JP7123326B1 - Information processing method, information processing program, information processing apparatus, and learning model generation method - Google Patents

Abstract

An object of the present invention is to provide an information processing method or the like for outputting the same person probability of a reference person and a referrer who are considered to be the same person.
In an information processing method, a computer acquires information on a standard contract, acquires attribute items of a standard person selected from parties and related parties of the standard contract from the information on the standard contract, acquire the attribute items of the referrer from the information of the reference contract that includes the referrer similar to the reference person as a party or related party, calculate the similarity between the attribute items of the reference person and the attribute items of the referrer, When the similarity between the attribute items of a person and the attribute items of a referrer is input, the calculated similarity is input to a learning model that outputs the same person probability indicating the probability that the reference person and the referrer are the same person. Then, the same person probability of the reference person and the reference person output by the learning model is acquired, and the process of outputting the same person probability and the parties and related parties of the reference contract is executed.
[Selection drawing] Fig. 14

Description

TECHNICAL FIELD The present invention relates to an information processing method and the like for outputting a set of customer information as name identification candidates.

When customer information is managed redundantly, it is not possible to properly grasp purchase histories, etc. As a result, there may be a problem in that direct mail is sent regarding products and services that the customer has already purchased. In order to prevent this, name identification is performed to collect duplicate customer information. It is difficult to perform name identification when items included in a plurality of pieces of customer information are different.

In response to such a situation, an information management system has been proposed for supporting name identification of information based on various services (Patent Document 1). In the information management system disclosed in Patent Document 1, when the first and second user-related information are newly acquired from the first and second service servers, for the combination of the newly calculated first score and second score , applying a predictive model to predict the likelihood of being the same person.

Japanese Patent Application Laid-Open No. 2021-033465

However, the information management system disclosed in Patent Literature 1 is mainly aimed at proposing new transactions to customers, and does not present the likelihood of being the same person. The present invention has been made in view of such circumstances. The purpose is to provide an information processing method or the like for outputting the same person probability of a reference person and a reference person who may be the same person.

In the information processing method disclosed in the present application, a computer acquires information on a standard contract, acquires from the information on the standard contract, the attribute items of the standard person selected from the parties and related parties of the standard contract, and obtaining the attribute items of the referrer from the information of a reference contract that includes, as a party or a related party, a referrer similar to the reference person whose attribute items partly match those of the reference person, and the attribute items of the reference person and the reference Same person probability indicating the probability that the reference person and the reference person are the same person when the similarity between the reference person's attribute items and the referrer's attribute items is calculated and the similarity between the reference person's attribute items and the referrer's attribute items is input input the calculated similarity to the learning model that outputs , obtain the same person probability of the reference person and the reference person output by the learning model, and party or related party other than the reference person included in the reference contract and adjusts the same person probability based on the degree of matching with a party or related party other than the referrer of the reference contract, and outputs the adjusted same person probability and the parties and related parties of the reference contract. It is characterized by executing processing.

In one aspect of the present application, it is possible to confirm the same person probability of a reference person and a reference person who may be the same person.

1 is an explanatory diagram showing a configuration example of a name identification system; FIG. It is a block diagram which shows the hardware structural example of an AP server. 3 is a block diagram showing a hardware configuration example of a user terminal; FIG. It is explanatory drawing which shows the example of customer DB. It is explanatory drawing which shows the example of contract DB. It is explanatory drawing which shows the example of candidate set DB. It is explanatory drawing which shows the example of similarity DB. It is explanatory drawing which shows the example of determination result DB. It is explanatory drawing which shows the example of result DB. It is explanatory drawing which shows the example of item result DB. It is explanatory drawing which shows the example of training DB. 6 is a flow chart showing an example of a procedure of model generation processing; 6 is a flow chart showing an example of a preprocessing procedure; 7 is a flow chart showing an example of a procedure of determination processing; 10 is a flow chart showing an example of a procedure of name identification processing; 6 is a flowchart illustrating an example of a procedure of relearning processing; FIG. 5 is an explanatory diagram showing an example of a display screen; FIG. 11 is an explanatory diagram showing an example of a name identification operation screen; FIG. 5 is an explanatory diagram showing an example of an item setting screen; FIG. 5 is an explanatory diagram showing an example of adjustment; 7 is a flowchart illustrating an example of a procedure of adjustment processing;

(Embodiment 1)
Embodiments will be described below with reference to the drawings. FIG. 1 is an explanatory diagram showing a configuration example of a name identification system. The name identification system 100 includes an AP server 1 and user terminals 2 . The AP server 1 and the user terminal 2 are connected by a network N so as to be able to communicate with each other. Also, the AP server 1 and the user terminal 2 can access the core system DB 200 via the network N. FIG.

The AP server 1 is composed of a server computer, a workstation, a PC (Personal Computer), and the like. Also, the AP server 1 may be composed of a multicomputer composed of a plurality of computers, a virtual machine or a quantum computer virtually constructed by software. Furthermore, the functions of the AP server 1 may be realized by a cloud service.

The user terminal 2 is composed of a notebook computer, a panel computer, a tablet computer, a smart phone, or the like. Although two user terminals 2 are shown in FIG. 2, the number may be one or three or more. Note that the user is an employee or the like who performs the name identification work.

The backbone system DB 200 is a database of the backbone system. The core system DB 200 stores customer data to be subjected to name identification processing. In the following description, we assume the customer data of a life insurance company. In a life insurance contract, in addition to the policyholder who is a party, the insured person, the person who bears the insurance premium (hereinafter simply referred to as the "payer"), the beneficiary of the benefits and insurance money (hereinafter simply referred to as the "recipient ”), etc. Here, the customer data includes not only data of the contractor but also data of related parties.

FIG. 2 is a block diagram showing a hardware configuration example of an AP server. The AP server 1 includes a control section 11 , a main storage section 12 , an auxiliary storage section 13 , a communication section 15 and a reading section 16 . A bus B connects the control unit 11 , the main storage unit 12 , the auxiliary storage unit 13 , the communication unit 15 and the reading unit 16 .

The control unit 11 has one or more arithmetic processing units such as a CPU (Central Processing Unit), an MPU (Micro-Processing Unit), and a GPU (Graphics Processing Unit). The control unit 11 reads and executes the control program 1P (program, program product) stored in the auxiliary storage unit 13 to perform various information processing, control processing, etc. related to the AP server 1, and the contract acquisition unit, It implements functional units such as an attribute acquisition unit, a reference attribute acquisition unit, a calculation unit, a probability acquisition unit, and an output unit.

The main storage unit 12 is SRAM (Static Random Access Memory), DRAM (Dynamic Random Access Memory), flash memory, or the like. The main storage unit 12 mainly temporarily stores data necessary for the control unit 11 to execute arithmetic processing.

The auxiliary storage unit 13 is a hard disk, SSD (Solid State Drive), or the like, and stores the control program 1P and various DBs (Databases) necessary for the control unit 11 to execute processing. The auxiliary storage unit 13 stores a customer DB 131, a contract DB 132, a candidate set DB 133, a similarity DB 134, a determination result DB 135, a result DB 136, an item result DB 137, and a training DB 138. Also, the auxiliary storage unit 13 stores a learning model 141 . The auxiliary storage unit 13 may be an external storage device externally connected to the AP server 1 separately. Various DBs and the like stored in the auxiliary storage unit 13 may be stored in a database server or cloud storage different from the AP server 1 .

The communication unit 15 communicates with the user terminal 2 via the network N. FIG. Alternatively, the control unit 11 may use the communication unit 15 to download the control program 1P from another computer via the network N or the like and store it in the auxiliary storage unit 13 .

The reader 16 reads portable storage media 1a including CD (Compact Disc)-ROM and DVD (Digital Versatile Disc)-ROM. The control unit 11 may read the control program 1P from the portable storage medium 1a via the reading unit 16 and store it in the auxiliary storage unit 13 . Alternatively, the control unit 11 may read the control program 1P from the semiconductor memory 1b.

FIG. 3 is a block diagram showing a hardware configuration example of a user terminal. The user terminal 2 includes a control section 21 , a main memory section 22 , an auxiliary memory section 23 , a communication section 24 , an input section 25 and a display section 26 . Each configuration is connected by a bus B.

The control unit 21 has one or more arithmetic processing units such as CPU, MPU, and GPU. The control unit 21 provides various functions by reading and executing the control program 2P (program, program product) stored in the auxiliary storage unit 23 .

The main storage unit 22 is SRAM, DRAM, flash memory, or the like. The main storage unit 22 mainly temporarily stores data necessary for the control unit 21 to execute arithmetic processing.

The auxiliary storage unit 23 is a hard disk, SSD, or the like, and stores various data necessary for the control unit 21 to execute processing. The auxiliary storage unit 23 may be an external storage device externally connected to the user terminal 2 separately. Various DBs and the like stored in the auxiliary storage unit 23 may be stored in a database server or cloud storage.

The communication unit 24 communicates with the AP server 1 via the network N. FIG. Alternatively, the control unit 21 may use the communication unit 24 to download the control program 2P from another computer via the network N or the like and store it in the auxiliary storage unit 23 .

The input unit 25 is a keyboard or mouse. The display unit 26 includes a liquid crystal display panel and the like. The display unit 26 displays customer information output by the AP server 1 and the like. Also, the input unit 25 and the display unit 26 may be integrated to form a touch panel display. Note that the user terminal 2 may display on an external display device.

Next, the database used by the name identification system 100 will be described. FIG. 4 is an explanatory diagram showing an example of a customer DB. The customer DB 131 stores data of customers (including contractors and related parties). The customer DB 131 includes a customer number column, a name column, a furigana column, a gender column, a date of birth column, an address column, and a phone number column. The customer number column stores customer numbers that can identify customers. The name column stores the name of the customer. The furigana column stores the furigana of the customer name. The sex column stores alphabets indicating the customer's sex. If the customer is male, the gender column stores M. The gender column stores F if the customer is female. The date of birth column stores the date of birth of the customer. The address column stores the customer's address. The phone number column stores the customer's phone number.

FIG. 5 is an explanatory diagram showing an example of the contract DB. The contract DB 132 stores the contents of insurance contracts. The contract DB 132 includes a policy number column, a policyholder column, an insured person column, a bearer column, and a beneficiary column. The policy number column stores policy numbers attached to insurance policies. The policyholder column stores the customer number of the policyholder. The insured person column stores the customer number of the insured person. The payer column stores the customer number of the insurance premium payer. The beneficiary column stores the information of the beneficiary of the insurance money. The payee column contains a number column and a customer column. The number column stores the sequential number of the recipient. The sequential numbers are assigned to identify the records, and the order of the numbers has no particular meaning. The customer column stores the recipient's customer number.

As for the customer DB 131 and the contract DB 132, the core system DB 200 should hold the master data. Therefore, the customer DB 131 and the contract DB 132 stored in the AP server 1 may be copied from the core system DB 200, or may be copied from the core system DB 200 only for items necessary for name identification.

FIG. 6 is an explanatory diagram showing an example of the candidate set DB. The candidate set DB 133 stores combination information of a plurality of customers to be identified. The candidate set DB 133 includes a PID column, a reference person column, a reference contract column, a target person column, and a target contract column. The PID column stores PIDs that identify candidate sets. The reference person column stores customer information that is used as a reference for name identification. The reference person column includes a customer number column and a position column. The customer number column stores the reference person's customer number. The position column stores the position in the contract of the reference party. A position is a policyholder, insured, bearer or beneficiary. The reference contract column stores policy numbers of insurance contracts linked to the reference party. The target person column stores information on customers to be identified. The target person column includes a number column, a customer number column and a position column. The number column stores the sequential number of the subject. The sequential numbers are assigned to identify the records, and the order of the numbers has no particular meaning. The customer number column stores the customer number of the target person. The position column stores the target's position in the contract. The target contract column stores the policy number of the insurance contract linked to the target person.

FIG. 7 is an explanatory diagram showing an example of a similarity DB. The similarity DB 134 stores the similarity between the reference person and the target person for each attribute item. The similarity DB 134 stores a PID column, a number column, a name column, a kana column, a gender column, a date of birth column, an address column, and a telephone number column. The PID column stores PIDs that identify candidate sets. The number column stores the sequential number of the subject. The sequential numbers are assigned to identify the records, and the order of the numbers has no particular meaning. The name column stores the degree of similarity between the name of the reference person and the name of the subject. The reading column stores the degree of similarity between the reading of the reference person's name and the reading of the target person's name. The gender column stores the degree of similarity between the gender of the reference person and the gender of the subject. The gender column stores 1 if the genders match, 0 if they do not match, and 0.5 if either gender is unknown. The date of birth column stores the degree of similarity between the date of birth of the reference person and the date of birth of the subject. The address column stores the degree of similarity between the reference person's address and the target person's address. The telephone number column stores the degree of similarity between the reference person's telephone number and the target person's telephone number. For example, the similarity may be a cosine similarity between the reference person and the target person expressed as a vector using Word2Vec or Doc2Vec. Further, the value of the reference person and the value of the target person may be compared with prefix matching character strings, and the degree of similarity may be calculated from the length of the matching character string and the length of the entire character string.

FIG. 8 is an explanatory diagram showing an example of the determination result DB. The determination result DB 135 stores the result of determination by the learning model 141 as to whether the reference person and the target person are the same person. The PID column stores PIDs that identify candidate sets. The reference person column stores the information of the reference person. The reference person column includes a customer number column and a position column. The customer number column stores the reference person's customer number. The position column stores the position in the contract in which the reference party is included. The reference contract column stores policy numbers of contracts that include the reference party. The target person column stores the information of the target person. The target person column includes a number column, a customer number column and a position column. The number column stores sequential numbers. The sequential numbers are assigned to identify the records, and the order of the numbers has no particular meaning. The customer number column stores the customer number of the target person. The position column stores the position of the subject. The subject contract column stores the policy number of the contract that includes the subject. The evaluation column stores evaluation values. The evaluation value is a value from 0 to 100, for example. 0 indicates that the probability that the reference person and the target person are the same person is 0%. 100 indicates that the probability that the reference person and the subject person are the same person is 100%.

FIG. 9 is an explanatory diagram showing an example of the result DB. The result DB 136 refers to the determination result of the learning model 141 and stores the determination result as to whether or not the user has determined that the reference person and the target person are the same person. The result DB 136 includes a PID column, a reference person column, a target person column, an evaluation column and a result column. The PID column stores PIDs that identify candidate sets. The reference person column stores the customer number of the reference person. The target person column stores the information of the target person. The target person column includes a number column, a customer number column and a position column. The number column stores sequential numbers. The sequential numbers are assigned to identify the records, and the order of the numbers has no particular meaning. The customer number column stores the customer number of the target person. The position column stores the position of the subject. The evaluation column stores evaluation values. The evaluation value stores the same value as the value of the evaluation column in the record in which the value of the PID column and the value of the number column of the subject column match in the determination result DB 135 . The result column stores judgment results. For example, when the user determines that the reference person and the target person are the same person, ◯ is stored in the result column. If the user determines that the reference person and the target person are not the same person, x is stored in the result column.

FIG. 10 is an explanatory diagram showing an example of the item result DB. The item result DB 137 stores whether the user has determined to unify the attribute items of the reference person and the target person to either the reference person or the target person when the user determines that the reference person and the target person are the same person. do. The item result DB 137 includes a PID column, a reference person column, a target person column, a name column, a kana column, a sex column, a date of birth column, an address column, and a telephone number column. The PID column stores PIDs that identify candidate sets. The reference person column stores the customer number of the reference person. The target person column stores the customer number of the target person. The name column, the kana column, the sex column, the date of birth column, the address column, and the telephone number column each store whether the value is determined to be the standard person or the subject person. For example, if the user decides to rewrite the value of the subject to the value of the reference person, each column stores the reference person. Each column stores a subject if the user decides to rewrite the reference person's value to the subject's value.

FIG. 11 is an explanatory diagram showing an example of a training DB. The training DB 138 stores training data used when generating or re-learning the learning model 141 . The training DB 138 includes a number column, a label column, a name column, a furigana column, a gender column, a date of birth column, an address column, and a phone number column. The number column stores sequential numbers. The sequential numbers are assigned to identify the records, and the order of the numbers has no particular meaning. The label column stores the correct judgment value. 0 indicates that it should be determined that they are not the same person, and 1 indicates that it should be determined that they are the same person. A name column, a furigana column, a gender column, a date of birth column, an address column, and a telephone number column are the degree of similarity for each item between the reference person and the target person. Items included in the training data can be generated from the similarity DB 134 or the result DB 136 . Therefore, after starting to use the learning model 141, training data can be created from the similarity DB 134 or the result DB 136, and the learning model 141 can be re-learned.

Next, the learning model 141 will be described. The learning model 141 is, for example, a neural network such as a CNN (Convolution Neural Network). The learning model 141 has an input layer that receives similarity for each item, an output layer that outputs the probability that the reference person and the target person are the same person, and an intermediate layer that outputs the similarity between the reference person and the target person. .

The input layer has a plurality of neurons that receive input of similarity for each item, and passes the input similarity to the intermediate layer. The intermediate layer has a plurality of neurons, and has a configuration in which a convolution layer that convolves the similarity of each item input from the input layer and a pooling layer that maps the similarity convoluted in the convolution layer are alternately connected. and outputs the similarity between the reference person and the target person. The output layer uses a softmax function or the like to convert the degree of similarity from the intermediate layer into a probability that the reference person and the target person are the same person, and outputs the result. The learning model 141 is not limited to CNN, and may be a learning model constructed by other learning algorithms such as neural networks, Bayesian networks, and decision trees other than CNN.

Next, information processing performed by the name identification system 100 will be described. FIG. 12 is a flow chart showing an example of the procedure of model generation processing. Model generation processing is processing for generating the learning model 141 . The control unit 11 of the AP server 1 acquires one record of training data from the training DB 138 (step S1). The control unit 11 performs learning (step S2). The control unit 11 inputs the degree of similarity for each item included in the training data to the input layer, and obtains the probability that the reference person and the target person are the same person from the output layer through arithmetic processing in the intermediate layer. The probability is a value ranging from 0 to 1. Note that the probability may be a discrete value (0 or 1). The control unit 11 compares the probability output from the output layer with the value of the label included in the training data, that is, the correct value, and adjusts the arithmetic processing in the intermediate layer so that the output value from the output layer approaches the correct value. Optimize the parameters used. The parameters are, for example, weights (coupling coefficients) between neurons, coefficients of activation functions used in each neuron, and the like. Although the parameter optimization method is not particularly limited, for example, the control unit 11 optimizes various parameters using the error backpropagation method. The control unit 11 determines whether or not there is unprocessed training data (step S3). When the control unit 11 determines that there is unprocessed training data (YES in step S3), the process returns to step S1, and learning is performed using the unprocessed training data. When the control unit 11 determines that there is no unprocessed training data (NO in step S3), the learning model 141 including the parameters optimized by learning is stored in the auxiliary storage unit 13 (step S4), and the processing is performed. exit.

FIG. 13 is a flow chart showing an example of a preprocessing procedure. The pre-processing is processing for making preparations such as creating a combination of a reference person and a subject person. First, as a premise, it is assumed that the control unit 11 acquires the added or changed customer information and contract information from the core system DB 200 as difference data. The control unit 11 normalizes the differential data (step S11). For example, the description format of "1-3-5 Nagatacho" included in the address is changed to the description format of "1-3-5 Nagatacho". In the telephone number, the description format of "(03) 123-4567" is changed to the description format of "03-123-4567". In the date of birth, the description format of "September 2, 2001" is changed to the description format of "2001/09/02". The example of change is merely an example, and other forms may be used as long as the form is unified. The control unit 11 creates a candidate set (step S12). A candidate set is a combination of one reference person and one or more target persons to be identified. For example, the control unit 11 uses a customer whose customer information has been added or changed as a reference person. Alternatively, the control unit 11 uses the parties to the contract or related persons included in the added or changed contract information as the reference person. The target person is a customer who searches the customer DB 131 and matches the reference person in predetermined items, such as the phonetic name of the name, and the date of birth or gender. The control unit 11 stores the candidate pair in the candidate pair DB 133 (step S13), and terminates the process. Steps S12 and S13 are repeatedly executed according to the number of reference persons.

FIG. 14 is a flow chart showing an example of the procedure of determination processing. The determination process is a process of determining whether or not the reference person and the target person are the same person. The control unit 11 acquires a set of data from the candidate set DB 133 (step S21). The control unit 11 acquires the attributes of the reference person and the target person included in the candidate set from the customer DB 131, and calculates the degree of similarity for each item (step S22). The control unit 11 makes a determination (step S23). The control unit 11 inputs the degree of similarity for each item to the learning model 141 and acquires the probability output by the learning model 141 . The processing is performed according to the number of subjects. The control unit 11 stores the determination result in the determination result DB 135 (step S24). The control unit 11 determines whether or not there is unprocessed candidate group data (step S25). If the control unit 11 determines that there is unprocessed candidate group data (YES in step S25), the process returns to step S21 to process the unprocessed candidate group data. When the control unit 11 determines that there is no unprocessed candidate set data (NO in step S25), the process ends. Note that the similarity calculation (step S22) may be performed in preprocessing.

It is assumed that preprocessing and determination processing are performed as a daily batch. The core system DB 200 creates difference data of customer information and contract information in daily batches. After the difference data is created, the AP server 1 may execute preprocessing and determination processing in order.

FIG. 15 is a flow chart showing a procedure example of the name identification process. The name identification process is a process for the user to determine whether to perform the name identification between the reference person and the reference person based on the determination result DB 135 . The user performs an operation to start name identification processing on the user terminal 2 . The control unit 21 of the user terminal 2 transmits a request for name identification candidate set data to the AP server 1 (step S31). The control unit 11 of the AP server 1 receives the request (step S32). The control unit 11 selects unprocessed candidate set data (step S33). The control unit 11 generates an operation screen and transmits it to the user terminal 2 (step S34). The control unit 21 of the user terminal 2 receives the screen (step S35). The control unit 21 displays the received screen (step S36). The user looks at the operation screen, judges whether the reference person and the reference person should be named, and performs an operation according to the judgment. The control unit 21 accepts the operation (step S37). The control unit 21 determines whether or not the received operation is completed (step S38). When the control unit 21 determines that the received operation is not completed (NO in step S38), it stores the details of the operation in the auxiliary storage unit 23 (step S39), and returns the process to step S36. The contents of the operation are an instruction operation for merging the reference person or the referrer, and an operation for instructing for each attribute item whether the attribute value should be matched to the reference person or the referrer when merging names. When the control unit 21 determines that the received operation is completed (YES in step S38), it transmits the determination result of the user to the AP server 1 (step S40). Note that the control unit 11 creates the determination result based on the operation content stored in the auxiliary storage unit 23 or the like. The control unit 11 of the AP server 1 receives the judgment result (step S41). The control unit 11 generates training data for re-learning based on the determination result (step S42). Control unit 11 stores the determination result in result DB 136 and item result DB 137, stores the training data in training DB 138 (step S43), and ends the process. Note that if there is no referrer to be named, the control unit 11 does not store in the item result DB 137 . When the user continues to perform the name identification process, the name identification process is performed again by the user's operation. The contents of the item result DB 137 are transmitted to the core system, and the data are sorted in the core system DB 200 . The judgment results stored in the result DB 136 and item result DB 137 are examples of association information.

FIG. 16 is a flow chart showing an example of a procedure of relearning processing. The control unit 11 of the AP server 1 acquires training data for re-learning from the training DB 138 (step S61). The control unit 11 uses the training data to learn the learning model 141 (step S62). The control unit 11 determines whether or not there is unprocessed training data (step S63). When the control unit 11 determines that there is unprocessed training data (YES in step S63), the process returns to step S61, and learning is performed using the unprocessed training data. When the control unit 11 determines that there is no unprocessed training data (NO in step S63), it compares the performance of the learning model 141 before and after relearning (step S64). For example, training data for relearning is input to the learning model 141 before and after relearning, and it is determined whether the accuracy rate is improved. The control unit 11 determines whether or not the performance of the learning model 141 has been improved (step S65). When the control unit 11 determines that the performance of the learning model 141 is improved (YES in step S65), it updates the learning model 141 to the one after re-learning (step S66), and terminates the process. When the control unit 11 determines that the performance of the learning model 141 has not been improved (NO in step S65), the process ends.

FIG. 17 is an explanatory diagram showing an example of a display screen. The display screen includes reference contract information 171 , reference contract information 172 and 173 and an attribute table 174 . The standard contract information 171 is information of the standard contract including the standard person. The reference contract information 171 includes the policy number and the names of the policyholder, insured, bearer and beneficiary. The reference contract information 172 and 173 is information of a reference contract including a referrer. The reference contract information 172, 173 includes the policy number and the names of the policyholder, insured, bearer and beneficiary. The attribute table 174 is a table that enables visual recognition of differences in attribute items between the reference person and the referrer.

In FIG. 17, the reference person is Ichiro Tanaka. Of the referrers, Ichiro Tanaka 1731 as the payer included in the reference contract information 173 has a probability of being the same person as the reference person (probability of same person) 1732 is 98%, and Ichiro Tanaka 1731 as the payee has a same person probability 1732. is determined to be 85%.

FIG. 18 is an explanatory diagram showing an example of the name identification operation screen. The name identification operation screen is a screen for performing the name identification operation. For example, when the user performs a predetermined operation after determining the reference person and the referrer to be identified with reference to the display screen shown in FIG. 17, the control unit 21 displays the name identification operation screen. The name identification operation screen is displayed in such a manner that the positional relationship with the reference person differs depending on the same person probability of the referrer. In FIG. 18, the referrer with a higher same-person probability is displayed at a position closer to the reference person. On the name identification operation screen, the user drags and drops the referrer icon 182 to be identified to the reference person icon 181 . This operation is an instruction operation for merging the names of the two. On the name identification operation screen, the positional relationship with the reference person is displayed differently depending on the same person probability of the referrer. The reference person icon 181 is an example of a reference person display area.

FIG. 19 is an explanatory diagram showing an example of the item setting screen. The item setting screen is a screen for designating which person's item value should be unified for each attribute item when matching the reference person and the reference person. By operating the directional icons 191 and 192, the user designates who adopts the value for each item. A direction icon 191 indicates that the attribute items of the reference person are unified with the attribute items of the referrer. A direction icon 192 indicates that the attribute items of the referrer are unified with the attribute items of the reference person. For example, the user gives an instruction by the following operation. When the direction icon 191 is mouse-clicked, it switches to the direction icon 192 , and when the direction icon 192 is mouse-clicked, it switches to the direction icon 191 . After the directions for all items have been determined, the user selects the confirm button 193 . When the user selects the confirm button 193, the user terminal 2 stores the settings. When the user selects the cancel button 194, the item settings are discarded.

By the user's operation, the reference person and the information of the referrer to be identified, and the information of which person to unify each attribute item are generated. The generated information is transmitted from the user terminal 2 to the AP server 1 as a determination result.

This embodiment has the following effects. Since the probability that the reference person and the reference person are the same person is displayed, the user can accurately specify the combination of the reference person and the reference person to be identified. In addition, on the display screen shown in FIG. 17, in addition to the parties (policyholders), related parties such as the insured person, the bearer, and the beneficiary are also displayed. By looking at the related persons, the user can determine whether the reference person and the referrer are the same person. Furthermore, by referring to the attribute table 174 of FIG. 17, the user can determine whether the reference person and the reference person are the same person. The name identification operation screen shown in FIG. 18 and the item setting screen shown in FIG. 19 allow the user to intuitively instruct the name identification, so that it is possible to suppress the occurrence of operation errors.

(Embodiment 2)
This embodiment is a form in which the reference contract and the target contract are compared in determining whether or not the reference person and the target person are the same person. In the following description, contents similar to those of the first embodiment are omitted as much as possible, and contents different from the first embodiment are mainly described.

In the present embodiment, the degree of matching between parties or parties other than the standard person or the target person is examined in comparison between the standard contract including the standard person and the target contract including the target person. Then, when the same person is included in the reference contract and the target contract other than the reference person or the target person, the same person probability between the reference person and the target person is adjusted. FIG. 20 is an explanatory diagram showing an example of adjustment. FIG. 20A shows the case of comparison only between a conventional reference person and a subject. FIG. 20B shows a case where parties or parties other than the reference person or the target person are compared. In FIG. 20, there are three addresses stored as addresses, which are indicated by numbers. As for the telephone number, 03 indicates that the fixed telephone number is stored, and 090 indicates that the mobile telephone number is stored.

In FIG. 20A, the reference person and the target person have different addresses and different telephone numbers. The reference person's phone number is for a landline phone and the subject's phone number is for a mobile phone. Therefore, since the phone numbers are significantly different, the probability that the reference person and the target person are the same person is determined to be 55%. In FIG. 20B, the insured person of the reference contract and the insured person of the target contract are the same Hanako Tanaka. Since the customer IDs match, Hanako Tanaka of both contracts is the same person. Therefore, the doujin probability is adjusted from 55% to 92%.

FIG. 21 is a flow chart showing an example of the procedure of adjustment processing. The adjustment process is executed between steps S23 and S24 in the determination process shown in FIG. The control unit 11 of the AP server 1 acquires information on the reference contract and the target contract included in the candidate set (step S81). The control unit 11 searches the contract DB 132 using the security numbers of the reference contract and the target contract included in the candidate set as keys, and obtains information on the parties and related parties included in the reference contract and the parties and relationships included in the target contract. obtain the information of the person. The control unit 11 compares the parties and persons involved in both contracts, excluding the reference person and the target person (step S82). The control unit 11 determines whether or not the same person is included in both contracts (step S83). For example, when a person with the same customer ID is included in both contracts, YES is determined in step S83. If the control unit 11 determines that the same person is included between both contracts (YES in step S83), the control unit 11 adjusts the same person probability (step S84), and ends the process. For example, processing such as multiplying the same-person probability by a predetermined coefficient, adding a predetermined value to the same-person probability, and updating the same-person probability using a table prepared in advance is performed as adjustment. Note that after the adjustment, the same person probability becomes a larger value than before the adjustment. The adjustment range may be changed according to the number of the same person included in both contracts. If the control unit 11 determines that the same person is not included in both contracts (NO in step S83), the process ends.

In the above description, the determination result of the learning model 141 is adjusted according to the comparison result between the reference contract and the target contract, but it is not limited to this. Inputs to the learning model 141 may include information about parties other than the reference person included in the reference contract and information about parties other than the target person included in the target contract. In this case, the learning model 141 is generated or re-learned by including the information of the persons concerned other than the reference person and the information of the persons concerned other than the target person in the training data. When the learning model 141 determines whether or not the reference person and the target person are the same person, the information of the persons concerned other than the reference person and the information of the persons concerned other than the target person are included as inputs.

This embodiment has the following effects in addition to the effects of the first embodiment. Comparing the standard contract that includes the standard company and the target contract that includes the target company, and if the same person is included between the parties and related parties of both contracts excluding the standard company and the target company, the standard company And increase the same person probability of the target person. This allows the user to determine with certainty that the reference person and the target person are the same person.

The technical features (components) described in each embodiment can be combined with each other, and new technical features can be formed by combining them.
The embodiments disclosed this time are illustrative in all respects and should not be considered restrictive. The scope of the present invention is indicated by the scope of the claims rather than the above-described meaning, and is intended to include all modifications within the scope and meaning equivalent to the scope of the claims.

100 System 200 Core system DB
1 AP server 11 control unit 12 main storage unit 13 auxiliary storage unit 131 customer DB
132 Contract DB
133 Candidate set DB
134 Similarity DB
135 Judgment result DB
136 Result DB
137 item result DB
138 training database
141 learning model 15 communication unit 16 reading unit 1P control program 1a portable storage medium 1b semiconductor memory 2 user terminal B bus N network

Claims (12)

the computer
Get the information of the standard contract,
Acquiring the attribute items of the standard person selected from the parties and related parties of the standard contract from the information of the standard contract,
obtaining the attribute items of the referrer from information of a reference contract that includes, as a party or a related party, a referrer similar to the reference person whose attribute items partially match the reference person;
calculating the degree of similarity between the attribute items of the reference person and the attribute items of the referrer;
When the similarity between the reference person attribute item and the referrer attribute item is input, the calculated similarity is input to the learning model that outputs the same person probability indicating the probability that the reference person and the referrer are the same person. Input,
Acquiring the same person probability of the reference person and the reference person output by the learning model,
Adjust the same person probability based on the degree of matching between a party or related party other than the reference person included in the reference contract and a party or related party other than the referrer of the reference contract,
An information processing method, characterized by executing a process of outputting the adjusted same-person probability and the parties and related parties of the reference contract. calculating the same person probability of each of the plurality of referrers with respect to one of the reference persons;
2. The information processing method according to claim 1 , wherein the same person probability of each of the plurality of referrers and the parties and related parties of the referral contract are output. outputting the reference person, the plurality of the reference persons and the respective same person probabilities based on the same person probability of each of the reference persons so that the positional relationship with the reference person is displayed in a different manner; 3. The information processing method according to claim 2 . 4. The information processing method according to any one of claims 1 to 3 , further comprising: outputting parties and related parties of the reference contract. If multiple candidates are the same person or presumed to be the same person among the parties and related parties of the reference contract and the said parties and related parties of the reference contract, said same person or said candidate person output so that it is visibly
5. The information processing method according to claim 4, characterized in that: 6. The information processing method according to any one of claims 1 to 5, wherein differences between the attribute items of the reference person and the attribute items of the referrer are output so as to be visually recognizable. In a screen on which the reference person and the referrer are displayed, when an icon indicating the reference person is moved to the display area of the reference person, the correspondence information of the reference person and the reference person is generated. 7. The information processing method according to any one of claims 1 to 6. obtaining the correspondence information;
8. The information processing method according to claim 7, wherein the learning model is re-learned based on the attribute items of the reference person and the attribute items of the referrer. 9. The method according to claim 7 or 8, wherein the association information includes information as to which of the value of the reference person and the value of the referrer is used for unifying each of the attribute items. Information processing methods. Get the information of the standard contract,
Acquiring the attribute items of the standard person selected from the parties and related parties of the standard contract from the information of the standard contract,
obtaining the attribute items of the referrer from information of a reference contract that includes, as a party or a related party, a referrer similar to the reference person whose attribute items partially match the reference person;
calculating the degree of similarity between the attribute items of the reference person and the attribute items of the referrer;
When the similarity between the reference person attribute item and the referrer attribute item is input, the calculated similarity is input to the learning model that outputs the same person probability indicating the probability that the reference person and the referrer are the same person. Input,
Acquiring the same person probability of the reference person and the reference person output by the learning model,
Adjust the same person probability based on the degree of matching between a party or related party other than the reference person included in the reference contract and a party or related party other than the referrer of the reference contract,
An information processing program for causing a computer to perform a process of outputting the adjusted doujin probability and the parties and related parties of the reference contract. a contract acquisition unit that acquires information on a standard contract;
an attribute acquisition unit that acquires attribute items of a reference person selected from parties and related parties of the reference contract from the information of the reference contract;
a reference attribute acquisition unit that acquires attribute items of the referrer from information of a reference contract that includes, as a party or a related party, a referrer similar to the reference person whose attribute items partly match those of the reference person;
a calculation unit that calculates the degree of similarity between the attribute items of the reference person and the attribute items of the referrer;
When the similarity between the reference person attribute item and the referrer attribute item is input, the calculated similarity is input to the learning model that outputs the same person probability indicating the probability that the reference person and the referrer are the same person. a probability acquisition unit that acquires the same person probability of the reference person and the reference person that is input and output by the learning model;
an adjustment unit that adjusts the same person probability based on the degree of matching between a party or related party other than the reference person included in the reference contract and a party or related party other than the referrer of the reference contract;
An information processing apparatus, comprising: an output unit that outputs the adjusted same-person probability and the parties and related parties of the reference contract. Acquire the attribute items of the reference person related to the reference contract and the parties concerned related to the reference contract other than the reference person,
Acquiring the attribute items of the referrer related to the reference contract determined to be the same person as the reference person and the related parties related to the reference contract other than the referrer;
calculating the degree of similarity between the attribute items of the reference person and the persons concerned other than the reference person and the attribute items of the reference person and persons concerned other than the reference person;
Based on the training data including the calculated degree of similarity, the attribute items of the reference person related to the reference contract and the parties concerned related to the reference contract other than the reference person , the reference person related to the reference contract, and the reference contract other than the reference person A learning model generation method for generating a learning model that outputs a probability that the reference person and the referrer are the same person when a degree of similarity with an attribute item of a related person is input.

Images