribute values, specifically, on male and wearing red pants. However, in the above correction example, these first attribute values ​​are not subject to correction. The reason for this will become clear in Feature 3 below.

[0045] - Feature 2 - As in the above correction example, the correction unit 12 can determine the certainty of the second attribute value in the corrected search query based on the possibility that each attribute value indicated in the reference information is confused with another attribute value.

[0046] There are various methods for calculating the certainty of the second attribute value using the possibility that each attribute value indicated in the reference information is confused with another attribute value. In the above correction example, the correction unit 12 subtracts the certainty of the first attribute value set by the user from 100% and distributes the result to the multiple second attribute values ​​in accordance with the possibility that each of the first attribute value and the second attribute value is confused.

[0047] Specifically, in the search query of the above corrected example, the confidence level of the 30s (first attribute value) is 70%. Therefore, the correction unit 12 allocates 30%, which is the value obtained by subtracting 70% from 100%, to the multiple second attribute values ​​(20s, 40s, 50s) according to the possibility of confusing the 30s with the 20s (20%; 40s; 14%; 50s; 7%).

[0048] In the above correction example, the certainty factor (70%) of the first attribute value set by the user is not corrected. As another example, the certainty factor (70%) of the first attribute value set by the user may be corrected. For example, the reference information shown in FIG. 3 indicates the possibility of confusing people in their 30s with people in their 20s (20%), the possibility of confusing people in their 30s with people in their 40s (14%), and the possibility of confusing people in their 30s with people in their 50s (7%). From this, the possibility of not confusing people in their 30s with other ages can be considered to be 59% (=100%-(20%+14%+7%)). Therefore, the correction unit 12 may use the possibility indicated in the reference information as the certainty factor and generate a corrected search query as follows: (Another example of a corrected search query) "Male: 100%, and people in their 30s: 59%, and people in their 20s: 20%, and people in their 40s: 14%, and people in their 50s: 7%, and people wearing red pants: 100%"

[0049] Feature 3: As in the above correction example, the correction unit 12 can select, as a correction target, an attribute value whose certainty level satisfies a predetermined condition. The correction unit 12 does not select, as a correction target, an attribute value whose certainty level does not satisfy the predetermined condition.

[0050] The predetermined condition is "the certainty is less than a reference value." The reference value may be 100% or a smaller value. If the certainty is relatively high, the user is confident in setting the attribute value in the search query. It is considered that there is no need to modify such an attribute value.

[0051] In the above correction example, the confidence levels for both male and wearing red pants are 100%, which are above the reference value. Therefore, these attribute values ​​are not subject to correction. Furthermore, other attribute values ​​that may be confused with these attribute values ​​are not included in the corrected search query.

[0052] As a modified example, all attribute values ​​included in the search query may be subject to correction, regardless of the confidence level.

[0053] 1 , the search unit 13 searches the registered information registered in the database for registered information that matches the corrected search query generated by the correction unit 12. The registered information is an image, but may also be text, audio, or the like.

[0054] Below, a specific description will be given of the search processing by the search unit 13. Here, the case where a person is the search target will be described as an example, but the search processing can also be performed in a similar manner when the search target is a place or other object.

[0055] First, index information such as that shown in Fig. 4 is generated by analyzing a plurality of pieces of registered information. There are no particular restrictions on the timing of generating the index information. The index information may be generated in advance of the timing of a search, or may be generated at the timing of a search.

[0056] The index information shown in FIG. 4 indicates person identification information, image identification information, attribute values ​​of each person, and their certainty levels.

[0057] "Person identification information" is information that identifies each of a plurality of people detected from the image that is the registration information.

[0058] "Image identification information" is information that identifies an image in which each person appears. The image identification information may indicate the image file name. Alternatively, the image identification information may indicate the file name of the video file and information indicating a part of the video file (time information such as the elapsed time from the beginning).

[0059] "Attribute values ​​and their certainty" indicates the certainty that each person has each attribute value. Note that to distinguish the certainty indicated in the search query and revised search query from the certainty indicated in the index information, the certainty indicated in the index information will hereinafter be referred to as the index certainty. For example, the attribute values ​​and their index certainty of each person detected from the registered information are calculated using a classifier generated by machine learning.

[0060] The search unit 13 calculates the matching degree with the revised search query for each person indicated in the index information. Then, the search unit 13 searches for images including a person whose matching degree satisfies a predetermined matching condition as images matching the revised search query. The matching condition may be "matching degree is equal to or greater than a threshold value." Alternatively, the matching condition may be "included in the top predetermined percentage or top predetermined number of people whose matching degree is among the multiple people indicated in the index information." Note that the matching conditions exemplified here are merely examples and are not limited to these.

[0061] The matching degree is calculated based on the attribute value and confidence indicated in the revised search query, and the index confidence of the attribute value indicated in the index information (the attribute value indicated in the revised search query). The higher the index confidence of the attribute value indicated in the revised search query, the higher the matching degree.

[0062] For example, the matching degree M is M=C １ ×IC １ +C ２ ×IC ２ +...C ｎ ×IC ｎ It may be calculated using the following formula:

[0063] n is the serial number of the attribute value included in the corrected search query. For example, if the corrected search query has the following content, n is the serial number of each of the six attribute values: male, 30s, 20s, 40s, 50s, and wearing red pants. (Corrected search query) "Male: 100%, and 30s: 70%, and 20s: 14%, and 40s: 10%, and 50s: 6%, and wearing red pants: 100%."

[0064] C ｎ is the confidence of the n-th attribute value included in the revised search query.

[0065] IC ｎ is the index confidence of the "n-th attribute value included in the corrected search query" of the person to be processed (target for calculating the matching degree) indicated by the index information.

[0066] The above-mentioned calculation formula is merely an example, and the present invention is not limited to this. For example, the search unit 13 may search for C ｎ ×IC ｎ The statistical values ​​of the attributes may be calculated, and the total value obtained by adding up the statistical values ​​for each attribute may be calculated as the matching degree M. The statistical values ​​may be an average value, a median value, a mode value, a maximum value, a minimum value, etc. The corrected search query includes three attributes: gender, age, and clothing characteristics. Therefore, the search unit 13 calculates C corresponding to each of the gender, age, and clothing characteristics. ｎ ×IC ｎ Then, the search unit 13 calculates the total value of the calculated three statistical values ​​as the matching degree M. In this way, the contribution rate of each of the multiple attributes to the matching degree M can be made equal.

[0067] The search unit 13 can output the search results. For example, the search unit 13 can output the matched registration information. As an example, the search unit 13 may output the matched registration information via a display or a projection device provided in the search device 10. Additionally, if the search device 10 is a server, the search unit 13 can transmit the matched registration information to a client terminal.

[0068] When there are multiple pieces of matching registered information, the search unit 13 may display the registered information in order of the degree of matching. The search unit 13 can output information displaying the matching registered information using any display method.

[0069] Next, an example of the processing flow of the search device 10 will be described with reference to the flowchart of FIG.

[0070] First, the search device 10 receives an input of a search query in which an attribute value and a certainty factor of the attribute value are set for at least one attribute (S10).

[0071] Next, the search device 10 modifies the search query set in S10 based on the attribute value and confidence level indicated by the search query, and generates a modified search query (S11).

[0072] Next, the search device 10 searches the registered information registered in the database for registered information that matches the corrected search query generated in S11 (S12).

[0073] Then, the search device 10 outputs a search result indicating registered information that matches the corrected search query (S13).

[0074] Next, an example of a process for generating a revised search query will be described with reference to the flowchart of FIG.

[0075] First, the search device 10 identifies one attribute value from the search query (S20).

[0076] Next, the search device 10 determines whether the certainty of the identified attribute value is equal to or greater than a reference value (S21).

[0077] If the certainty of the identified attribute value is not equal to or greater than the reference value (No in S21), the search device 10 determines a second attribute value corresponding to the identified attribute value and the certainty of the second attribute value (S22). The second attribute value corresponding to the identified attribute value is an attribute value that may be confused with the identified attribute value. For example, the second attribute value corresponding to the identified attribute value is an attribute value that is associated with the identified attribute value and registered as a confused attribute value in the reference information such as that shown in FIG. 3. The method for determining the certainty of the second attribute value has been described above, and therefore will not be described here.

[0078] On the other hand, if the certainty of the identified attribute value is equal to or greater than the reference value (Yes in S21), the search device 10 does not execute the process of S22.

[0079] Thereafter, if there is an attribute value that has not been specified in the search query (Yes in S23), the process returns to S20 and the same process is repeated.

[0080] When all attribute values ​​in the search query have been identified (No in S23), the search device 10 generates a revised search query based on the search query and the determination content of S22 (S24). Specifically, the search device 10 generates the revised search query by assigning the second attribute value and its certainty determined in S22 to the search query. The second attribute value and its certainty determined in S22 are assigned to the search query using, for example, an AND condition, but is not limited to this.

[0081] "Effects" According to the search device 10 of the second embodiment, the same effects as those of the search device 10 of the first embodiment are achieved.

[0082] Furthermore, when a search query is set based on a user's vague memory or on information heard from another person, the attribute values ​​of the search target may be confused within a certain range, resulting in an incorrect search query. The search device 10 of the second embodiment generates a corrected search query by adding attribute values ​​that may be confused with the attribute values ​​set in the search query to the search query. According to the search device 10 of the second embodiment, even if an incorrect search query is set due to confusion, the correct attribute values ​​will be included in the corrected search query, so there is a possibility that the desired search target will be found.

[0083] Furthermore, the search device 10 of the second embodiment can determine the certainty of other attribute values ​​in the corrected search query based on the possibility that each attribute value set in the search query is confused with each other attribute value. According to the search device 10 of the second embodiment, the certainty of other attribute values ​​in the corrected search query can be appropriately set. As a result, the accuracy of the search is improved.

[0084] Furthermore, the search device 10 of the second embodiment can target attribute values ​​whose certainty is less than a reference value for correction, and can exclude attribute values ​​whose certainty is equal to or greater than the reference value for correction. Unnecessarily correcting attribute values ​​that a user has set with confidence increases the processing load on the computer unnecessarily. By targeting attribute values ​​whose certainty is less than a reference value for correction, and excluding attribute values ​​whose certainty is equal to or greater than the reference value for correction, this inconvenience can be alleviated.

[0085] Third Embodiment A search device 10 according to a third embodiment modifies attribute values ​​set in a search query in a different manner depending on the confidence level of each attribute value set in the search query. When the confidence level is high, confusion is unlikely to have occurred, whereas when the confidence level is low, confusion is likely to have occurred. Taking this into consideration, the search device 10 according to the third embodiment modifies attribute values ​​set in a search query in a different manner depending on the confidence level of the attribute value set in the search query. This will be described in detail below.

[0086] The correcting unit 12 can generate a corrected search query from the search query based on the reference information as shown in FIG.

[0087] The reference information shown in FIG. 7 indicates "other attribute values ​​that may be confused with each attribute value" and "the possibility of confusing each attribute value with each other attribute value" when the certainty level set in the search query is within a predetermined range. One attribute value is identified by the values ​​in the "Attribute" column and the "Attribute Value" column. The "Certainty Level" column indicates the certainty level of the identified attribute value. The "Confusing Attribute Value" column indicates other attribute values ​​that may be confused with the attribute value identified by the values ​​in the "Attribute" and "Attribute Value" columns when the certainty level is within the range indicated by the value in the "Certainty Level" column. Note that the numerical value associated with each attribute value indicated in the "Confusing Attribute Value" column indicates the possibility of confusing each attribute value with each other attribute value.

[0088] 7 indicates that when the search query is set to "30s" and the confidence level is set to "90% or more but less than 100%," the person being searched for may actually be in their 20s or 40s. Furthermore, the reference information shown in FIG. 7 indicates that there is a 7% chance that the person being searched for is actually in their 20s and a 4% chance that the person being searched for is actually in their 40s.

[0089] 7 indicates that when the search query is set to "30s" and the confidence level is set to "80% or more but less than 90%," the person being searched for may actually be in their 20s, 40s, or 50s. Furthermore, the reference information shown in FIG. 7 indicates that the person being searched for may actually be in their 20s (14%), 40s (12%), or 50s (1%).

[0090] Such reference information can be generated by any means. For example, the administrator of the search device 10 (a person in charge at the manufacturer) may ask multiple subjects to look at people of different ages and respond to their guesses about how old they think the people are and their confidence level. The administrator of the search device 10 may then aggregate the responses to generate reference information such as that shown in FIG. 7 . The administrator of the search device 10 may also conduct the experiment under different conditions. For example, the administrator may ask multiple subjects to look at people of different ages without providing them with detailed information, and then, after a certain period of time (e.g., several days), ask them to respond to their guesses about how old the people are and their confidence level. As another example of how to generate reference information, the administrator of the search device 10 may generate reference information such as that shown in FIG. 7 based on their own experience and intuition. The method of generating reference information illustrated here is merely an example and is not limited to this.

[0091] Here, a specific example of a process for correcting a search query based on reference information shown in FIG. 7 will be described.

[0092] The correction unit 12 can generate the following corrected search query from the search query shown below, for example, based on the reference information shown in FIG. 7 . In the following correction, the search query is assigned AND conditions of 20s: 14%, 40s: 10%, and 50s: 6%. (Search query) "Male: 100%, and 30s: 70%, and wearing red pants: 100%" (Corrected search query) "Male: 100%, and 30s: 70%, and 20s: 14%, and 40s: 10%, and 50s: 6%, and wearing red pants: 100%"

[0093] When the correction target is people in their 30s included in the search query, the correction unit 12 generates a corrected search query based on the attribute value: 30s and confusion attribute values ​​associated with a confidence level of 70% or more and less than 80% in the reference information shown in Fig. 7. The specific method for generating the corrected search query is the same as in the second embodiment.

[0094] As described above, the correction unit 12 determines another attribute value (second attribute value) to be included in the corrected search query and its certainty based on the reference information shown in Fig. 7 and the attribute value and certainty set in the search query. The reference information shown in Fig. 7 indicates other attribute values ​​that may be confused with the attribute value set in the search query when the certainty of the attribute value satisfies a predetermined condition, and the possibility of this.

[0095] Other configurations of the search device 10 of the third embodiment are similar to those of the search device 10 of the first and second embodiments.

[0096] The search device 10 of the third embodiment achieves the same effects as the search devices 10 of the first and second embodiments. Furthermore, the search device 10 of the third embodiment can modify attribute values ​​set in a search query in a different way depending on the confidence level of each attribute value set in the search query. When the confidence level is high, confusion is unlikely to have occurred, whereas when the confidence level is low, confusion is likely to have occurred. Taking these points into consideration, the accuracy of the search can be improved by appropriately modifying each attribute value set in the search query to generate a modified search query.

[0097] <Fourth embodiment> The search device 10 of the fourth embodiment modifies attribute values ​​set in a search query in different ways depending on the attributes of a "witness who witnessed the search target and submitted an attribute value to be included in the search query."

[0098] The presence or absence of confusion and the manner of confusion may vary depending on the attributes of the witness. For example, if the witness is a child, the likelihood of confusing the search target's age, which is in their 20s to 40s, is higher than if the witness is an adult. Also, witnesses of a different nationality from the search target are more likely to confuse the search target's age than witnesses of the same nationality as the search target. Also, witnesses of a different gender from the search target are more likely to confuse the characteristics of the search target's clothing and belongings than witnesses of the same gender as the search target. Also, witnesses with different fashion preferences from the search target are more likely to confuse the characteristics of the search target's clothing and belongings than witnesses with the same fashion preferences as the search target.

[0099] Taking these points into consideration, the search device 10 of the fourth embodiment modifies the attribute values ​​set in the search query in different ways depending on the attributes of the "witness who witnessed the search target and submitted the attribute values ​​to be included in the search query." This will be described in detail below.

[0100] The correcting unit 12 can generate a corrected search query from the search query based on the reference information as shown in FIG. 8 .

[0101] 8 indicates "other attribute values ​​that may be confused with each attribute value set based on the eyewitness testimony of a witness who satisfies a predetermined witness condition" and "the possibility that each attribute value may be confused with each other attribute value." One attribute value is specified by the values ​​in the "Attribute" column and the "Attribute Value" column.

[0102] The "Witness Conditions" column shows the conditions for witnesses. The conditions for witnesses are defined using gender, age, nationality, hairstyle, clothing characteristics, possession characteristics, etc.

[0103] The "Confusing Attribute Value" column indicates other attribute values ​​that may be confused with the attribute value specified by the values ​​in the "Attribute" and "Attribute Value" columns in the case of a witness who meets predetermined witness conditions. The numerical value associated with each attribute value shown in the "Confusing Attribute Value" column indicates the possibility of confusion.

[0104] The reference information shown in Figure 8 indicates that if the witness is "under 12 years old" and the search query is set to "30s," the person being searched for may actually be in their 20s, 40s, or 50s. Furthermore, the reference information shown in Figure 8 indicates that the person being searched for may actually be in their 20s (32%), 40s (30%), or 50s (3%).

[0105] 8 indicates that if the witness is "between 18 and under 60 years old" and the search query is set to "30s," the person being searched for may actually be in their 20s, 40s, or 50s. Furthermore, the reference information shown in FIG. 8 indicates that the person being searched for may actually be in their 20s (20%), 40s (14%), or 50s (7%).

[0106] Such reference information can be generated by any means. For example, the administrator of the search device 10 (a person in charge at the manufacturer) may ask multiple subjects to look at people of different ages and respond to how old they think the people are. The administrator of the search device 10 may then aggregate the responses by eyewitness attributes to generate reference information such as that shown in FIG. 8 . The administrator of the search device 10 may also conduct the experiment under different conditions. For example, the administrator may ask multiple subjects to look at people of different ages without providing details, and then, after a certain period of time (e.g., several days), ask them to respond to how old they think the people are. As another example of how to generate reference information, the administrator of the search device 10 may generate reference information such as that shown in FIG. 8 based on their own experience and intuition. The methods of generating reference information illustrated here are merely examples, and are not limited to these.

[0107] Here, a specific example of a process for correcting a search query based on reference information shown in FIG. 8 will be described.

[0108] First, the input receiving unit 11 receives an input for setting attributes of a witness in addition to a search query. That is, the user inputs the attributes of a witness to the search device 10.

[0109] The correction unit 12 can generate the following corrected search query from the search query shown below, for example, based on the reference information shown in FIG. 8 . In the following correction, the search query is assigned AND conditions of 20s: 14%, 40s: 10%, and 50s: 6%. (Search query) "Male: 100%, and 30s: 70%, and wearing red pants: 100%" (Corrected search query) "Male: 100%, and 30s: 70%, and 20s: 14%, and 40s: 10%, and 50s: 6%, and wearing red pants: 100%"

[0110] For example, suppose the attribute of the input witness is 40 years old. If a person in their 30s included in the search query is to be corrected, the correction unit 12 generates a corrected search query based on the attribute value: 30s and the confusing attribute value associated with the witness condition: 18 years old or older and younger than 60 years old in the reference information shown in Fig. 8. The specific method for generating the corrected search query is the same as in the second embodiment.

[0111] As described above, the correction unit 12 determines another attribute value (second attribute value) to be included in the corrected search query and its certainty based on the set witness attribute, the reference information as shown in Fig. 8, and the attribute value and certainty set in the search query. The reference information shown in Fig. 8 indicates other attribute values ​​that may be confused with each attribute value when the witness attribute satisfies a predetermined witness condition, and the possibility of this being confused.

[0112] Other configurations of the search device 10 of the fourth embodiment are similar to those of the search devices 10 of the first to third embodiments.

[0113] The search device 10 of the fourth embodiment achieves the same effects as the search devices 10 of the first to third embodiments. Furthermore, the search device 10 of the fourth embodiment can modify the attribute values ​​set in the search query in different ways depending on the attributes of the "witness who witnessed the search target and submitted the attribute values ​​to be included in the search query."

[0114] The presence or absence of confusion and the manner of confusion may vary depending on the attributes of the witness. For example, if the witness is a child, the likelihood of confusing the search target's age, which is in their 20s to 40s, is higher than if the witness is an adult. Also, witnesses of a different nationality from the search target are more likely to confuse the search target's age than witnesses of the same nationality as the search target. Also, witnesses of a different gender from the search target are more likely to confuse the characteristics of the search target's clothing and belongings than witnesses of the same gender as the search target. Also, witnesses with different fashion preferences from the search target are more likely to confuse the characteristics of the search target's clothing and belongings than witnesses with the same fashion preferences as the search target.

[0115] Taking these points into consideration, the accuracy of the search can be improved by appropriately correcting each attribute value set in the search query and generating a corrected search query.

[0116] Fifth Embodiment A search device 10 according to a fifth embodiment modifies attribute values ​​set in a search query in different ways depending on the environment in which the search target was observed.

[0117] The presence or absence of confusion and the manner of confusion may vary depending on the environment at the time of sighting. For example, the appearance and color of an object may differ depending on the timing, such as night, evening, daytime, or morning. The appearance and color of an object may also differ depending on the weather, such as sunny, rainy, or cloudy. The appearance and color of an object may also differ between outdoors and indoors. Furthermore, when there are many people around, the possibility of confusing the attributes of an object may increase, but when there are few or no people around, the possibility of confusing the attributes of an object may decrease.

[0118] Taking these points into consideration, the search device 10 of the fifth embodiment modifies the attribute values ​​set in the search query in different ways depending on the environment when the search target was observed. This will be described in detail below.

[0119] The correcting unit 12 can generate a corrected search query from the search query based on the reference information as shown in FIG.

[0120] 9 indicates "other attribute values ​​that may be confused with each attribute value set based on sightings in an environment that satisfies predetermined sighting environmental conditions" and "the possibility that each attribute value may be confused with each other attribute value." One attribute value is specified by the values ​​in the "Attribute" column and the "Attribute Value" column.

[0121] The "Environmental Conditions at the Time of Sighting" column shows the environmental conditions at the time of sighting. The environmental conditions at the time of sighting are defined using the time of day, whether it was daytime or nighttime, the weather, whether it was indoors or outdoors, whether there were many people around, etc.

[0122] The "confusing attribute value" column indicates other attribute values ​​that may be confused with the attribute value specified by the values ​​in the "attribute" and "attribute value" columns when the environment satisfies the specified sighting environmental conditions. The numerical value associated with each attribute value shown in the "confusing attribute value" column indicates the possibility of confusion.

[0123] 9 indicates that when the environment at the time of sighting was "nighttime" and the search query was set to "30s," the person being searched for may actually be in their 20s, 40s, or 50s. Furthermore, the reference information shown in FIG. 9 indicates that the person being searched for may actually be in their 20s (26%), 40s (25%), or 50s (6%).

[0124] 9 indicates that when the environment at the time of sighting was "daytime" and the search query was set to "30s," the person being searched for may actually be in their 20s, 40s, or 50s. Furthermore, the reference information shown in FIG. 9 indicates that there is a 20% chance that the person being searched for is actually in their 20s, 40s, or 50s, and there is a 14% chance that the person being searched for is actually in their 20s, 40s, or 50s.

[0125] Such reference information can be generated by any means. For example, an administrator of the search device 10 (a person in charge at a manufacturer) may ask multiple subjects to look at people of different ages in a specified environment and respond to how old they think the people are. The administrator of the search device 10 may then aggregate the responses by environment to generate reference information such as that shown in FIG. 9 . The administrator of the search device 10 may also conduct the experiment under different conditions. For example, the administrator may ask multiple subjects to look at people of different ages in a specified environment without providing details, and then, after a certain period of time (e.g., several days), ask them to respond to how old they think the people are. As another example of how to generate reference information, the administrator of the search device 10 may generate reference information such as that shown in FIG. 9 based on their own experience and intuition. The methods of generating reference information illustrated here are merely examples and are not limiting.

[0126] Here, a specific example of a process for correcting a search query based on reference information shown in FIG. 9 will be described.

[0127] First, the input receiving unit 11 receives an input for setting the environment at the time of sighting in addition to the search query. That is, the user inputs the environment at the time of sighting to the search device 10.

[0128] The correction unit 12 can generate the following corrected search query from the search query shown below, for example, based on the reference information shown in FIG. 9 . In the following correction, the search query is assigned AND conditions of 20s: 14%, 40s: 10%, and 50s: 6%. (Search query) "Male: 100%, and 30s: 70%, and wearing red pants: 100%" (Corrected search query) "Male: 100%, and 30s: 70%, and 20s: 14%, and 40s: 10%, and 50s: 6%, and wearing red pants: 100%"

[0129] For example, suppose the environment at the time of the input sighting was daytime. When the search query includes people in their 30s as the target of correction, the correction unit 12 generates a corrected search query based on the confusing attribute values ​​associated with the attribute value: 30s and the environmental condition at the time of sighting: daytime in the reference information shown in Fig. 9. The specific method for generating the corrected search query is the same as that in the second embodiment.

[0130] As described above, the correction unit 12 determines other attribute values ​​(second attribute values) to be included in the corrected search query and their certainty levels based on the set environment at the time of sighting, the reference information as shown in Fig. 9, and the attribute values ​​and certainty levels set in the search query. The reference information shown in Fig. 9 indicates other attribute values ​​that may be confused with each attribute value when the environment at the time of sighting satisfies predetermined environment conditions at the time of sighting, and the possibility of this being confused.

[0131] Other configurations of the search device 10 of the fifth embodiment are similar to those of the search devices 10 of the first to fourth embodiments.

[0132] The search device 10 of the fifth embodiment achieves the same effects as the search devices 10 of the first to fourth embodiments. Furthermore, the search device 10 of the fifth embodiment can change the way attribute values ​​set in a search query depending on the environment when the search target is seen.

[0133] The presence or absence of confusion and the manner of confusion may vary depending on the environment at the time of sighting. For example, the appearance and color of an object may differ depending on the timing, such as night, evening, daytime, or morning. The appearance and color of an object may also differ depending on the weather, such as sunny, rainy, or cloudy. The appearance and color of an object may also differ between outdoors and indoors. Furthermore, when there are many people around, the possibility of confusing the attributes of an object may increase, but when there are few or no people around, the possibility of confusing the attributes of an object may decrease.

[0134] Taking these points into consideration, the accuracy of the search can be improved by appropriately correcting each attribute value set in the search query and generating a corrected search query.

[0135] Sixth Embodiment A search device 10 according to a sixth embodiment modifies an attribute value set in a search query in a different manner depending on other attribute values ​​set in the search query.

[0136] As an example, clothing characteristics set in a search query can be used to modify the age of a search target. For example, if the search query includes "wearing a miniskirt," the age of the person being searched for can be narrowed down to a certain extent based on the characteristics. It is also preferable to use different methods of narrowing down the search results depending on whether the certainty of the attribute value in the search query is set high or low.

[0137] Taking these points into consideration, the search device 10 of the sixth embodiment modifies attribute values ​​set in a search query in a different way depending on other attribute values ​​set in the search query. This will be explained in detail below. Note that, in order to distinguish between the "other attribute values" set in the search query and "other attribute values" that may be confused with each attribute value, the "other attribute values" set in the search query will be referred to as "simultaneously set attribute values" below.

[0138] The correcting unit 12 can generate a corrected search query from the search query based on the reference information as shown in FIG.

[0139] 10 indicates "other attribute values ​​that may be confused with each attribute value" and "the possibility that each attribute value may be confused with each other attribute value" when simultaneously set attribute values ​​set together with each attribute value in a search query satisfy a predetermined simultaneously set attribute value condition. One attribute value is identified by the values ​​in the "Attribute" column and the "Attribute Value" column.

[0140] The "Concurrently set attribute value conditions" column shows conditions related to concurrently set attribute values ​​set in a search query. The concurrently set attribute value conditions are defined by what attribute values ​​are included in the search query, and with what degree of certainty what attribute values ​​are included in the search query. Alternatively, the concurrently set attribute value conditions may be defined by what attribute values ​​are not included in the search query, etc.

[0141] If the "Confusing Attribute Value" column satisfies a predetermined simultaneous attribute value condition, it indicates other attribute values ​​that may be confused with the attribute value identified by the values ​​in the "Attribute" and "Attribute Value" columns. The numerical value associated with each attribute value shown in the "Confusing Attribute Value" column indicates the possibility of confusion.

[0142] 10 indicates that if the search query is set to "30s" and then "wearing a miniskirt" is set with a confidence level of 90% or higher, the person being searched for may actually be in their 20s or 40s. Furthermore, the reference information shown in FIG. 10 indicates that there is a 30% chance that the person being searched for is actually in their 20s and 40s (11%).

[0143] 10 indicates that if the search query is set to "30s" and "wearing a miniskirt" is not set with a confidence level of 90% or higher, the person being searched for may actually be in their 20s, 40s, or 50s. Furthermore, the reference information shown in FIG. 10 indicates that the person being searched for may actually be in their 20s (20%), 40s (14%), or 50s (7%).

[0144] Such reference information can be generated by any means. For example, the administrator of the search device 10 (a person in charge at the manufacturer) may ask multiple subjects to look at people of each age group and respond with how old they think those people are. The administrator of the search device 10 then aggregates the responses. The administrator of the search device 10 (a person in charge at the manufacturer) may also aggregate the trends in the attribute values ​​of multiple randomly selected subjects and calculate the trends. This allows the administrator of the search device 10 to calculate the likelihood that a person of each attribute value (e.g., 20s, 30s, 40s, etc.) has a specific attribute value (e.g., wearing a miniskirt). The administrator of the search device 10 may integrate the aggregated results using any method to generate reference information such as that shown in FIG. 10. The administrator of the search device 10 may also conduct the experiment under different conditions. For example, the administrator may ask multiple subjects to look at people of each age group without providing them with detailed information, and then, after a certain period of time (e.g., several days), ask them to respond with how old they think those people are. As another example of how to generate reference information, the administrator of the search device 10 may generate reference information as shown in Fig. 10 based on his or her own experience and intuition. Note that the methods of generating reference information illustrated here are merely examples, and are not limited to these.

[0145] Here, a specific example of a process for correcting a search query based on reference information shown in FIG. 10 will be described.

[0146] The correction unit 12 refers to the reference information shown in FIG. 10 and identifies which "simultaneously set attribute value conditions" the attribute value to be corrected satisfies. Then, the correction unit 12 generates a corrected search query based on the confusing attribute values ​​associated with the identified simultaneously set attribute value conditions. The specific method for generating the corrected search query is the same as in the second embodiment.

[0147] As described above, the correction unit 12 determines other attribute values ​​(second attribute values) to be included in the corrected search query and their certainty levels based on reference information such as that shown in Fig. 10. The reference information shown in Fig. 10 indicates other attribute values ​​that may be confused with each attribute value when the other attribute values ​​set in the search query satisfy a predetermined simultaneously set attribute value condition, and the likelihood of this being confused.

[0148] Other configurations of the search device 10 of the sixth embodiment are similar to those of the search devices 10 of the first to fifth embodiments.

[0149] The search device 10 of the sixth embodiment achieves the same effects as the search devices 10 of the first to fifth embodiments. Furthermore, the search device 10 of the sixth embodiment can modify an attribute value set in a search query in a different way depending on other attribute values ​​set in the search query.

[0150] As an example, clothing characteristics set in a search query can be used to modify the age of a search target. For example, if the search query includes "wearing a miniskirt," the age of the person being searched for can be narrowed down to a certain extent based on the characteristics. It is also preferable to use different methods of narrowing down the search results depending on whether the certainty of the attribute value in the search query is set high or low.

[0151] Taking these points into consideration, the accuracy of the search can be improved by appropriately correcting each attribute value set in the search query and generating a corrected search query.

[0152] Seventh Embodiment A search device 10 according to a seventh embodiment modifies an attribute value set in a search query in a different way depending on the amount of time that has elapsed since the search target was sighted.

[0153] The presence or absence of confusion and the type of confusion may vary depending on the time elapsed since the search target was first seen. Generally, the greater the elapsed time, the more likely confusion will occur. Also, the greater the elapsed time, the greater the extent of confusion.

[0154] Taking this into consideration, the search device 10 of the seventh embodiment modifies the attribute values ​​set in the search query in a different way depending on the amount of time that has elapsed since the search target was seen. This will be described in detail below.

[0155] The correction unit 12 corrects the search query based on reference information specifying multiple patterns of confusion attribute values ​​and confusion possibilities according to the elapsed time since the search target was sighted, and generates a corrected search query. The correction method is the same as in the above-mentioned embodiment. The user inputs the elapsed time since the search target was sighted into the search device 10.

[0156] Such reference information can be generated by any means. For example, the administrator of the search device 10 (a person in charge at the manufacturer) may ask multiple subjects to look at people of different age groups in a specified environment, and then, after a specified time has passed, ask them to respond about how old they think the people are. The elapsed time may be different for each subject. The administrator of the search device 10 may then aggregate the response results by elapsed time to generate the reference information described above. As another example of how to generate reference information, the administrator of the search device 10 may generate the reference information described above based on his or her own experience or intuition. Note that the method of generating reference information illustrated here is merely an example, and is not limited to this.

[0157] Other configurations of the search device 10 of the seventh embodiment are similar to those of the search devices 10 of the first to sixth embodiments.

[0158] The search device 10 of the seventh embodiment achieves the same effects as the search devices 10 of the first to sixth embodiments. Furthermore, the search device 10 of the seventh embodiment can modify the attribute values ​​set in the search query in different ways depending on the amount of time that has elapsed since the search target was sighted.

[0159] The presence or absence of confusion and the type of confusion may vary depending on the time elapsed since the search target was first seen. Generally, the greater the elapsed time, the more likely confusion will occur. Also, the greater the elapsed time, the greater the extent of confusion.

[0160] Taking these points into consideration, the accuracy of the search can be improved by appropriately correcting each attribute value set in the search query and generating a corrected search query.

[0161] Eighth Embodiment A search device 10 according to an eighth embodiment generates a revised search query from a search query using a method that combines two or more of the methods described in the third to seventh embodiments.

[0162] That is, the search device 10 of the eighth embodiment modifies the attribute values ​​set in the search query in a different way depending on at least two of the following: "the degree of certainty of each attribute value set in the search query," "the attributes of the witness who witnessed the search target and provided the attribute value to be included in the search query," "the environment when the search target was witnessed," "other attribute values ​​set in the search query," and "the amount of time elapsed since the search target was witnessed."

[0163] Specifically, when the reference information satisfies a predetermined condition defined using at least two of the above-described multiple items, other attribute values ​​that may be confused with each attribute and the possibility of such confusion are indicated. Then, the correction unit 12 generates a corrected search query from the search query using the reference information.

[0164] Other configurations of the search device 10 of the eighth embodiment are similar to those of the search devices 10 of the first to seventh embodiments. According to the search device 10 of the eighth embodiment, the same operational effects as those of the search devices 10 of the first to seventh embodiments are realized. Furthermore, the possibility that each attribute value is confused with another attribute value can be defined in more detail by categorizing the cases. As a result, the search query can be more appropriately corrected and a corrected search query can be generated.

[0165] Ninth Embodiment A search device 10 according to a ninth embodiment generates a corrected search query from a search query using a learning model that has learned confusion tendencies as described in the above embodiments.

[0166] That is, the corrector 12 generates a corrected search query from the search query using a learning model that has learned the confusion tendency as described in the above embodiment.

[0167] For example, in response to an input of an attribute value, the learning model outputs confusing attribute values ​​that may be confused with that attribute value and the likelihood of that confusing attribute value being confused with that attribute value.

[0168] In addition, the learning model may output confusing attribute values ​​that may be confused with the attribute value and the probability of that attribute value being confused in response to the input of the attribute value and the certainty of that attribute value. This learning model has learned the confusion tendency described in the third embodiment.

[0169] In addition, the learning model may output, in response to input of an attribute value and an attribute of a witness, a confusing attribute value that may be confused with the attribute value and the probability of that confusion. This learning model has learned the confusion tendency described in the fourth embodiment.

[0170] In addition, the learning model may output confusing attribute values ​​that may be confused with the attribute value and the likelihood of that being confused, depending on the input of the attribute value and the environment at the time of sighting. This learning model has learned the confusion tendency described in the fifth embodiment.

[0171] In addition, the learning model may output confusing attribute values ​​that may be confused with an attribute value and other attribute values ​​set in a search query, and the likelihood of that being confused. This learning model has learned the confusion tendency described in the sixth embodiment.

[0172] In addition, the learning model may output confusing attribute values ​​that may be confused with the attribute value and the probability of confusion, in response to input of the attribute value and the elapsed time since the search target was seen. This learning model has learned the confusion tendency described in the seventh embodiment.

[0173] Additionally, the learning model may output confusing attribute values ​​that may be confused with the attribute value and the likelihood of that attribute value being confused in response to input of an "attribute value" and "at least two of the following: the confidence level of the attribute value, the attribute of the witness, the environment at the time of the sighting and other attribute values ​​set in the search query, and the elapsed time since the search target was sighted." This learning model has learned the confusion tendency described in the eighth embodiment.

[0174] Other configurations of the search device 10 of the ninth embodiment are similar to those of the search devices 10 of the first to eighth embodiments. According to the search device 10 of the ninth embodiment, the same effects as those of the search devices 10 of the first to eighth embodiments are realized.

[0175] <Modifications> Modifications applicable to all of the above-described embodiments will be described below. These modifications also achieve the same effects as the above-described embodiments.

[0176] "Variation 1" (Search query) "30s: 20% and 40s: 70%" As in the above example, in a search query, multiple attribute values ​​may be set corresponding to one attribute. In the above example, two attribute values, 30s and 40s, are set corresponding to age.

[0177] In such a case, the correction unit 12 generates a corrected search query based on the multiple attribute values. Specifically, the correction unit 12 determines "attribute values ​​corresponding to the attributes" and "their certainty" to be included in the corrected search query based on the multiple attribute values. In the above example, the correction unit 12 determines the attribute value of age to be included in the corrected search query and its certainty based on the two attribute values ​​of 30s and 40s.

[0178] There are various methods for correction, but the following describes an example in which the search query in the above example is corrected based on the reference information in FIG.

[0179] Example 1 As an example, the correction unit 12 may include, in the corrected search query, all of the other attribute values ​​that may be confused with each of the multiple attribute values ​​set in the search query.

[0180] In the reference information of FIG. 3, other attributes that may be confused with 30's are shown as 20's, 40's, and 50's.

[0181] Furthermore, the reference information in FIG. 3 indicates that other attributes that may be confused with "forties" include "thirties" and "fifties."

[0182] In this example, the correction unit 12 includes in the corrected search query all of the people in their 20s, 40s, and 50s who may be confused with people in their 30s, all of the people in their 30s and 50s who may be confused with people in their 40s, and all of the people in their 30s and 40s who were originally included in the search query.

[0183] There are various ways to set the confidence level. For example, the correction unit 12 may set the confidence level of each attribute value set in the search query and a statistical value of the future possibility of each attribute value indicated in the reference information as the confidence level of each attribute value set in the corrected search query. The statistical value may be an average value, a weighted average value, a maximum value, a minimum value, a mode value, a median value, or the like.

[0184] The reference information in FIG. 3 indicates a 7% probability that someone in their 30s is confused with someone in their 50s. The reference information in FIG. 3 also indicates a 21% probability that someone in their 40s is confused with someone in their 50s. Therefore, the correction unit 12 sets the statistical values ​​of 7% and 21% as the confidence level for the 50s in the corrected search query. When adopting a weighted average, the confidence level of each attribute value set in the search query can be used as the weight. That is, the weight of 7% is the confidence level (20%) for the 30s set in the search query. And the weight of 21% is the confidence level (70%) for the 40s set in the search query.

[0185] - Example 2 - As another example, the correction unit 12 may include, in the corrected search query, other attribute values ​​that may be confused with the attribute value with the highest confidence among multiple attribute values ​​of the same attribute set in the search query.

[0186] In the search query of the above example, the attribute value with the highest certainty is 40s. In this example, the correction unit 12 includes in the corrected search query the 30s and 50s, which may be confused with 40s, as well as the 30s and 40s originally included in the search query. The method of setting the certainty is the same as in Example 1 above.

[0187] Example 3 As another example, the correction unit 12 may include, in the corrected search query, other attribute values ​​that may be confused with all of the attribute values ​​set in the search query.

[0188] In the search query of the above example, another attribute value that may be confused with all of the multiple attribute values ​​that have been set is "50s." In this example, the correction unit 12 includes "50s" and "30s" and "40s," which were originally included in the search query, in the corrected search query. The method of setting the confidence level is the same as in Example 1 above.

[0189] "Variation 2" Attributes to be modified by the modifying unit 12 and attributes not to be modified by the modifying unit 12 may be defined in advance. Then, the modifying unit 12 may modify only the attribute values ​​of the attributes defined as the attributes to be modified, and may not modify the attribute values ​​of the attributes not defined as the attributes to be modified.

[0190] For example, attributes that are unlikely to be confused, such as gender, can be excluded from correction, while attributes that are likely to be confused, such as age, can be included in correction.

[0191] "Variation 3" The search unit 13 may perform both a search (hereinafter referred to as a "first search") based on the search query (a search query set by the user) received by the input receiving unit 11 and a search (hereinafter referred to as a "second search") based on the corrected search query generated by the correction unit 12. The search device 10 may then output the results of the first search and the second search to the user, and output information inquiring of the user as to whether the search query may be corrected. The search device 10 may then correct the search query in response to a user input indicating that the search query may be corrected.

[0192] In this case, the search query accepted by the input accepting unit 11 is displayed in the search query field on the search screen. Then, with the search query displayed in the search query field, the results of the first search and the second search are displayed on the search screen. Then, in response to a user input indicating that the search query may be modified, the content of the search query field is changed from the search query accepted by the input accepting unit 11 to the modified search query generated by the modifying unit 12.

[0193] When outputting the results of the first search and the results of the second search to the user, the search device 10 may output only a predetermined number of the top results of both searches to the user. Furthermore, when outputting the results of the first search and the results of the second search to the user, the search device 10 may highlight the differences between the two results.

[0194] Although the embodiments of the present invention have been described above with reference to the drawings, these are merely examples of the present invention, and various other configurations may be adopted. The configurations of the above-described embodiments may be combined with each other, or some of the configurations may be replaced with other configurations. Furthermore, various modifications may be made to the configurations of the above-described embodiments without departing from the spirit of the invention. Furthermore, the configurations and processes disclosed in the above-described embodiments and modified examples may be combined with each other.

[0195] In addition, in the flowcharts used in the above description, multiple steps (processes) are described in order. However, the order of the steps performed in each embodiment is not limited to the order described. In each embodiment, the order of the steps shown in the drawings can be changed to the extent that the content is not affected. Furthermore, the above-described embodiments can be combined to the extent that the content is not contradictory.

[0196] Some or all of the above embodiments may be described as, but are not limited to, the following supplementary notes. 1. A search device comprising: an input accepting means for accepting a search query in which an attribute value and a certainty of the attribute value are set for at least one attribute; a correcting means for correcting the set search query based on the attribute value and the certainty and generating a corrected search query; and a search means for searching registered information registered in a database for registered information that matches the corrected search query. 2. The search device described in 1, wherein the correcting means, when the search query includes a first attribute value of a first attribute, generates the corrected search query including the first attribute value of the first attribute and a second attribute value of the first attribute, and the first attribute value and the second attribute value are different from each other. 3. The search device described in 2, wherein the correcting means generates the corrected search query including, as the second attribute value, the attribute value that is indicated to be possibly confused with the first attribute value in reference information indicating other attribute values ​​that may be confused with each of the attribute values. 4. The search device described in 3., wherein the correction means determines the second attribute value to be included in the corrected search query based on the reference information indicating other attribute values ​​that may be confused with each of the attribute values ​​when the certainty of the attribute value set in the search query satisfies a predetermined condition, and the certainty of the first attribute value set in the search query. 5. The search device described in 3 or 4, wherein the input accepting means further accepts input setting an attribute of a witness, and the correction means determines the second attribute value to be included in the corrected search query based on the reference information indicating other attribute values ​​that may be confused with each of the attribute values ​​when the attribute of the witness satisfies a predetermined condition, and the set attribute of the witness. 6. The search device described in any of 3 to 5, wherein the input accepting means further accepts input setting an environment at the time of sighting, and the correction means determines the second attribute value to be included in the corrected search query based on the reference information indicating other attribute values ​​that may be confused with each of the attribute values ​​when the environment at the time of sighting satisfies a predetermined condition, and the set environment at the time of sighting.7. The search device according to any one of 3 to 6, wherein the correction means determines the second attribute value to be included in the corrected search query based on the reference information indicating other attribute values ​​that may be confused with each of the attribute values ​​when the other attribute values ​​set in the search query satisfy a predetermined condition. 8. The search device according to any one of 3 to 7, wherein the reference information further indicates a possibility that each of the attribute values ​​may be confused with each of the other attribute values, and the correction means determines the certainty of the second attribute value in the corrected search query based on the possibility. 9. The search device according to 8, wherein the reference information indicates a possibility that each of the attribute values ​​may be confused with each of the other attribute values ​​for each of the certainty of the attribute value set in the search query, for each attribute of a witness, for each environment at the time of sighting, or for each of the other attribute values ​​set in the search query. 10. The search device according to any one of 1 to 9, wherein the correction means generates the corrected search query based on the multiple attribute values ​​when multiple attribute values ​​are set corresponding to one attribute in the search query. 11. 12. The search device according to any one of 1 to 10, wherein the correction means generates the corrected search query from the search query based on a learning model. 12. The search device according to any one of 1 to 6, wherein the correction means targets the attribute value whose confidence level is less than a reference value for correction. 13. A search method in which one or more computers receive a search query in which an attribute value and a confidence level of the attribute value are set for at least one attribute, correct the set search query based on the attribute value and the confidence level, and generate a corrected search query, and search for registered information that matches the corrected search query from registered information registered in a database. 14. A program that causes a computer to function as: input accepting means that accepts a search query in which an attribute value and a confidence level of the attribute value are set for at least one attribute; correction means that corrects the set search query based on the attribute value and the confidence level, and generates a corrected search query; and search means that search for registered information that matches the corrected search query from registered information registered in a database.

[0197] This application claims priority based on Japanese Patent Application No. 2022-204368, filed December 21, 2022, the disclosure of which is incorporated herein in its entirety by reference.

[0198] 10 Search device 11 Input reception unit 12 Correction unit 13 Search unit 1A Processor 2A Memory 3A Input / output I / F 4A Peripheral circuit 5A Bus

Claims

1. an input receiving means for receiving a search query in which an attribute value and a certainty of the attribute value are set for at least one attribute; a correction means for correcting the set search query based on the attribute value and the confidence factor to generate a corrected search query; a search means for searching for registered information that matches the corrected search query from registered information registered in a database; A search device having the above configuration.

2. The correction means is If the search query includes a first attribute value of a first attribute, generating the modified search query including the first attribute value of the first attribute and a second attribute value of the first attribute; The search device according to claim 1 , wherein the first attribute value and the second attribute value are different from each other.

3. The correction means is The search device according to claim 2, wherein the modified search query is generated by including, as the second attribute value, an attribute value that is indicated as possibly being confused with the first attribute value in reference information indicating other attribute values that may be confused with each of the attribute values.

4. The correction means is The search device according to claim 3 , wherein the second attribute value to be included in the corrected search query is determined based on the reference information indicating other attribute values that may be confused with the attribute value set in the search query when the certainty of the attribute value set in the search query satisfies a predetermined condition, and the certainty of the first attribute value set in the search query.

5. the input receiving means further receives an input for setting attributes of a witness; The correction means is The search device described in claim 3 determines the second attribute value to be included in the revised search query based on the reference information indicating other attribute values that may be confused with each of the attribute values when the witness attribute satisfies a predetermined condition, and the set witness attribute.

6. the input receiving means further receives an input for setting an environment at the time of sighting; The correction means is The search device described in claim 3 determines the second attribute value to be included in the revised search query based on the reference information indicating other attribute values that may be confused with each of the attribute values when the environment at the time of sighting satisfies specified conditions, and the set environment at the time of sighting.

7. The correction means is The search device described in claim 3 determines the second attribute value to be included in the corrected search query based on the reference information indicating other attribute values that may be confused with each of the attribute values when the other attribute values set in the search query satisfy a predetermined condition.

8. the reference information further indicates a likelihood that each of the attribute values is confused with each of the other attribute values; The search device according to claim 3 , wherein the correction means determines the degree of certainty of the second attribute value in the corrected search query based on the possibility.

9. One or more computers receiving a search query in which an attribute value and a confidence level of the attribute value are set for at least one attribute; modifying the set search query based on the attribute value and the confidence factor to generate a modified search query; A search method for searching for registered information that matches the modified search query from registered information registered in a database.

10. Computer, an input receiving means for receiving a search query in which an attribute value and a certainty factor of the attribute value are set for at least one attribute; a correction means for correcting the set search query based on the attribute value and the confidence factor to generate a corrected search query; a search means for searching for registered information that matches the corrected search query from registered information registered in a database; A program that functions as a