JP7649205B2 - Human relationship extraction device and human relationship extraction method - Google Patents

Description

The present invention relates to a human relationship extraction device and a human relationship extraction method, and is suitable for use in a human relationship extraction device and a human relationship extraction method that determine whether a subject is a high-risk individual in various screening procedures.

Traditionally, in various situations such as opening an account or registering a residence certificate, a person is screened to determine whether or not he or she is a high-risk person. The above-mentioned "risk" can be, for example, the possibility that the person will be involved in a crime in the near future. The screening agency that performs the above screening generally maintains a blacklist that lists known high-risk persons (e.g., criminals and their associates), and it is known that high-risk persons are often connected in some way to those on the blacklist. Therefore, during the screening, whether or not the person is a high-risk person is determined by determining the relationship between the person and those on the blacklist.

For example, Patent Document 1 discloses a technique for resolving entities based on their relationship to a common entity. Patent Document 1 discloses a method of "comparing two entities and resolving them to a single entity if the two entities satisfy an entity resolution threshold based on a set of one or more commonly related entities." When using Patent Document 1 to determine whether or not there is a relationship between a high-risk individual and a blacklisted individual, by treating the target and the blacklisted individual as two entities, it is possible to determine that there is a relationship between the target and the blacklisted individual if the two entities have attributes in common, such as a telephone number or address.

Special Publication No. 2014-529129

As mentioned above, the technique in Patent Document 1 makes it possible to extract a relationship between a target person and a blacklisted person when the values of basic information linked to the target person, such as a telephone number or address, match those of the target person and a blacklisted person. However, there are not many cases where the basic information matches perfectly, and it must be said that the risk of overlooking high-risk individuals who have a relationship with a blacklisted person remains high.

Furthermore, the relationship between a high-risk person and a blacklisted person cannot be sufficiently determined based on only basic information linked to an individual; it is conceivable that a relationship may exist based on information surrounding the individual (for example, the organization to which the person belongs, such as the place of employment). However, with conventional technology, it was not possible to extract the relationship between a subject and a blacklisted person by taking into account such information surrounding the subject. Specifically, for example, if a blacklisted person also works at the subject's workplace, it was possible to extract the relationship using conventional technology, but if a blacklisted person works at a different workplace similar to the subject's workplace, the conventional technology was not able to extract the relationship.

The present invention has been made in consideration of the above points, and aims to propose a human relationship extraction device and method that can determine whether a subject is a high-risk individual by taking into account not only basic information linked to an individual, but also the relationship between the individual and an organization, and the relationship between organizations.

In order to solve such problems, the present invention provides a human relationship extraction device that uses a computer to determine whether a person to be screened is a high-risk person having a high level of risk, the human relationship extraction device comprising a processor that executes a predetermined process by executing a program, and a storage device that stores the program and data, the storage device stores at least information indicating a factual relationship between people, information indicating a factual relationship between organizations, information indicating a factual relationship between organizations, and information indicating a person on a blacklist, as known factual information input from outside, the processor, when the person is designated, acquires factual information linked to the person of the subject and registers it in the storage device, further registers organizational relationship analogy information indicating an analogical relationship between organizations inferred from the factual information in the storage device, executes a first related person extraction to extract a first related person of the subject based on the various factual information stored in the storage device, and performs the first related person extraction to extract the first related person of the subject based on the various factual information stored in the storage device, and performs the first related person extraction to extract the first related person of the subject based on the various factual information stored in the storage device. a first determination as to whether the subject is a high-risk person or not based on whether the first related persons extracted in the first related persons extraction include a person who belongs to the blacklist; a second determination as to whether the subject is a high-risk person or not based on whether the first related persons extracted in the second related persons extraction include a person who belongs to the blacklist; a relevance score indicating the degree to which the second related persons extracted in the second related persons extraction include a person who belongs to the blacklist; a second determination as to whether the subject is a high-risk person or not based on the calculated relevance score ;

In addition, in order to solve such problems, the present invention provides a human relationship extraction method by a human relationship extraction device that judges whether or not a person to be screened is a high-risk person having a high level of risk, the computer constituting the human relationship extraction device has a processor that executes predetermined processing by executing a program, and a storage device that stores the program and data, the storage device stores at least information indicating a factual relationship between people, information indicating a factual relationship between organizations and people, information indicating a factual relationship between organizations and organizations, and information indicating a person on a blacklist, as known factual information input from the outside, the processor includes a fact information registration step in which, when the subject is designated, the processor acquires fact information linked to the subject and registers it in the storage device, an inference information registration step in which the processor further registers organizational relationship inference information indicating an inference relationship between organizations inferred from the fact information, a first related person extraction step in which the processor extracts a first related person of the subject based on various factual information stored in the storage device, and the processor extracts a first related person of the subject based on various factual information stored in the storage device in addition to the first related person extracted in the first related person extraction step. a second related person extraction step of extracting, as a second related person of the target, a person who has a history of affiliation to an organization that has the analogous relationship with an organization to which the target has a history of affiliation, based on the organizational relationship analogy information registered in the analogy information registration step; a first judgment step in which the processor judges whether the target is a high-risk person or not, based on whether the first related people extracted in the first related person extraction step include a person on the blacklist; a second judgment step in which the processor calculates a relevance score indicating the degree to which the second related people extracted in the second related person extraction step are included in the person on the blacklist, and judges whether the target is a high-risk person or not, based on the calculated relevance score; a final judgment step in which the processor finally judges the target to be a high-risk person when a judgment result that the target is a high-risk person is obtained in at least one of the first and second judgment steps; and an output step in which the processor outputs the judgment result in the final judgment step and the judgment results in the first and second judgment steps .

According to the present invention, it is possible to determine whether a subject is at high risk by taking into consideration not only basic information linked to the individual, but also the relationship between the individual and the organization, and the relationship between the organizations.

1 is a block diagram showing an example of the configuration of a human relationship extraction device 100 according to an embodiment of the present invention. FIG. 2 is a block diagram showing an example of a hardware configuration of a computer 10 that realizes the human relationship extraction device 100. FIG. 2 is a diagram illustrating an example of a human relationship table 111. FIG. 13 is a diagram showing an example of an organization-person relationship table 112. FIG. 13 is a diagram showing an example of an organizational relationship fact table 113. FIG. 2 is a diagram showing an example of a person basic information table 114. FIG. 13 is a diagram showing an example of a BL applicable person table 115. FIG. 13 is a diagram showing an example of a subject relationship information table 116. FIG. 11 is a diagram showing an example of an individual similarity table 117. FIG. 13 is a diagram showing an example of an organizational relationship inference table 118. FIG. 11 is a diagram showing an example of an extraction result table 119. 13 is a flowchart showing an example of a processing procedure for human relationship extraction processing. 1 is a schematic diagram showing an example of a connection relationship between a subject A and the surrounding environment. FIG. 1 is a schematic diagram showing an example of a connection relationship between subject A and an extended surrounding environment. FIG. FIG. 13 is a diagram showing an example of a determination result screen.

One embodiment of the present invention will be described below with reference to the drawings. Note that the embodiment described below does not limit the invention according to the claims, and not all of the combinations of features described in the embodiment are necessarily essential to the solution of the invention. In the following description, various information may be described using expressions such as "table", "list", and "queue", but the various information may be expressed using data structures other than these. To indicate that the information is not dependent on the data structure, "XX table", "XX list", etc. may be referred to as "XX information". When describing the content of each piece of information, expressions such as "identification information", "identifier", "name", "ID", and "number" are used, but these are interchangeable.

(1) Device Configuration Fig. 1 is a block diagram showing an example of the configuration of a human relationship extraction device 100 according to one embodiment of the present invention. As shown in Fig. 1, the human relationship extraction device 100 includes an input relationship registration unit 101, an organizational relationship analogy unit 102, a related person extraction unit 103, and a high-risk person determination unit 104 as processing units providing predetermined functions, and includes a human relationship table 111, an organization-human relationship table 112, an organizational relationship fact table 113, a human basic information table 114, a BL corresponding person table 115, a target person relationship information table 116, an individual similarity table 117, an organizational relationship analogy table 118, and an extraction result table 119 as various data for holding predetermined information.

The human relationship extraction device 100 is a device that determines whether a subject who is the subject of screening is a high-risk person who poses a high level of risk by executing a human relationship extraction process that extracts relationships between the subject and blacklisted persons (BL persons) registered on a blacklist (BL). Details will be described later with reference to FIG. 12 and other figures, but the human relationship extraction process extracts related persons of the subject from the surrounding environment based on relationships between people, between organizations, and between organizations shown from factual information, and relationships between organizations inferred from factual information, and determines whether the subject is a high-risk person based on whether the related persons are BL persons or other criteria, and visualizes the determination result to detect high-risk persons with a high degree of accuracy.

The input relationship registration unit 101 has a function of accepting input of various factual information, registering the input factual information in various data (person relationship table 111, organization-person relationship table 112, organization relationship fact table 113, person basic information table 114, BL applicable person table 115), and storing it in a storage device (memory 12 or storage device 13 shown in FIG. 2). The organizational relationship analogy unit 102 has a function of analogizing the relationship between organizations and registering the analogy result in the organizational relationship analogy table 118. The related person extraction unit 103 has a function of registering basic information linked to the subject to be judged in the subject relationship information table 116, and extracting people (related persons) who have a relationship with the subject based on the various registered data. The related person extraction result by the related person extraction unit 103 is registered in the extraction result table 119. The high-risk person determination unit 104 has a function of determining whether the subject is a high-risk person or not based on the related person extraction result by the related person extraction unit 103, and outputting the determination result to the user.

Details of the various data held by the human relationship extraction device 100 will be described later with reference to Figures 3 to 11. The human relationship table 111, organization-human relationship table 112, organization relationship fact table 113, person basic information table 114, and BL applicable person table 115 are data about factual information already known about people and organizations that is input and registered in the human relationship extraction device 100 before a determination is made as to whether or not the subject is a high-risk person. The subject relationship information table 116 is data about factual information about the subject of human relationship extraction that is input and registered in the human relationship extraction device 100 when the subject is specified. The individual similarity table 117, organization relationship analogy table 118, and extraction result table 119 are data that are generated and registered during the human relationship extraction process.

Figure 2 is a block diagram showing an example of the hardware configuration of a computer 10 that realizes the human relationship extraction device 100.

The computer 10 is configured with a CPU 11, a memory 12, a storage device 13, an input device 14, an output device 15, and a network device 16, all connected to each other via a bus 17.

The CPU 11 is a processor that executes a program, such as a CPU (Central Processing Unit). The memory 12 is a main storage device, such as a RAM (Random Access Memory), that stores programs and data referenced when the programs are executed. The storage device 13 is an auxiliary storage device that stores programs and data referenced when the programs are executed, such as a HDD (Hard Disk Drive) or SSD (Solid State Drive). The input device 14 is a device that accepts input from an operator, such as a keyboard, mouse, or touch panel. The output device 15 is a device that outputs the results of processing executed by the CPU 11 or data held in the memory 12 or the storage device 13, such as a liquid crystal display device or an organic EL (Electro Luminescence) display. The network device 16 is various interfaces such as a communication interface and an input/output interface. The bus 17 is an internal communication line that connects the devices in the computer 10.

Each processing unit of the human relationship extraction device 100 shown in FIG. 1 is realized by executing a processing program corresponding to the function of each processing unit by the hardware resources of the computer 10 described above. Specifically, the input relationship registration unit 101, organizational relationship analogy unit 102, related party extraction unit 103, and high-risk individual determination unit 104 shown in FIG. 1 are realized by the CPU 11 executing a predetermined program stored in the memory 12 or storage device 13 while referring to related data. The high-risk individual determination unit 104 is realized by the CPU 11 executing a predetermined program and outputting the execution result (determination result screen 120 in FIG. 15 described later) to the output device 15 (or a user-side terminal).

In addition, the various data of the human relationship extraction device 100 shown in FIG. 1 (human relationship table 111, organization-human relationship table 112, organizational relationship fact table 113, human basic information table 114, BL matched person table 115, target person relationship information table 116, personal similarity table 117, organizational relationship analogy table 118, and extraction result table 119) are realized by the memory 12 or the storage device 13.

The above programs executed by the CPU 11 in the computer 10 and some or all of the data referenced by the programs may be stored in advance in the memory 12 or storage device 13, or may be configured to be stored in the storage device 13 of the computer 10 from a non-transient storage device or non-transient storage medium of another device connected to the network via the network device 16, as necessary. In addition, in the following description, for simplicity, the human relationship extraction device 100 is configured to be realized by one computer 10, but the configuration of the human relationship extraction device 100 according to this embodiment is not limited to this, and may be realized by multiple computers 10, or may be realized including a database device, cloud storage, etc.

(2) Data Structure Fig. 3 is a diagram showing an example of the person relationship table 111. The person relationship table 111 is a table that holds information indicating the factual relationship between people, and has columns of a person 1111, a related person 1112, and a relationship 1113. In the person relationship table 111, a relationship 1113 between the related person 1112 and each person 1111 is recorded. In this embodiment, regarding the factual relationship between people, for example, when person A and person B are in a family relationship or when person A and person B have the same address, a relationship is considered to exist, and in each of the above cases, "family" or "living together" is written in the relationship 1113.

FIG. 4 is a diagram showing an example of the organization-person relationship table 112. The organization-person relationship table 112 is a table that holds information indicating the factual relationship between an organization and a person, and has columns of organization 1121, person 1122, relationship 1123, relationship start 1124, and relationship end 1125. In the organization-person relationship table 112, for each organization 1121, for a person 1122 related to (or related to) the organization, the relationship 1123, and further the start time (relationship start 1124) and end time (relationship end 1125) of the relationship are recorded. In this embodiment, for example, when person A belongs to organization P for a certain period of time, the relationship is considered to exist, and in that case, "affiliation" is written in the relationship 1123. Specifically, for example, the record in the first row of FIG. 4 indicates that person C belonged to organization P from April 2017 to March 2019. The column value of the relationship 1123 can be set as appropriate to something other than "affiliation." For example, if the person was directly employed, "affiliation" can be entered, and if the person worked under a dispatch contract or contracted work, "dispatch" or "contractor" can be entered. The relationship 1123 can also be entered by distinguishing the position (job title, etc.) in the organization 1121.

Figure 5 is a diagram showing an example of the organizational relationship fact table 113. The organizational relationship fact table 113 is a table that holds information showing the factual relationship between organizations, and has columns for the relevant organization 1131, the related organization 1132, and the relationship 1133. In the organizational relationship fact table 113, the related organization 1132 and the relationship 1133 with the related organization are recorded for each relevant organization 1131. In this embodiment, regarding the factual relationship between organizations, for example, when organizations P and Q are group companies, or when organizations P and Q have the same representative, a relationship is considered to exist, and in the above cases, "group company" or "same representative" is written in the relationship 1133, respectively.

Figure 6 is a diagram showing an example of a person basic information table 114. The person basic information table 114 is a table that holds basic information linked to an individual, and has columns for name 1141, nationality 1142, occupation 1143, address 1144, age 1145, and place of employment 1146.

Figure 7 shows an example of the BL recipient table 115. The BL recipient table 115 is a table that holds the registration information of the blacklist, and lists the names of BL recipients.

8 is a diagram showing an example of the subject relationship information table 116. The subject relationship information table 116 is a table that holds basic information linked to the subject as factual information related to the subject, and has columns of related party 1161, category 1162, relationship 1163, relationship start 1164, and relationship end 1165. The registered contents of the subject relationship information table 116 are input together with the designation of the subject in the human relationship extraction process. Related party 1161 describes a subject that is directly related to the subject, and includes not only "person" but also "organization" (the organization to which the subject belongs). Category 1162 describes the type of related party 1161, which is "person" or "organization". Relationship 1163 describes the relationship between the subject and related party 1161. Then, relationship start 1164 and relationship end 1165 describe the period during which the subject belonged to the organization when an organization is described in related party 1161 (in other words, category 1162 is "organization").

Figure 9 is a diagram showing an example of the individual similarity table 117. The individual similarity table 117 is a table that holds individual similarities that indicate inferred relationships between people. Details will be described later, but the individual similarities are calculated by the organizational relationship inference unit 102 based on information belonging to each individual registered in the person basic information table 114. In this example, the individual similarity is expressed as a numerical value in the range of "0.0" to "1.0", and the closer the value is to 1.0, the higher the similarity (there is inferred to be a relationship).

Figure 10 is a diagram showing an example of the organizational relationship analogy table 118. The organizational relationship analogy table 118 is a table that holds information indicating an analogical relationship between organizations, and has columns for the relevant organization 1181, related organizations 1182, and organizational similarity 1183. The relevant organization 1181 and related organizations 1182 are organizations that have an analogical relationship, and the organizational similarity 1183 is the similarity (organizational similarity) of the relevant organization 1181 to the related organization 1182. As will be described in detail later, the organizational similarity is calculated by the organizational relationship analogy unit 102 based on the personal similarity of people belonging to each organization, and a combination of organizations that has a predetermined organizational similarity or higher is recorded in the organizational relationship analogy table 118. In this example, the organizational similarity is expressed as a numerical value in the range of "0.0" to "1.0", and the closer the value is to 1.0, the higher the similarity (it is inferred that there is a relationship). In this embodiment, an analogical relationship between organizations is inferred when, for example, the attribute information of person A who belongs to organization P and person B who belongs to organization Q are similar, or when the financial information, corporate size, or products handled by organizations P and Q are similar.

FIG. 11 is a diagram showing an example of the extraction result table 119. The extraction result table 119 is a table that holds information on the extraction result of related persons of the subject, and is registered by the related person extraction unit 103. The extraction result table 119 has columns of related persons 1191, BL applicability 1192, factual relationship 1193, analogical relationship 1194, and related organization 1195. Related persons 1191 records people extracted as related persons of the subject. BL applicability 1192 records whether the related person is a BL related person, factual relationship 1193 records whether the related person has a direct relationship with the subject based on factual information, and analogical relationship 1194 records whether the related person has an analogical relationship with the subject based on analogical information. In addition, related organization 1195 records organizations to which the subject or related person belongs (or belonged).

(3) Human Relationship Extraction Process The human relationship extraction process executed by the human relationship extraction device 100 will be described in detail below with appropriate reference to the specific data examples in Figures 3 to 11. The human relationship extraction process is a process in which, when a target person is specified, the human relationship extraction device 100 extracts related persons who have a strong relationship with the target person and compares them with blacklisted persons to extract relationships between the target person and blacklisted persons, and determines whether the target person is a high-risk person.

Figure 12 is a flowchart showing an example of the processing procedure for human relationship extraction processing. In Figure 12, the processing of steps S1 and S2 is executed at any timing before the target person is specified. On the other hand, the processing of steps S3 to S5 is executed after the target person is specified and the target person's relationship information (information to be registered in the target person relationship information table 116) is input. In this explanation, it is assumed that "target person A" is specified.

(3-1) Registration of Input Relationships First, in step S1, the input relationship registration unit 101 registers information indicating various pre-prepared factual relationships in the person relationship table 111, the organization-person relationship table 112, the organization relationship fact table 113, the person basic information table 114, and the BL relevant person table 115.

(3-2) Inference of Organizational Relations Next, in step S2, the organizational relation inference unit 102 infers relationships between organizations. In step S2, the organizational relation inference unit 102 calculates the similarity between organizations (organization similarity) based on the similarity of people belonging to each organization (individual similarity), infers that an organizational relation exists when the organizational similarity exceeds a predetermined threshold, and registers the inference result in the organizational relation inference table 118.

Here, we explain the process of step S2 in detail.

First, the organizational relationship analogy unit 102 extracts attribute information belonging to individuals (e.g., nationality 1142, occupation 1143, address 1144, age 1145) from the person basic information table 114, and calculates the similarity between individuals (personal similarity) using a predetermined calculation method based on the extracted attribute information. The calculation method of the personal similarity is not limited to a specific method, but for example, the percentage (degree) of matching of the extracted attribute information between the target individuals is taken as the personal similarity. Specifically, in the case of the combination of X1 and X3 shown in the person basic information table 114 of FIG. 6, two items of the attribute information of nationality, occupation, address, and age match, nationality "Country A" and address "Tokyo", so this personal similarity is 0.5 (2/4). The organizational relationship analogy unit 102 performs such calculation of personal similarity for all combinations of people registered in the person basic information table 114. The individual similarity table 117 shown in FIG. 9 is a data table that records the calculation results of these individual similarities.

Next, the organizational relationship analogy unit 102 extracts pairs having a personal similarity of a predetermined threshold value (e.g., 0.8) or more from among combinations of people belonging to different organizations for each organization listed in the workplace 1146 of the person basic information table 114. Specifically, in the case of the person basic information table 114 of FIG. 6 and the personal similarity table 117 of FIG. 9, one pair of X1 belonging to P company and X4 belonging to R company is extracted. The personal similarity of the pair is 1.0. Next, the organizational relationship analogy unit 102 calculates the ratio (degree) of people extracted as a pair to the total number of people in each organization as the organizational similarity of the organization. In the above specific example, according to the person basic information table 114, there are two people, X1 and X2, who belong to P company, and the only person from P company extracted as a pair is X1, so the organizational similarity of P company to R company is 0.5 (1/2). The organizational similarity based on Company R (i.e., the organizational similarity of Company R to Company P) is 0.33 (1/3) because there are three people (X3, X4, X5) who belong to Company R. In this way, by calculating the organizational similarity of each organization to the other organization, it is possible to calculate similarity that takes the size of the organizations into account.

The organizational relationship inference unit 102 then considers that an inter-organizational relationship can be inferred for those organizational similarities calculated for each organization that are equal to or greater than a predetermined threshold (e.g., 0.3), and registers these in the organizational relationship inference table 118. In the above specific example, the organizational similarity of Company P to Company R of "0.5" and the organizational similarity of Company R to Company P of "0.33" are both equal to or greater than the predetermined threshold "0.3", and are therefore registered in the organizational relationship inference table 118 as shown in FIG. 10.

In this manner, in step S2, the organizational relationship inference unit 102 infers the relationships between organizations, and the inference results are registered in the organizational relationship inference table 118.

(3-3) Extraction of related parties based on factual information Next, in step S3, the related party extraction unit 103 extracts people who have a relationship (direct relationship) with the target person A specified by the user based on factual information as related parties of the target person A (first related party extraction).

Here, the process of step S3 will be described in detail. In step S3, the stakeholder extraction unit 103 executes the following first to fourth stages of processing in order.

Figure 13 is a schematic diagram showing an example of a connection relationship between subject A and the surrounding environment. Figure 13 shows an example of a direct relationship that becomes clear from factual information about the surrounding environment of subject A, with the notation "family" or "affiliation" indicating the relationship between the subjects. Furthermore, the number written below the notation "affiliation" indicates the period of said affiliation. Below, the first related party extraction will be explained using the relationship shown in Figure 13 as a concrete example.

In the first stage, the related person extraction unit 103 registers the factual information about the subject A that was entered when the subject A was specified in the subject relationship information table 116.

In the second stage, the related person extraction unit 103 extracts "people" related to the subject A based on the subject relationship information table 116 and the various relationship tables registered in step S1 (person relationship table 111, organization-person relationship table 112, organization relationship fact table 113). All information in each table referenced in this second stage is factual information that does not include analogies.

The second stage processing will be explained in detail using the tables in Figures 3, 4, 5, and 8.

First, the related party extraction unit 103 refers to the target relationship information table 116 (Figure 8) and confirms that the related party 1161 in the first record is "B" and the category 1162 is "person". Then, using "B" as a key, it searches the person relationship table 111 (Figure 3) and the organization-person relationship table 112 (Figure 4) and attempts to extract a target (which may be a person or an organization) related to "B". However, since there is no information about "B" in the person relationship table 111 in Figure 3 and the organization-person relationship table 112 in Figure 4, no target related to "B" is extracted.

Next, the related party extraction unit 103 refers to the second record in the target relationship information table 116, and confirms that the related party 1161 is "P" and the category 1162 is "organization." Then, using "P" as a key, it searches the organization-person relationship table 112 (Figure 4) and the organization relationship fact table 113 (Figure 5) and attempts to extract targets (which may be people or organizations) related to "P." In this case, Q, C, and D are extracted as targets related to "P." Furthermore, if any targets are extracted, the related party extraction unit 103 repeats the same search using each of the extracted targets as a key.

Then, the related person extraction unit 103 repeatedly executes the above search to select "people" from the multiple targets extracted, and sets these (B, C, D, and E in this example) as candidates for related persons of target A.

In the third stage, the related person extraction unit 103 checks the period of affiliation to the organization of each of the candidates related to the subject A extracted in the second stage, and excludes candidates whose period of affiliation does not overlap with the period of affiliation of the subject A from the related persons of the subject A. Specifically, according to the subject relationship information table 116 in FIG. 8, the period during which the subject A belonged to the organization P was from April 2018 to March 2019. On the other hand, according to the organization-person relationship table 112 in FIG. 4, the period during which D belonged to the organization P was from April 2013 to March 2016, and the affiliation periods do not overlap. Note that C and E are not excluded because there is no factual information regarding their affiliation to the organization. Therefore, by the processing in the third stage, D is excluded, and B, C, and E are regarded as related persons of the subject A. Note that the candidates to be excluded in the third stage may include not only candidates whose period of affiliation to the organization does not overlap with the subject, but also candidates subordinate to the candidate (for example, family members).

In the fourth stage, the related person extraction unit 103 registers the related persons of subject A extracted at the end of the third stage in the extraction result table 119.

In detail, the related party extraction unit 103 creates a new record for each related party in the extraction result table 119, registers the related party in related party 1191, and registers "○" in factual relationship 1193. Furthermore, in the fourth stage, the related party extraction unit 103 compares the related parties of the target person A with the BL relevant party table 115 (FIG. 7) to confirm whether or not there is a BL relevant party, and registers the confirmation result in BL relevant party 1192 of the relevant record in the extraction result table 119. Specifically, in the fourth stage of step S3, four people, B, C, E, and F, are registered in the extraction result table 119 as related parties of the target person A. Of these, in the record of C, who is a BL relevant party, "○" is registered in the column value of BL relevant party 1192, and in the records of B, D, and E, who are not BL relevant parties, "-" is registered in the column value of BL relevant party 1192. Additionally, the organization to which the subject or related person belongs (or belonged) is registered in the related organization 1195 of each record.

As described above, in step S3, the first related party extraction can extract related parties who have a direct relationship with subject A based on factual information, and the extraction results can be registered in the extraction result table 119.

(3-4) Extraction of related parties based on analogical information Next, in step S4, the related party extraction unit 103 extracts, in addition to people who have a relationship based on analogical information (direct relationship) extracted in step S3, people who have a relationship based on analogical information (analogous relationship) as related parties of the subject A (second related party extraction).

Here, the process of step S4 will be described in detail. In step S4, the stakeholder extraction unit 103 executes the following first to fourth stages of processing in order.

Figure 14 is a schematic diagram showing an example of the connection relationship between subject A and the expanded surrounding environment. In addition to the direct relationship shown in Figure 13, Figure 14 shows an example of an analogical relationship that becomes clear from analogical information regarding the surrounding environment of subject A. As also shown in the legend, in Figure 14, direct relationships based on factual information are represented by solid lines, and analogical relationships based on analogical information are represented by dashed lines. The other notation methods in Figure 14 are the same as those in Figure 13. Below, the second related party extraction will be explained using the relationships shown in Figure 14 as a specific example.

In the first stage, the related party extraction unit 103 focuses on the organizations extracted in the second stage of step S3 (first related party extraction) and extracts organizations that have an analogical relationship with the organization from the organizational relationship inference table 118. Specifically, since organization P was extracted in the second stage of the first related party extraction, the related party extraction unit 103 extracts organizations R and S as organizations that have an analogical relationship with organization P by referring to related organizations 1182 of records in which "P" is entered in the corresponding organization 1181 in the organizational relationship inference table 118 of FIG. 10. Note that in the first stage, the related party extraction unit 103 may obtain related parties based on direct relationships extracted in step S3 (first related party extraction) as related parties of the extracted target person A.

In the second stage, the related party extraction unit 103 extracts people who have a direct relationship with the organizations extracted in the first stage, as candidates for related parties who have an analogical relationship with the target person A, based on the organization-person relationship table 112 and the person relationship table 111. Specifically, from the organization-person relationship table 112 in FIG. 4, R1 and R2 are extracted as people who have a history of belonging to organization R, and S1 and S2 are extracted as people who have a history of belonging to organization S. Note that, since people related to R1, R2, S1, and S2 are not registered in the person relationship table 111 in FIG. 3, no other people are extracted.

In the third stage, the related party extraction unit 103 checks the period of affiliation to the organization of each of the candidates for related parties of the subject A based on the analogical relationship extracted in the second stage, and excludes candidates whose period of affiliation does not overlap with the period of affiliation to the organization of the subject A from the related parties based on the analogical relationship of the subject A. Specifically, the affiliation periods of R1, R2, S1, and S2 to the organization R or organization S can be read from the organization-person relationship table 112 in FIG. 4, and the related party extraction unit 103 compares these with the period of affiliation of the subject A to the organization P, which is "April 2018 to March 2019." As a result, since the period during which R1 and S1 belonged to the organization does not overlap with the period during which the subject A belonged to the organization P, R1 and S1 are excluded from the related parties based on the analogical relationship of the subject A. In other words, by the processing of the third stage, R2 and S2 are regarded as related parties based on the analogical relationship of the subject A. In addition, the candidates to be excluded in the third stage may include not only candidates whose period of affiliation to the organization does not overlap with that of the target candidate, but also candidates who are subordinate to the target candidate (e.g., family members).

In the fourth stage, the related person extraction unit 103 registers the related persons of subject A extracted at the end of the third stage in the extraction result table 119.

In detail, the related party extraction unit 103 creates a new record for each related party in the extraction result table 119, registers the related party in related party 1191, registers "-" in factual relationship 1193, and registers "○" in analogical relationship 1194. Furthermore, in the fourth stage, the related party extraction unit 103 compares the related parties of the target A newly registered in the extraction result table 119 with the BL relevant party table 115 (FIG. 7) to confirm whether or not there is a BL relevant party, and registers the confirmation result in BL relevant party 1192 of the relevant record in the extraction result table 119. Specifically, in the fourth stage of step S4, R2 and S2 are registered in the extraction result table 119 as new related parties, and according to the BL relevant party table 115 in FIG. 7, both R2 and S2 are BL relevant parties. Therefore, in the records of R2 and S2 in the extraction result table 119, "○" is registered in the column value of BL relevant party 1192. Additionally, the organizations to which the subject and related parties belong (or belonged) are registered in the related organizations 1195 of each record. Note that in the extraction results table 119, four people B, C, E, and F are registered as related parties to subject A through a direct relationship in the fourth stage of step S3, and "○" is also registered in the analogical relationship existence 1194 of the records of these related parties.

As described above, in step S4, the second related party extraction can extract related parties who have an analogical relationship with subject A based on the analogy information, and the extraction result can be additionally registered in the extraction result table 119.

(3-4) Determination of High-Risk Individual Next, in step S5, the high-risk individual determination unit 104 refers to the extraction result table 119 registered in steps S3 and S4 to determine whether or not subject A is a high-risk individual, and outputs the final determination result to the user, after which the human relationship extraction process is terminated.

Here, the high-risk individual determination in this embodiment includes the first and second determinations described below.

First, in the first judgment, the high-risk individual judgment unit 104 uses the extraction result of the first related person extraction in step S3 to judge whether or not subject A is a high-risk individual based on whether or not there is a BL person among related persons who have a strong relationship (direct relationship) with subject A based on factual information.

Specifically, in the first determination, the high-risk individual determination unit 104 refers to the extraction result table 119 and checks whether there is a record in which both BL relevance 1192 and factual relationship 1193 are marked "○", thereby determining whether related parties who have a direct relationship with subject A are BL related parties. In the case of the extraction result table 119 in FIG. 11, referring to factual relationship 1193 shows that B, C, E, and F are related parties who have a direct relationship, and referring to BL relevance 1192 shows that C is a BL related party. In other words, since subject A has a direct relationship with C, who is a BL related party, the high-risk individual determination unit 104 determines subject A to be a high-risk person in the first determination.

According to this first determination, when subject A has a direct relationship with a BL patient, subject A can be determined to be a high-risk individual from the viewpoint that subject A has a strong relationship with the BL patient.

Next, in the second determination, the high-risk individual determination unit 104 uses the extraction results of the first related person extraction in step S3 or the second related person extraction in step S4 to determine whether or not subject A is a high-risk individual based on the extent to which BL-related individuals exist in subject A's surrounding environment.

Below, the second judgment will be explained in detail, with related parties of subject A who are not only directly related but also extended to analogical relationships (i.e., related parties extracted in the first and second related party extractions) as the "surrounding environment of subject A". As another example of the second judgment, the human relationship extraction device 100 according to this embodiment may perform the second judgment with only related parties who have a direct relationship with subject A (i.e., related parties extracted in the first related party extraction) as the "surrounding environment of subject A", in which case it is not necessary to execute the process of step S4 (second related party extraction) in the human relationship extraction process.

In the second determination, the high-risk person determination unit 104 calculates a BL relevance score based on the proportion of BL-related persons among the total number of related persons extracted as related persons of the subject A, and determines the subject A to be a high-risk person if the calculated BL relevance score is equal to or greater than a predetermined threshold. Specifically, the high-risk person determination unit 104 can obtain the total number of related persons from the total number of records by referring to the extraction result table 119, and can obtain the number of BL-related persons from the number of records in which "○" is registered in BL relatedness 1192. In the case of the extraction result table 119 in FIG. 11, the total number of related persons is six (B, C, E, F, R2, S2), of which three are BL-related persons (C, R2, S2), so the high-risk person determination unit 104 calculates the BL relevance score of the subject A to be 0.5 (3/6). Here, if a predetermined threshold value set in advance for the BL relevance score is, for example, 0.4, the BL relevance score of subject A is equal to or greater than the threshold value, so the high-risk individual determination unit 104 determines that subject A is a high-risk individual in the second determination.

According to this second determination, even if subject A does not have a direct relationship with a BL person, if there are many BL people in subject A's surrounding environment (including, for example, by analogy), subject A can be determined to be a high-risk person from the perspective that subject A is highly likely to be related to a BL person.

In addition, before making the first and second judgments described above, the high-risk individual judgment unit 104 may refer to the BL corresponding individual table 115 to check whether or not subject A is a BL corresponding individual, and if subject A is a BL corresponding individual, may judge the subject to be a high-risk individual.

Then, the high-risk individual determination unit 104 displays a determination result screen based on the results of the first and second determinations, thereby notifying the user, who is the examiner, of not only the final determination result as to whether or not subject A is a high-risk individual, but also the detailed determination results of each determination and any necessary countermeasures.

Figure 15 is a diagram showing an example of a judgment result screen. The judgment result screen 120 shown in Figure 15 is a display example of the judgment result screen output in step S5, and is displayed on the display of the human relationship extraction device 100 or the user's terminal, etc.

According to FIG. 15, the final result of the high-risk individual determination is displayed in area 121 of the determination result screen 120. When the high-risk individual determination unit 104 determines that the subject is a high-risk individual in at least either the first determination or the second determination, it displays "Corresponding to high-risk individual" in area 121.

In area 122 of the judgment result screen 120, detailed judgment results based on the first judgment and recommended measures for dealing with the judgment are displayed. Specifically, when the high-risk individual judgment unit 104 judges the subject A as a high-risk individual in the first judgment, it checks "BL applicable" in area 122 and displays that the subject A is related to a BL applicable individual and confirmation is required. Furthermore, the high-risk individual judgment unit 104 displays in area 122 the related person of the BL applicable individual who has a direct relationship with the subject A (for example, C in this example), and the relationship with the subject A. These display contents can be generated, for example, from the information of a record in which "○" is registered in both the BL applicable or not 1192 and the factual or not 1193 in the extraction result table 119.

In area 123 of the judgment result screen 120, detailed judgment results based on the second judgment and recommended measures for dealing with the judgment are displayed. Specifically, when the high-risk individual judgment unit 104 judges the subject A as a high-risk individual in the second judgment, it checks the "BL relevant individual relevance" in area 122 and displays that the subject A has a high relationship with a BL relevant individual and therefore confirmation is required. In addition, the high-risk individual judgment unit 104 displays in area 123 the BL relevant individual score of the subject A calculated in the second judgment, the total number of related parties (number of subjects) and the number of BL relevant individuals in the calculation. In this example, as described above, the BL relevant individual score is 0.5, the total number of related parties is 6, and the number of BL relevant individuals is 3. Furthermore, the high-risk individual determination unit 104 displays in the area 123, as information about the organizations that require confirmation, the organization to which the subject A belongs, the organization to which the BL relevant person who has an analogical relationship with the subject A belongs, the total number of related parties (number of subjects) in the organization to which the BL relevant person belongs, and the number of BL relevant persons. In this example, the organization to which the subject A belongs is organization P, and the BL relevant persons (R2, S2) who have an analogical relationship with the subject A are organizations R and S. The total number of related parties in organization R is only the BL relevant person R2, and the total number of related parties in organization S is only the BL relevant person S2. These display contents can be generated, for example, from the information of a record in the extraction result table 119 in which "x" is registered in the factual relationship presence/absence 1193 and "o" is registered in the analogical relationship presence/absence 1194.

In addition, on the judgment result screen 120, specific information may be highlighted, for example, in areas 122 and 123, by displaying the organization to which the BL person belongs in a different color or font from the others.

By outputting (displaying) the judgment result screen 120 as described above, the user can know whether or not the subject is a high-risk individual as a result of analyzing the related parties extracted while taking into account the subject's surrounding environment. Furthermore, such a judgment result screen 120 not only presents detailed information about the first and second judgments that determined whether or not the subject is a high-risk individual, but also items that require confirmation. Therefore, the user, who is the examiner, can reliably understand points where the examination should be stopped, and can prevent high-risk individuals from being overlooked.

As described above, the human relationship extraction device 100 according to this embodiment can extract related parties who have a direct relationship with the subject based on factual information, and if there is a BL related party among those related parties, the subject can be determined to be a high-risk supporter in the first judgment. At this time, the factual information includes not only basic information linked to the individual (person basic information table 114, subject relationship information table 116), but also the relationship between the individual and the organization (organization-person relationship table 112) and the relationship between organizations (organization relationship fact table 113), so that related parties can be extracted taking into account the subject's surrounding environment. Therefore, the human relationship extraction device 100 according to this embodiment can accurately identify high-risk individuals who are suspected of having a direct connection with a BL related party.

In addition, according to the human relationship extraction device 100 of this embodiment, related parties who have a direct or analogical relationship (indirect relationship) with the subject can be extracted based on factual information or analogical information, and whether the subject is a high-risk person in the second judgment can be determined based on the proportion of BL related parties among these related parties. Since the analogical information includes information indicating analogical relationships between organizations (organizational relationship analogy table 118), the use of the analogical information makes it possible to expand the surrounding environment of the subject and extract related parties. This method of determining high-risk persons by the second judgment takes into consideration the fact that when the proportion of BL related parties is high in the surrounding environment of the subject, the presence of brokers or the like who support crimes is assumed. Therefore, according to the human relationship extraction device 100 of this embodiment, even if a direct relationship between the subject and a BL related party is not found, it is possible to accurately identify high-risk persons who are suspected of having an indirect (environmental) connection with a BL related party.

In addition, in the human relationship extraction device 100 according to this embodiment, in the first and second judgments for determining high-risk individuals, a process is performed to exclude from related parties those whose period of affiliation to the organization does not overlap with that of the target person, thereby improving the accuracy of extraction of related parties.

10 Computer 11 CPU
12 Memory 13 Storage device 14 Input device 15 Output device 16 Network device 17 Bus 100 Human relationship extraction device 101 Input relationship registration unit 102 Organizational relationship analogy unit 103 Related person extraction unit 104 High-risk person determination unit 111 Person relationship table 112 Organization-person relationship table 113 Organizational relationship fact table 114 Person basic information table 115 BL applicable person table 116 Target person relationship information table 117 Individual similarity table 118 Organizational relationship analogy table 119 Extraction result table 120 Determination result screen

Claims (5)

A human relationship extraction device that uses a computer to determine whether a person to be screened is a high-risk person having a high level of risk,
A processor that executes a program to perform a predetermined process;
A storage device for storing the program and data;
Equipped with
the storage device stores at least information indicating a relationship between people, information indicating a relationship between organizations and people, information indicating a relationship between organizations and organizations, and information indicating a person on a blacklist, as known factual information input from outside;
The processor,
When the subject is designated, factual information associated with the subject is acquired and registered in the storage device;
further registering organizational relationship analogy information indicating an analogy relationship between organizations inferred from the factual information in the storage device;
Executing a first related person extraction to extract a first related person of the subject based on the various factual information stored in the storage device;
execute a second related person extraction to extract, based on the organizational relationship inference information, a person who has a history of affiliation to an organization that has the analogous relationship with an organization to which the target person has a history of affiliation, as a second related person of the target person, in addition to the first related person extracted by the first related person extraction;
execute a first determination as to whether or not the target person is a high-risk person based on whether or not the first related person extracted in the first related person extraction includes a person on the blacklist;
calculating a relevance score indicating the degree to which the second related person extracted in the second related person extraction corresponds to a person on the blacklist, and performing a second determination on the basis of the calculated relevance score to determine whether the target person is a high-risk person;
When at least one of the first and second determinations indicates that the subject is a high-risk individual, the subject is finally determined to be a high-risk individual;
a determination result of the final determination and determination results of the first and second determinations . The factual information stored in the storage device includes basic information about a plurality of individuals that indicates basic information associated with each of the individuals,
The processor,
Calculating an inter-personal similarity indicating a similarity between individuals based on the person basic information;
Inferring relationships between organizations based on the inter-personal similarity between people belonging to each organization;
The human relationship extraction device according to claim 1 , further comprising: registering the inferred analogical relationship between organizations in the organizational relationship inference information. The processor,
The human relationship extraction device according to claim 1, characterized in that, based on information indicating the factual relationship between the organization and the person, people whose period of affiliation to the organization does not overlap with the period of affiliation of the target person to the organization are excluded from the first and second related persons of the target person. The processor,
The human relationship extraction device according to claim 1, characterized in that, when the first or second judgment results in a judgment that the subject is a high-risk person, an output is given recommending checking with people or organizations that are thought to have relationships with the subject on the blacklist . A human relationship extraction method using a human relationship extraction device that determines whether a person to be examined is a high-risk person having a high level of risk,
The computer constituting the human relationship extraction device has a processor that executes a predetermined process by executing a program, and a storage device that stores the program and data,
the storage device stores at least information indicating a relationship between people, information indicating a relationship between organizations and people, information indicating a relationship between organizations and organizations, and information indicating a person on a blacklist, as known factual information input from outside;
a fact information registration step in which, when the subject is designated, the processor acquires fact information associated with the subject and registers the fact information in the storage device;
an inference information registration step in which the processor further registers, in the storage device, organizational relationship inference information indicating an inferred relationship between organizations inferred from the fact information;
A first related person extraction step in which the processor extracts a first related person of the subject based on various types of the factual information stored in the storage device;
a second related person extraction step in which the processor extracts, as a second related person of the subject, a person who has a history of affiliation to an organization that has the analogous relationship with an organization to which the subject has a history of affiliation, based on the organizational relationship inference information registered in the analogy information registration step, in addition to the first related person extracted in the first related person extraction step;
a first determination step in which the processor determines whether the target person is a high-risk person based on whether the first related person extracted in the first related person extraction step includes a person on the blacklist;
a second determination step in which the processor calculates an association score indicating the degree to which the second related person extracted in the second related person extraction step is included in the blacklist, and determines whether the target person is a high-risk person based on the calculated association score;
a final determination step in which the processor finally determines that the subject is a high-risk individual when a determination result that the subject is a high-risk individual is obtained in at least one of the first and second determination steps;
an output step in which the processor outputs a determination result in the final determination step and the determination results in the first and second determination steps ;
A human relationship extraction method comprising:

Images