JP6802134B2 - Operator selection device, operator selection system, program and operator selection method - Google Patents

Description

The present invention relates to an operator selection device, an operator selection system, a program, and an operator selection method for assigning operators to incoming customers.

A device that assigns an operator to a customer who receives a call is known. Patent Document 1 discloses a technique for assigning an operator having a low degree of tension among a plurality of operators.

JP-A-2003-244328

When assigning an operator to a customer who receives a call, it is preferable to select the operator so as to satisfy the allocation condition that the customer satisfaction is high and the time required for answering the call is short. However, the ability to correspond to the category to which the incoming call content belongs differs depending on the skill level of the operator, and the fatigue level changes depending on the passage of time and the content of the telephone response. Therefore, the selection of the operator satisfying the allocation condition is not constant. By using conventional techniques, it is possible to select an operator who is not tired or tense. However, the operator who always responds to emotionlessness without considering the customer is preferentially selected, which may deviate from the customer satisfaction.

Therefore, the present invention has been made in view of these points, and is an operator selection device and an operator that assigns operators to customers who receive a call so that the customer satisfaction is high and the time required for answering the call is shortened. It is an object of the present invention to provide a selection system, a program and an operator selection method.

The operator selection device according to the first aspect of the present invention is a machine learning model learned to output the response capability value of the operator regarding the incoming call by inputting the information regarding the incoming call of the customer and the biometric information of the operator. An input information acquisition unit that acquires information about the customer's incoming call and biometric information of a vacant operator who does not correspond to the customer at a timing that satisfies a predetermined acquisition condition, and an input information acquisition unit that acquires the customer. A model application unit that acquires the response capability value when each of the vacant operators corresponds to each of the vacant operators by inputting information on the customer's incoming call and the biometric information of the vacant operator into the machine learning model, and the response capability value. It is provided with an operator allocation unit that assigns the corresponding vacant operator to each of the customers by using the evaluation function determined in use.

The input information acquisition unit may acquire information on the customer's incoming call and biometric information of the vacant operator at a predetermined time and at a timing when the acquisition condition of any of the predetermined number of incoming calls of the customer is satisfied. ..

The information regarding the customer's incoming call may include the category to which the customer's incoming call content belongs.
The information regarding the customer's incoming call may include the customer attribute of the customer.

The operator selection device sets a calculation parameter for calculating the response ability value, and the shorter the response time indicating the time taken by the operator to respond to a call, the higher the response ability value, and the customer receives a call response. The index value of the response ability value, the biometric information of the operator, and the customer, which are determined based on the response ability mapping mapped so that the higher the customer satisfaction level indicating the evaluation of the operator, the higher the response ability value. A model generation unit that generates the machine learning model learned based on the customer response history including the information regarding the incoming call, the customer satisfaction, the biometric information of the operator, and the response time may be further provided. ..

The model generation unit has a large difference between the customer satisfaction and the self-evaluation when the self-evaluation, which is the evaluation of the operator who responded to the telephone by the operator, is higher than the customer satisfaction. The corresponding ability value is corrected so that the index value becomes lower, and the corresponding ability value is determined and corrected based on the corresponding ability mapping mapped so that the higher the customer satisfaction is, the higher the corresponding ability value is. The machine learning model learned based on the index value and the customer response history including the self-evaluation may be generated.

When the personal characteristic of the operator shows a tendency to give a high self-evaluation, the model generation unit corrects the index value to be low, and the personal characteristic of the operator shows a tendency to make the self-evaluation low. In some cases, the index value may be corrected to be high.

The operator selection system according to the second aspect of the present invention acquires the operator selection device and the biometric information of the operator at predetermined time intervals, and obtains the acquired biometric information of the operator at a predetermined timing. It is provided with an operator terminal for transmitting to.

The program according to the third aspect of the present invention is machine learning learned to output the response capability value of the operator regarding the incoming call to the computer by inputting the information regarding the incoming call of the customer and the biometric information of the operator. A step of acquiring a model, a step of acquiring information on the customer's incoming call at a timing satisfying a predetermined acquisition condition, and a step of acquiring biometric information of a vacant operator who does not correspond to a customer at a timing satisfying a predetermined acquisition condition. And the step of inputting the information about the customer's incoming call and the biometric information of the operator into the machine learning model to acquire the response ability value when each vacant operator responds to each incoming customer, and the response ability value. The step of assigning the corresponding free operator to each of the customers is executed by using the evaluation function determined by using.

The operator selection method according to the fourth aspect of the present invention is learned to output the response ability value of the operator regarding the incoming call by inputting the information regarding the incoming call of the customer and the biometric information of the operator executed by the computer. The step of acquiring the obtained machine learning model, the step of acquiring the information regarding the customer's incoming call at the timing when the predetermined acquisition condition is satisfied, and the biometric information of the vacant operator who does not respond to the customer at the timing when the predetermined acquisition condition is satisfied. And the step of inputting the information about the customer's incoming call and the biometric information of the operator into the machine learning model to acquire the response ability value when each vacant operator responds to each incoming customer. It is characterized by having a step of assigning a corresponding free operator to each of the customers by using an evaluation function determined by using the corresponding ability value.

According to the present invention, it is possible to assign an operator to a customer who receives a call so that the customer satisfaction is high and the time required for answering the call is shortened.

It is a figure for demonstrating the outline of the operator selection system. It is a figure for demonstrating the outline of the operator selection system. It is a figure which shows the structure of an operator terminal. It is a figure which shows the structure of the CTI server. It is a figure which shows the structure of the operator selection apparatus. It is a figure which shows the structure of the operator state database. It is a figure which shows the structure of the incoming call information database. It is a figure which shows the structure of the correspondence capacity estimation database. It is a figure which shows the structure of the customer correspondence history database. It is a figure which shows the structure of the correspondence capacity evaluation summary database. It is a figure which shows the structure of the correspondence capacity mapping. It is a sequence diagram which shows the flow of processing by an operator selection system.

[Overview of operator selection system S]
1 and 2 are diagrams for explaining the outline of the operator selection system S. The operator selection system S uses a machine learning model so that the customer satisfaction is high and the time required for telephone response is shortened from among a plurality of vacant operators who do not respond to telephone calls when receiving a call from a customer. It is a system to select. Further, the operator selection system S is a system that updates the machine learning model M based on the information regarding the telephone correspondence performed by the operator.

The operator selection system S includes an operator terminal 2, a CTI (Computer Telephony Integration) server 3, and an operator selection device 4.
The operator terminal 2 is, for example, a PC (Personal Computer). The operator terminal 2 is a terminal used by an operator to answer a telephone call to a customer. The operator terminal 2 has a sensor for acquiring the operator's biological information (for example, an electrocardiogram, a pulse, an autonomic nerve state, etc.).

The CTI server 3 is, for example, a server. The CTI server 3 receives an incoming call from a customer who has made a call to the call center via the base station 1. The CTI server 3 transfers the received customer's telephone call to the operator terminal 2 used by the operator selected by the operator selection device 4.

The operator selection device 4 is, for example, a server. The operator selection device 4 uses the machine learning model M so that the customer satisfaction is high and the time required for telephone response is shortened from among a plurality of vacant operators who do not respond to telephone calls when receiving a call from the customer. Calculate the combination of incoming customers and vacant operators. The machine learning model M is a model trained to output a corresponding ability value when a predetermined information is input. The response ability value is a numerical value indicating the high suitability of the operator for the business of responding to the customer by telephone when receiving a call from the customer. The response ability value takes into consideration the skill level of the operator, the fatigue level at the time of incoming call, the attribute of the incoming customer, the category of the telephone content, and the like.

Customers C1 and C2 shown in FIG. 1 are in a state of making a call to the call center ((1) in FIG. 1), and customer C3 is in a state in which the operator OP4 is in charge of making a call. In this case, when the CTI server 3 receives incoming calls from customers C1 and C2 via the base station 1 ((2) in FIG. 1), the operator selection device 4 receives information regarding incoming calls from customers C1 and C2 (for example,). Send customer attributes including gender and age, telephone content categories, etc.) ((3) in Fig. 1).

When the operator selection device 4 acquires the information regarding the incoming call of the customers C1 and C2, the operator selection device 4 acquires the biometric information of the vacant operators OP1, OP2, and OP3 that do not correspond to the telephone ((4) in FIG. 1). The operator selection device 4 uses the machine learning model M corresponding to each operator to input information on incoming calls of customers C1 and C2 and biometric information of the operator corresponding to the machine learning model M, and for each of the customers C1 and C2. Acquire the corresponding ability values of the operators OP1, OP2, and OP3 ((5) in FIG. 1).

The operator selection device 4 uses an evaluation function from 6 patterns in which empty operators OP1, OP2, OP3 and customers C1 and C2 are combined based on the acquired response ability value, and has high customer satisfaction and telephone response. A combination is calculated so that the time required for the above is shortened ((6) in FIG. 1). Specifically, the operator selection device 4 uses an evaluation function to calculate the combination that maximizes the total value of the corresponding ability values. For example, when the operator selection device 4 calculates the combination of the vacant operator OP1 for the customer C1 and the vacant operator OP2 for the customer C2, which maximizes the total value of the corresponding ability values, the combination information indicating the combination is transmitted to the CTI server 3. It is transmitted ((7) in FIG. 1).

Based on the combination information transmitted from the operator selection device 4, the CTI server 3 transfers the customer C1's telephone to the empty operator OP1 shown in the combination information, and transfers the customer C2's telephone to the operator OP2 ((FIG. 1). 8)). The vacant operators OP1 and OP2 to which the telephone is transferred start telephone correspondence to the customers C1 and C2, respectively, via the CTI server 3 ((9) in FIG. 1).

When the operators OP1, OP2, and OP3 finish the telephone response, the CTI server 3 switches to the automatic voice requesting the answer to the questionnaire. Each of the customers C1, C2, and C3 responds to the evaluation of each of the operators OP1, OP2, and OP3 corresponding to the telephone by a questionnaire according to the automatic voice ((1) in FIG. 2). The CTI server 3 acquires customer satisfaction from the questionnaire results answered by customers C1, C2, and C3 ((2) in FIG. 2). The CTI server 3 transmits the correspondence information associated with the operator and the customer to the operator selection device 4 at the customer satisfaction level and the response time required for the telephone response ((3) in FIG. 2).

The operators OP1, OP2, and OP3 input self-evaluation for telephone correspondence to the operator terminal 2, respectively. The operator terminals 2a, 2b, and 2c transmit the self-evaluation input by each of the operators OP1, OP2, and OP3 to the operator selection device 4 ((4) in FIG. 2). The operator selection device 4 updates the customer correspondence history based on the correspondence information and the self-evaluation ((5) in FIG. 2). After that, the operator terminals 2a, 2b, and 2c transmit the biometric information of the operators OP1, OP2, and OP3 acquired at predetermined time intervals to the operator selection device 4 ((6) in FIG. 2). Then, the operator selection device 4 generates a machine learning model M learned based on the updated customer correspondence history and the biometric information of the operators OP1, OP2, OP3 transmitted from the operator terminals 2a, 2b, and 2c (FIG. 2). (7)).

As described above, the operator selection system S calculates the combination of the incoming customer and the vacant operator based on the response ability value output from the machine learning model M learned based on the customer response history and the operator's biometric information. As a result, the operator can be assigned so that the customer satisfaction is high and the time required for answering the telephone is shortened.
Hereinafter, the details of the operator terminal 2, the CTI server 3, and the operator selection device 4 will be described.

[Configuration of operator terminal 2]
FIG. 3 is a diagram showing the configuration of the operator terminal 2. The operator terminal 2 includes a headset 20, a communication unit 21, a sensor unit 22, a storage unit 23, and a control unit 24.
The headset 20 has a microphone for the operator to input voice and a speaker for outputting customer's voice.

The communication unit 21 has a role as a telephone exchange for talking with a customer via the base station 1. In addition, the communication unit 21 includes a communication controller for transmitting and receiving data to and from the CTI server 3 and the operator selection device 4.

The sensor unit 22 measures the operator's electrocardiographic waveform or heart rate from a sensor attached to a device worn by the operator (for example, a wearable or a headset).
The storage unit 23 is, for example, a storage medium such as a ROM (Read Only Memory), a RAM (Random Access Memory), and a hard disk. The storage unit 23 stores various programs executed by the control unit 24.

The control unit 24 is, for example, a CPU (Central Processing Unit). The control unit 24 controls the function related to the operator terminal 2 by executing the program stored in the storage unit 23. By executing the program, the control unit 24 functions as a biometric information acquisition unit 241, a call unit 242, a self-evaluation input unit 243, and a transmission unit 244.

The biometric information acquisition unit 241 acquires the biometric information of the operator based on the electrocardiographic waveform or the heart rate measured by the sensor unit 22 at predetermined time intervals. Further, the biometric information acquisition unit 241 obtains the biometric information of the operator based on the electrocardiographic waveform or heart rate measured by the sensor unit 22 when the biometric information acquisition instruction is notified from the communication unit 242 or the operator selection device 4. get. The predetermined time may be set for each operator according to the total time of telephone response or the number of telephone responses, or may be set according to the characteristics of the operator (for example, easy to get nervous and easy to get tired). The biological information is, for example, an electrocardiogram or a state of the autonomic nerve (for example, whether or not the autonomic nerve is out of balance) obtained by a known method from the electrocardiogram.

The biological information acquisition unit 241 associates the acquired biological information of the operator with the operator's identification number and stores it in the storage unit 23. The operator identification number may be obtained from the account information logged in to the operator terminal 2. When the biometric information acquisition unit 241 is notified of the biometric information acquisition instruction from the communication unit 242 or the operator selection device 4, the biometric information acquisition unit 241 inputs the acquired biometric information of the operator to the transmission unit 244.

The call unit 242 receives the call transferred from the CTI server 3 via the communication unit 21 and controls the headset 20 so that the operator can make a voice call with the incoming customer. After the call is completed, the call unit 242 notifies the biometric information acquisition unit 241 of a biometric information acquisition instruction for acquiring the biometric information of the operator. When the call ends, the call unit 242 notifies the CTI server 3 that the call has ended via the self-evaluation input unit 243 and the communication unit 21.

When the self-evaluation input unit 243 receives a call end notification from the call unit 242, the self-evaluation input unit 243 displays a self-evaluation input screen for subjectively evaluating whether or not the operator has successfully responded this time on a display unit (not shown). .. On the self-evaluation input screen, for example, on a 5-point scale, "I couldn't do it well", "There were some things I couldn't do well", "Normal", "Somewhat worked", and "Very worked". Includes the options shown. When the operator inputs any of the options, the self-evaluation input unit 243 inputs the input information to the transmission unit 244.

The transmission unit 244 transmits the self-evaluation input from the self-evaluation input unit 243 to the operator selection device 4 via the communication unit 21. The transmission unit 244 transmits the biometric information of the operator acquired by the biometric information acquisition unit 241 to the operator selection device 4 at a predetermined timing via the communication unit 21. For example, in order for the operator selection device 4 to use the biometric information for learning, the transmission unit 244 transmits the biometric information stored in the storage unit 23 to the operator selection device 4 at the timing when the call center business is completed. When the biometric information is input from the biometric information acquisition unit 241, the transmission unit 244 transmits the input biometric information to the operator selection device 4 via the communication unit 21.

[Configuration of CTI server 3]
FIG. 4 is a diagram showing the configuration of the CTI server 3. The CTI server 3 includes a communication unit 31, a storage unit 32, and a control unit 33.
The communication unit 31 has a role as a telephone exchange for receiving a customer's telephone call to the call center via the base station 1 and transferring the received telephone call to the operator terminal 2. Further, the communication unit 31 includes a communication controller for transmitting and receiving data to and from the operator selection device 4.

The storage unit 32 is, for example, a storage medium such as a ROM, RAM, and a hard disk. The storage unit 32 stores the customer information database. The customer information database stores the customer number, the customer's name, the customer's gender, the customer's age, the customer's telephone number, and the like in association with each other. The storage unit 32 stores various programs executed by the control unit 33.

The control unit 33 is, for example, a CPU, and controls the functions related to the CTI server 3 by executing the program stored in the storage unit 32. By executing the program, the control unit 33 functions as an incoming call information acquisition unit 331, an allocation information acquisition unit 332, a transfer unit 333, and a corresponding information acquisition unit 334.

The incoming call information acquisition unit 331 acquires information regarding the incoming call of the customer. The information regarding the customer's incoming call is the customer attribute including the customer's gender and the customer's age, and the information input by the customer according to the automatic voice guidance (for example, the category to which the customer's incoming call content belongs). The incoming call information acquisition unit 331 acquires the customer attribute from the customer information database based on the caller ID obtained by the incoming call. The incoming call information acquisition unit 331 inputs the acquired information regarding the incoming call of the customer to the allocation information acquisition unit 332. If the telephone number is a new customer that does not exist in the customer database, the incoming call information acquisition unit 331 inputs only the category to the allocation information acquisition unit 332.

When the allocation information acquisition unit 332 receives information regarding the customer's incoming call from the incoming call information acquisition unit 331, the allocation information acquisition unit 332 transmits the information regarding the customer's incoming call to the operator selection device 4 via the communication unit 31. The allocation information acquisition unit 332 acquires the combination information generated by the operator selection device 4 via the communication unit 31 after transmitting the information regarding the customer's incoming call to the operator selection device 4. The combination information includes information for identifying an operator to be assigned to the incoming customer. The allocation information acquisition unit 332 inputs the acquired combination information to the transfer unit 333.

The transfer unit 333 transfers the incoming customer's call to the vacant operator based on the combination information input from the allocation information acquisition unit 332. Specifically, the transfer unit 333 transfers the incoming customer's telephone to the free operator indicated in the combination information.

The correspondence information acquisition unit 334 acquires information on the telephone correspondence after the operator finishes the telephone correspondence. Specifically, when the response information acquisition unit 334 receives the notification of the end of the call from the operator terminal 2, the response information acquisition unit 334 acquires the response time of the telephone response and the customer satisfaction level based on the questionnaire result input by the customer. For example, the response information acquisition unit 334 automatically guides the choices of "dissatisfied", "somewhat dissatisfied", "normal", "almost satisfied", and "satisfied" on a five-point scale for this response, and the customer automatically guides them. Get the choices you answered according to the voice. The correspondence information acquisition unit 334 transmits the correspondence information in which the operator's identification number and the customer number are associated with the acquired telephone response time and customer satisfaction to the operator selection device 4 via the communication unit 31.

[Configuration of operator selection device 4]
FIG. 5 is a diagram showing the configuration of the operator selection device 4. The operator selection device 4 includes a communication unit 41, a storage unit 42, and a control unit 43.
The communication unit 41 includes a communication controller for transmitting and receiving data to and from the operator terminal 2 and the CTI server 3.

The storage unit 42 is a storage medium such as a ROM, a RAM, or a hard disk. The storage unit 42 stores a customer response history database, an operator status database, an incoming call information database, a response capability estimation database, a response capability evaluation summary database, and a response capability mapping. The configuration of each database and mapping will be described later. Further, the storage unit 42 stores a plurality of machine learning models M associated with each operator. The storage unit 42 stores various programs executed by the control unit 43.

The control unit 43 is, for example, a CPU, and controls a function related to the operator selection device 4 by executing a program stored in the storage unit 42. By executing the program, the control unit 43 functions as an input information acquisition unit 431, a model acquisition unit 432, a model application unit 433, an operator allocation unit 434, a model generation information acquisition unit 435, and a model generation unit 436.

The input information acquisition unit 431 acquires the information to be input in order to output the corresponding ability value from the machine learning model M. Specifically, the input information acquisition unit 431 acquires the information regarding the customer's incoming call and the biometric information of the vacant operator who does not correspond to the customer at the timing when a predetermined acquisition condition is satisfied. More specifically, the input information acquisition unit 431 acquires the customer's incoming call information and the biometric information of the vacant operator at a timing that satisfies any of the acquisition conditions of the predetermined time and the predetermined customer's incoming call number. ..

Here, the predetermined time is, for example, a period from the transmission of information regarding the customer's incoming call from the CTI server 3 to the elapse of a preset time (for example, 1 minute). The predetermined number of incoming calls by the customer is, for example, the number from when the information regarding the incoming call of the customer is transmitted from the CTI server 3 to the number of incoming calls (for example, 10 calls) set in advance.

Information about a customer's incoming call includes categories and customer attributes of the customer. Customer attributes are, for example, the gender and age of the customer. By including the customer attribute in the customer's incoming call information in this way, the input information acquisition unit 431, for example, wants a young man to answer accurately in a short time, and an elderly woman to listen to the story properly and understand it easily. It is possible to output from the machine learning model M the response ability value of each operator considering that the required contents differ depending on the gender and age, such as requesting an explanation even if it takes a certain amount of time. Further, by including the category in the customer's incoming call information, the input information acquisition unit 431 can output the response ability value considering the skill level different for each operator from the machine learning model M, and in the education of the operator. It becomes easy to identify areas of weakness.

The input information acquisition unit 431 identifies a free operator by referring to the operator status database. FIG. 6 is a diagram showing the configuration of the operator status database. The operator status database stores the identification number of the operator and the correspondence status indicating whether or not the operator is responding to the telephone in association with each other. When the operator is responding to the telephone, the customer number is stored in the response status, and when the operator is not responding to the telephone, "0" is stored. In the example shown in FIG. 6, the input information acquisition unit 431 sets OP01, OP03, and OP01, OP03, and OP01, OP03, and OP01 as the identification numbers of the operators whose correspondence status is "0" indicating that the operator is a free operator who does not respond to the telephone from the operator status database. Acquire OP04.

When the input information acquisition unit 431 acquires the operator's identification number from the operator status database, the input information acquisition unit 431 acquires the operator's biometric information from the operator terminal 2 used by the empty operator of the identification number. Specifically, the input information acquisition unit 431 notifies the operator terminal 2 used by the vacant operator of the biometric information acquisition instruction via the communication unit 41, and acquires the biometric information of the operator from the notified operator terminal 2.

The input information acquisition unit 431 inputs the acquired biometric information of the operator to the model application unit 433. Further, the input information acquisition unit 431 inputs the acquired operator identification number to the model acquisition unit 432. In this way, the input information acquisition unit 431 assigns a vacant operator who does not respond to telephone calls, so that the selection is not biased only to a specific operator who has experience and high skill, and the overall vacancy status and experience. It is possible to make selections in consideration of balance, such as assigning them exceptionally to operators with a small amount.

The input information acquisition unit 431 stores the categories and customer attributes transmitted from the CTI server 3 in the incoming call information database. FIG. 7 is a diagram showing the configuration of the incoming call information database. The incoming call information database stores the customer number, the customer's gender, the customer's age, and the category in association with each other. As shown in FIG. 7, the input information acquisition unit 431 stores the category and the customer attribute transmitted from the CTI server 3 in the incoming call information database in association with the customer number.

The model acquisition unit 432 acquires the machine learning model M learned to output the response ability value regarding the operator's incoming call by inputting the information regarding the customer's incoming call and the operator's biometric information. Specifically, the model acquisition unit 432 acquires the machine learning model M corresponding to the operator's identification number input from the input information acquisition unit 431 from the storage unit 42. The machine learning model M is a model learned for each operator using, for example, a support vector machine (SVM). The model acquisition unit 432 may acquire the machine learning model M of each operator in advance.

The model application unit 433 acquires the response capability value when each vacant operator responds to each incoming customer by inputting the information regarding the customer's incoming call and the biometric information of the vacant operator into the machine learning model M. Specifically, the model application unit 433 converts the information stored in the incoming call information database and the biometric information of the vacant operator input from the input information acquisition unit 431 into the machine learning model M for each combination of the customer and the vacant operator. Enter and acquire the response capability value when a vacant operator responds to the incoming customer.

The model application unit 433 stores the acquired corresponding ability value in the corresponding ability estimation database. FIG. 8 is a diagram showing the configuration of a response capacity estimation database. The response capability estimation database stores the operator's identification number, the customer number of the incoming customer, and the response capability of the operator in association with each other. As shown in FIG. 8, the model application unit 433 stores the acquired response capability value for each combination in the response capability estimation database in association with the operator identification number and the customer number.

The operator allocation unit 434 allocates a corresponding free operator for each customer by using an evaluation function determined by using the corresponding ability value. The operator allocation unit 434 refers to, for example, a response capacity estimation database, and calculates a combination that maximizes the response capacity value when each of a plurality of customers who receive a call is assigned to each vacant operator (for example, the Hungarian method).

In the example shown in FIG. 8, the operator allocation unit 434 has six ways in which the three operators identified by OP01, OP03, and OP04 correspond to the three customers identified by C01, C02, and C03. From the combinations of, calculate the combination that maximizes the total value of the corresponding ability. In this case, the operator allocation unit 434 uses the customer with the customer number "C02" and the identification number "OP03" as the operator with the identification number "OP01", which has the maximum total value of the corresponding ability of "1.6". A combination of a customer having the customer number "C01" for a certain operator and a customer having the customer number "C03" for the operator having the identification number "OP04" is calculated. By doing so, the operator allocation unit 434 can calculate a combination in which the customer satisfaction is high as a whole and the response time is minimized. The operator allocation unit 434 transmits the calculated combination information to the CTI server 3 via the communication unit 41.

The operator allocation unit 434 updates the correspondence status of the operator status database based on the calculated combination. Based on the above combination, for example, the operator allocation unit 434 identifies the correspondence status corresponding to the identification number "OP01" of the operator status database as "C02" and the correspondence status corresponding to the identification number "OP03" as "C01". The correspondence status corresponding to the number "OP04" is updated to "C03".

The model generation information acquisition unit 435 acquires model generation information for generating the machine learning model M. Specifically, the model generation information acquisition unit 435 acquires the self-evaluation from the operator terminal 2 and acquires the response time and customer satisfaction from the CTI server 3. In addition, the model generation information acquisition unit 435 acquires the biometric information of the operator from the operator terminal 2 at predetermined time intervals and before and after the telephone response. The model generation information acquisition unit 435 updates the customer correspondence history database stored in the storage unit 42 based on the acquired self-evaluation, response time, and customer satisfaction.

FIG. 9 is a diagram showing the configuration of the customer correspondence history database. The customer response history database stores the operator's identification number, category, response time, customer satisfaction, self-evaluation, and the total number of telephone responses in association with each other. The response time stored in the customer response history database is the time obtained by converting the total value of the time required for telephone response into a decimal number. The customer satisfaction stored in the customer response history database is an average value obtained by dividing the total value of customer satisfaction by the total number of telephone responses. The self-evaluation stored in the customer response history database is an average value obtained by dividing the total value of the self-evaluation by the total number of telephone responses. As shown in FIG. 9, the model generation information acquisition unit 435 corresponds to the operator's identification number and category in the customer response history database based on the acquired self-evaluation, response time, and customer satisfaction, and respond time and customer satisfaction. Update the degree, self-evaluation, and number of responses.

When the model generation information acquisition unit 435 acquires the self-evaluation, the model generation information acquisition unit 435 updates the correspondence status corresponding to the operator who performed the self-evaluation in the operator status database. Specifically, the model generation information acquisition unit 435 updates the correspondence status corresponding to the identification number of the operator using the operator terminal 2 that has acquired the self-evaluation to "0".

The model generation information acquisition unit 435 inputs the operator's biometric information acquired at predetermined time intervals and the operator's biometric information acquired before and after the telephone response to the model generation unit 436. By doing so, the model generation information acquisition unit 435 uses the biometric information for each predetermined time, the amount of change in the degree of fatigue required from the biometric information before and after the telephone response, and the customer satisfaction into the machine learning model M. The machine learning model M can output a response ability value linked to customer satisfaction from the degree of fatigue obtained based on the operator's biological information at the time of incoming call. As a result, it is possible to estimate the fatigue tolerance of each operator, such as an operator who has a high degree of fatigue but a high degree of customer satisfaction, or an operator who has a low degree of fatigue but a low degree of customer satisfaction. And can educate operators.

The model generation unit 436 is a machine that learns the calculation parameters for calculating the response ability value based on the index value of the response ability value determined based on the response ability mapping, the biometric information of the operator, and the customer response history database. Generate a learning model M. As shown in FIG. 9, the customer interaction history database includes information about the incoming call the customer (e.g., category), and customer satisfaction, and between pairs latency time, the.

FIG. 10 is a diagram showing the configuration of a response capability evaluation summary database. The response capability evaluation summary database stores an evaluation summary for calculating the index value of the response capability by the response capability mapping. The response capability evaluation summary database stores the response time, which indicates whether or not it took a long time to respond to a call based on a predetermined standard value, the evaluation shown in five stages, and the evaluation summary for calculating the index value. ing. In the evaluation summary, for example, "◎" indicates the highest evaluation and "x" indicates the lowest evaluation.

FIG. 11 is a diagram showing a configuration of response capability mapping. The response capability mapping stores information to be trained by the machine learning model M. Based on the customer response history database, the response capability mapping indicates the time taken by the operator to answer the call. The shorter the response time, the higher the response capability value, and the customer satisfaction that indicates the evaluation of the operator who received the call. It is mapped so that the higher the degree, the higher the response ability value. Response capability mapping maps the response capability index values to response time and evaluation.

The index value is a measure of the response capacity quantified between 0 <index value ≤ 1 based on the response capacity evaluation summary database. The index value may be calculated by the model generation unit 436 or may be set by the manager of the call center. By associating the response time with the response ability value in this way, the model generation unit 436 can output the response ability value in consideration of the balance between the customer satisfaction and the response time from the machine learning model M. Further, the model generation unit 436 can change whether to emphasize the response time or the customer satisfaction for each company, each department, or each category by adjusting the response capability mapping.

In the model generation unit 436, when the self-evaluation, which is the evaluation of the operator who responded to the telephone by the operator, is higher than the customer satisfaction, the larger the difference between the customer satisfaction and the self-evaluation, the lower the index value. It may be corrected so as to be. For example, when the self-evaluation is "5" and the customer satisfaction is "2", the model generation unit 436 is an index of the ability mapping as compared with the case where the self-evaluation and the customer satisfaction match. It may be corrected so that the value becomes low.

In this case, the model generation unit 436 determines the index value of the responsiveness value, which is determined based on the responsiveness mapping mapped so that the higher the customer satisfaction is, the higher the responsiveness value, and the self-evaluation. Further, a machine learning model learned based on the customer correspondence history database including the database may be generated. By comparing the difference between the customer satisfaction and the operator's self-evaluation in this way, the model generation unit 436 can output the corresponding ability value in consideration of the operator's communication ability from the machine learning model M. In addition, it can be used for education to improve communication ability to grasp the feelings of the other party at the time of operator education, and it can be effective in resolving dissatisfaction on the operator side by presenting it as an objective evaluation at the time of personnel evaluation. ..

The model generation unit 436 corrects the index value so that it becomes low when the individual characteristic of the operator shows a tendency to give a high self-evaluation, and the index value when the individual characteristic of the operator shows a tendency to make a low self-evaluation. May be corrected so that The model generation unit 436 uses, for example, a subjective questionnaire (for example, the personality trait Big5 scale) that periodically acquires personal traits from the operator, and for an operator with a high tendency to extrovert, which tends to enhance self-evaluation. The index value may be corrected to be low, and the index value may be corrected to be high for an operator with a nervous tendency who tends to give a low self-evaluation. By doing so, the model generation unit 436 can output the corresponding ability value considering the character of the operator from the machine learning model M.

[Processing of operator selection system S]
Subsequently, the flow of processing by the operator selection system S will be described. FIG. 12 is a sequence diagram showing a processing flow by the operator selection system S. This sequence diagram starts when the customer calls the call center.

When the incoming call information acquisition unit 331 of the CTI server 3 receives an incoming call from the customer, the incoming call information acquisition unit 331 acquires the information regarding the incoming call of the customer. Specifically, the incoming call information acquisition unit 331 acquires customer attributes including the customer's gender and customer's age from the customer information database based on the caller ID obtained by the incoming call, and the incoming call input by the customer according to the automatic voice. Get the category to which the content belongs. The incoming call information acquisition unit 331 inputs the acquired customer attributes and categories to the allocation information acquisition unit 332 of the CTI server 3. When the customer attribute and category are input from the incoming call information acquisition unit 331, the allocation information acquisition unit 332 transmits the customer attribute and category to the operator selection device 4 via the communication unit 31.

When the input information acquisition unit 431 of the operator selection device 4 acquires the customer attributes and categories from the CTI server 3, it determines whether or not a predetermined acquisition condition is satisfied (S1). Specifically, the input information acquisition unit 431 determines whether or not the acquisition condition of any of the predetermined time and the predetermined number of incoming calls of the customer is satisfied. When the input information acquisition unit 431 determines that the predetermined acquisition condition is not satisfied, the process returns to S1.

On the other hand, when the input information acquisition unit 431 determines that the predetermined acquisition condition is satisfied, it identifies a vacant operator who is not responding to the telephone (S2). Specifically, the input information acquisition unit 431 acquires the identification number of the operator whose correspondence status is "0", which indicates that the operator is a free operator who does not respond to the telephone from the operator status database.

When the input information acquisition unit 431 acquires the operator's identification number from the operator status database, the input information acquisition unit 431 acquires the operator's biometric information from the operator terminal 2 used by the empty operator of the identification number. Specifically, first, the input information acquisition unit 431 notifies the operator terminal 2 used by the vacant operator of the biometric information acquisition instruction via the communication unit 41. When the biometric information acquisition unit 241 of the operator terminal 2 receives the biometric information acquisition instruction from the operator selection device 4, the biometric information acquisition unit 241 acquires the biometric information of the operator based on the electrocardiographic waveform or heart rate measured by the sensor unit 22. The transmission unit 244 of the operator terminal 2 transmits the biometric information of the operator acquired by the biometric information acquisition unit 241 to the operator selection device 4. Then, the input information acquisition unit 431 acquires the biometric information of the operator transmitted from the operator terminal 2.

The input information acquisition unit 431 inputs the acquired biometric information of the operator to the model application unit 433. The input information acquisition unit 431 stores the customer attributes and categories transmitted from the CTI server 3 in the incoming call information database. The model acquisition unit 432 acquires the machine learning model M corresponding to each of the operator identification numbers input from the input information acquisition unit 431 from the storage unit 42.

Subsequently, the model application unit 433 inputs the biometric information of the operator input from the input information acquisition unit 431 and the information stored in the incoming call information database into the machine learning model M, and combines the incoming customer and the vacant operator. For each case, the response capability value when a vacant operator responds to the incoming customer is acquired (S3), and the acquired response capability value is stored in the response capability estimation database.

The operator allocation unit 434 allocates a corresponding free operator for each customer by using an evaluation function determined by using the corresponding ability value. For example, the operator allocation unit 434 refers to the response capacity estimation database, and calculates the combination that maximizes the response capacity value when each customer who receives a call is assigned to each vacant operator based on the Hungarian law (S4). The operator allocation unit 434 transmits the calculated combination information to the CTI server 3 via the communication unit 41. In addition, the operator allocation unit 434 updates the operator status database based on the calculated combination.

The transfer unit 333 of the CTI server 3 transfers incoming calls from a plurality of customers to each of the plurality of vacant operators based on the input combination information via the communication unit 31. Then, each of the transferred operators starts telephone correspondence. The correspondence information acquisition unit 334 of the CTI server 3 acquires the information regarding the telephone correspondence after the operator finishes the telephone correspondence. Specifically, when the response information acquisition unit 334 receives the notification of the end of the call from the operator terminal 2, it acquires information on the telephone response including the response time and the customer satisfaction level based on the questionnaire result input by the customer. The correspondence information acquisition unit 334 transmits the acquired information regarding the telephone correspondence to the operator selection device 4 via the communication unit 31.

When the self-evaluation input unit 243 of the operator terminal 2 receives the call end notification from the call unit 242, the self-evaluation input unit 243 displays the self-evaluation input screen. When the operator inputs one of the five-level evaluation options from the self-evaluation input screen, the self-evaluation input unit 243 inputs the input self-evaluation to the transmission unit 244. The transmission unit 244 of the operator terminal 2 transmits the self-evaluation input from the self-evaluation input unit 243 to the operator selection device 4 via the communication unit 21.

The model generation information acquisition unit 435 of the operator selection device 4 acquires model generation information for generating the machine learning model M. Specifically, the model generation information acquisition unit 435 acquires the self-evaluation from the operator terminal 2 and acquires the response time and customer satisfaction from the CTI server 3.

The model generation information acquisition unit 435 updates the customer correspondence history database stored in the storage unit 42 based on the acquired model generation information (S5). The model generation information acquisition unit 435 inputs the biometric information of the operator at predetermined time intervals and before and after the telephone response to the model generation unit 436. Specifically, the model generation information acquisition unit 435 acquires the biometric information of the operator transmitted at a predetermined timing by the transmission unit 244 of the operator terminal 2 and inputs it to the model generation information 436. The model generation unit 436 is a machine that learns the calculation parameters for calculating the response ability value based on the index value of the response ability value determined based on the response ability mapping, the biometric information of the operator, and the customer response history database. A learning model M is generated (S6).

[Effects in this embodiment]
As described above, the operator selection system S according to the present embodiment calculates the combination of the incoming customer and the vacant operator based on the corresponding ability value of the operator output by the machine learning model M. Then, the operator selection system S assigns a customer to the operator based on the calculated combination. By doing so, the operator selection system S can be assigned in the optimum combination with high customer satisfaction and short response time. As a result, the operator selection system S can assign the optimum operator to the incoming customer.

Although the present invention has been described above using the embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments, and various modifications and changes can be made within the scope of the gist. is there. For example, the specific embodiment of the distribution / integration of the device is not limited to the above embodiment, and all or a part thereof may be functionally or physically dispersed / integrated in any unit. Can be done. Also included in the embodiments of the present invention are new embodiments resulting from any combination of the plurality of embodiments. The effect of the new embodiment produced by the combination has the effect of the original embodiment together.

1 Base station 2 Operator terminal 20 Headset 21 Communication unit 22 Sensor unit 23 Storage unit 24 Control unit 241 Biometric information acquisition unit 242 Call unit 243 Self-evaluation input unit 244 Transmission unit 3 CTI server 31 Communication unit 32 Storage unit 33 Control unit 331 Incoming call information acquisition unit 332 Allocation information acquisition unit 333 Transfer unit 334 Corresponding information acquisition unit 4 Operator selection device 41 Communication unit 42 Storage unit 43 Control unit 431 Input information acquisition unit 432 Model acquisition unit 433 Model application unit 434 Operator allocation unit 435 Model generation Information acquisition unit 436 Model generation unit

Claims (10)

A model acquisition unit that acquires a machine learning model that has been trained to output the response capability value of the operator regarding the incoming call by inputting the customer's incoming call information and the operator's biological information.
An input information acquisition unit that acquires information on the customer's incoming call and biometric information of a vacant operator who is not responding to the customer at the timing when a predetermined acquisition condition is satisfied.
A model application unit that acquires the response capability value when each of the vacant operators responds to each of the incoming customers by inputting information on the customer's incoming call and biometric information of the vacant operator into the machine learning model.
An operator allocation unit that allocates the corresponding vacant operator for each of the customers by using the evaluation function determined by using the corresponding ability value.
An operator selection device comprising. The input information acquisition unit acquires information on the customer's incoming call and biometric information of the vacant operator at a predetermined time and at a timing that satisfies the acquisition condition of any of the predetermined number of incoming calls of the customer.
The operator selection device according to claim 1. The information regarding the customer's incoming call includes the category to which the customer's incoming call content belongs.
The operator selection device according to claim 1 or 2. The information regarding the customer's incoming call includes the customer attributes of the customer.
The operator selection device according to any one of claims 1 to 3. As for the calculation parameter for calculating the response ability value, the shorter the response time indicating the time taken by the operator to respond to the call, the higher the response ability value, and the evaluation of the operator who received the call from the customer. The higher the customer satisfaction level, the higher the response ability value, and the index value of the response ability value, the operator's biometric information, and the information regarding the customer's incoming call, which are determined based on the response ability mapping. further comprising a customer satisfaction, and before Symbol response time, the model generating unit for generating the machine learning model trained based on the customer interaction history including,
The operator selection device according to any one of claims 1 to 4. The model generation unit has a large difference between the customer satisfaction and the self-evaluation when the self-evaluation, which is the evaluation of the operator who responded to the telephone by the operator, is higher than the customer satisfaction. The corresponding ability value is corrected so that the index value becomes lower, and the corresponding ability value is determined and corrected based on the corresponding ability mapping mapped so that the higher the customer satisfaction is, the higher the corresponding ability value is. Generate the machine learning model learned based on the customer response history including the index value and the self-evaluation.
The operator selection device according to claim 5. When the personal characteristic of the operator shows a tendency to give a high self-evaluation, the model generation unit corrects the index value to be low, and the personal characteristic of the operator shows a tendency to make the self-evaluation low. In some cases, the index value is corrected so as to be higher.
The operator selection device according to claim 6. The operator selection device according to any one of claims 1 to 7.
An operator terminal that acquires the biometric information of the operator at predetermined time intervals and transmits the acquired biometric information of the operator to the operator selection device at a predetermined timing.
Operator selection system with. On the computer
A step of acquiring a machine learning model trained to output a response ability value of the operator regarding the incoming call by inputting information about the customer's incoming call and the biometric information of the operator.
A step of acquiring information on the customer's incoming call at a timing that satisfies a predetermined acquisition condition, and
A step to acquire biometric information of a vacant operator who does not respond to customers at the timing when a predetermined acquisition condition is satisfied, and
A step of inputting information on the customer's incoming call and biometric information of the operator into the machine learning model to acquire the response ability value when each vacant operator responds to each incoming customer.
A step of assigning the corresponding free operator to each of the customers using the evaluation function determined by using the corresponding ability value, and
A program to execute. Computer runs,
A step of acquiring a machine learning model trained to output a response ability value of the operator regarding the incoming call by inputting information about the customer's incoming call and the biometric information of the operator.
A step of acquiring information on the customer's incoming call at a timing that satisfies a predetermined acquisition condition, and
A step to acquire biometric information of a vacant operator who does not respond to customers at the timing when a predetermined acquisition condition is satisfied, and
A step of inputting information on the customer's incoming call and biometric information of the operator into the machine learning model to acquire the response ability value when each vacant operator responds to each incoming customer.
A step of assigning the corresponding free operator to each of the customers using the evaluation function determined by using the corresponding ability value, and
An operator selection method characterized by having.

Images