Call center program and call center device

Abstract

To provide a call center program and a call center device, which do not cause operational problems due to a lack of an operator even if working population decreases.SOLUTION: In a step 201, user information is received. In a step 203, inquiry information is received from a user terminal 2. In a step 204, inquiry information is transmitted to a generative artificial brain 4. In a step 206, response information from the generative artificial brain 4 is received. In a step 207, response information from the generative artificial brain 4 is inputted and a corrected portion of inappropriate response information is estimated and extracted by using a learned response information portion learning model with an inappropriate response information portion as input teacher data and its corrected response information portion as output teacher data. When the inappropriate corrected response information portion is extracted, a system advances to a step 208, and the inappropriate response information portion is rewritten into the corrected response information portion. In a step 209, response information or response information including the corrected response information portion is transmitted to the user terminal 2.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a program for a call center (also called a contact center) and an apparatus for a call center. [Background technology]

[0002] In a conventional call center system, a user sends inquiry information to a call center by voice over the telephone, and in response, an operator at the call center creates response information and sends it to the user by voice over the telephone. Summary of the Invention [Problem to be solved by the invention]

[0003] However, in the conventional call centers described above, a shortage of operators is expected due to the recent decline in the working population, and this poses the problem of disruption to call center operations in the future.

[0004] Large-scale language models (also called generative artificial intelligence (AI) or interactive artificial intelligence (AI)) are well known. That is, generative AI is a supervised machine learning model in which a large amount of inquiry information is used as input training data and response information to the inquiry is used as output training data. When it receives inquiry information consisting of text from a user, it basically provides response information to the user. [Means for solving the problem]

[0005] In order to solve the above-mentioned problems, the call center program of the present invention is configured to cause a computer to execute an inquiry receiving procedure for receiving inquiry information from a user terminal, an inquiry information transmitting procedure for transmitting the inquiry information to a large-scale language model, a response information receiving procedure for receiving response information to the inquiry information from the large-scale language model, and a response information transmitting procedure for transmitting the response information to the user terminal.

[0006] In addition, the call center device of the present invention comprises an inquiry receiving means for receiving inquiry information from a user terminal, an inquiry information transmitting means for transmitting the inquiry information to a large-scale language model, a response information receiving means for receiving response information to the inquiry information from the large-scale language model, and a response information transmitting means for transmitting the response information to the user terminal. Effect of the Invention

[0007] According to the present invention, since a large-scale language model is used, even if the working population decreases in the future, there will be no disruption to business operations due to a shortage of operators. [Brief description of the drawings]

[0008] [Figure 1] 1 is a block circuit diagram showing an embodiment of a call center device according to the present invention; [Diagram 2] 4 is a flowchart for explaining the operation of a central processing unit of the call center device of FIG. [Diagram 3] FIG. 2 is a block diagram showing another embodiment of the call center device of FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] FIG. 1 is a block circuit diagram showing an embodiment of a call center device according to the present invention.

[0010] In FIG. 1, the call center device 1 is configured by a computer, and in detail, is configured by a central processing unit (CPU) 11, a read only memory (ROM) 12 for storing programs etc., a flash memory 13, a random access memory (RAM) 14 for storing temporary data etc., an input / output (I / O) interface 15 having a USB (registered trademark) terminal, an interrupt terminal INT etc., a voice recognition unit 16 connected to a user terminal (a telephone in this case) 3 via a telephone network, a voice conversion unit 17 connected to the user terminal (telephone) 3 via the telephone network, a communication interface 18 connected to the Internet 3, a history memory 19 consisting of, for example, a magnetic disk unit, a response information partial learning control model 20, etc.

[0011] The voice signal from the user terminal 2 is converted into voice text by the voice recognition unit 16. The voice recognition unit 16 can also be configured as a neural network that performs deep learning using a large number of voice signals (voice data) as input training data and the corresponding correct voice text as output training data. The neural network learns by adjusting the weights and biases of the multilayer perceptron by the backpropagation method from the input layer, multiple intermediate layers, and output layer.

[0012] Conversely, the voice text obtained in the call centre device 1 is converted into a voice signal (data) by the voice conversion unit 17 and sent to the user terminal 2 .

[0013] The generating artificial intelligence (AI) 4 is connected to the communication interface 18 via the Internet 3. The generating artificial intelligence 4 receives inquiry information from the call center device 1 and provides response information to the call center device 1. In addition, an information terminal (e.g., a personal computer) 5 operated by an operator (administrator) can also be connected to the communication interface 18 via the Internet 3.

[0014] The history memory 19 records the operation details within the call center device 1, and for example, inquiry information from the call center device 1 to the generating artificial intelligence 4 and response information from the generating artificial intelligence 4 to the call center device 1 are recorded in chronological order, i.e., together with time.

[0015] The operator extracts in advance inappropriate or undesirable contents from the response information part from the generating artificial intelligence 4 in the history memory 19, for example, contents not used in the industry or contents prohibited by copyright law, and sets a correct response information part for it. The response information part learning control model 20 is composed of a neural network consisting of an input layer, multiple intermediate layers, and an output layer. Therefore, the response information part learning control model 20 learns by deep learning using multiple inappropriate or undesirable response information parts as input teacher data and corrected response information parts corresponding to them as output teacher data.

[0016] Fig. 2 is a flowchart for explaining the operation of the CPU 11 of the call center device 1 of Fig. 1. This flowchart is stored in the ROM 12 or flash memory 13 of Fig. 1. This routine starts when a call from the user terminal 2 is received.

[0017] First, in step 201, an automatic response voice is sent to the user terminal 2, and in response to this, user information, in this case a telephone number, is received via the voice recognition unit 16. Next, in step 202, the user information is written into the history memory 19.

[0018] Next, in step 203, inquiry information is received from the user terminal 2 via the voice recognition unit 16. Then, in step 204, the inquiry information is transmitted to the generation artificial intelligence 4 via the communication interface 18 and the Internet 3. Then, in step 205, the inquiry information is written into the history memory 19.

[0019] Next, in step 206, response information is received from the generation artificial intelligence 4.

[0020] Next, in step 207, the inappropriate response information portion is used as input teaching data and the corrected response information portion is used as output teaching data to input the response information from the generation artificial intelligence 4 and estimate and extract the corrected response information portion of the inappropriate response information portion in response information portion learning model 20. If the corrected response information portion is extracted as a result, the process proceeds to step 208 and the inappropriate response information portion is rewritten to the corrected response information portion. On the other hand, if the corrected response information portion is not extracted, the process proceeds directly to step 209.

[0021] In step 209 , the response information or the response information including the corrected response information portion is transmitted to the user terminal 2 via the speech conversion unit 17 , and then in step 210 , the response information transmitted to the user terminal 2 is written in the history memory 20 .

[0022] In step 211, it is determined whether the user has hung up the phone and whether the time limit has been exceeded. If the user has not hung up the phone before the time limit, the above steps 203 to 210 are repeated. On the other hand, if the time limit has been exceeded or the user has hung up the phone, the process proceeds to step 212.

[0023] Next, in step 212, the routine waits for a predetermined time for re-inquiry information from the user via the voice recognition unit 16. If there is re-inquiry information, the routine proceeds to step 213, and if there is no re-inquiry information, the routine proceeds directly to step 214, where the routine ends.

[0024] In step 213, an operator or administrator responds. For example, the operator or administrator creates re-response information and transmits it to the user. If it takes time to create the re-response information, a voice signal "We will send you the re-response information by phone later" is first transmitted to the user, and then the re-response information is transmitted to the user by phone. In addition, when the re-response information is created, the inquiry information, the re-inquiry information, and the re-response information are transmitted to the generation artificial intelligence 4 via the communication interface 18, contributing to the re-learning of the generation artificial intelligence 4. Then, the process ends in step 214.

[0025] The call center device 1 in FIG. 2 can also be configured with hardware shown in FIG.

[0026] In Figure 3, the call center device comprises an inquiry receiving means 302 that receives inquiry information from a user terminal 301, an inquiry information transmitting means 304 that transmits the inquiry information to a large-scale language model 303, a response information receiving means 305 that receives response information to the inquiry information from the large-scale language model 303, and a response information transmitting means 306 that transmits the response information to the user terminal 301.

[0027] Although the call center device 1 and the user terminal 2 are connected by voice communication, the user terminal 2 and the call center device 1 may be connected by digital data communication.

[0028] Furthermore, the present invention can be applied to any modifications within the obvious scope of the above-described embodiment. [Explanation of symbols]

[0029] 1: Call center equipment 11: CPU 12:ROM 13: Flash memory 14: RAM 15: I / O interface 16: Voice recognition unit 17a: Voice conversion unit 18: Communication interface 19: History memory 20: Response information partial learning control model 301: User terminal 302: Inquiry information receiving means 303: Large-scale language model 304: Inquiry information transmission means 305: Response information receiving means 306: Response information transmission means

Claims

1. an inquiry receiving procedure for receiving inquiry information from a user terminal; a query information transmission step of transmitting the query information to a large-scale language model; a response information receiving step of receiving response information to the query information from the large-scale language model; a response information transmission procedure for transmitting the response information to the user terminal; A call center program for causing a computer to execute the above.

2. moreover, a corrected response information part estimation extraction step of estimating and extracting a corrected response information part by inputting the response information using a learned response information part learning control model with an inappropriate response information part as input teacher data and a corrected response part of the inappropriate response information part as output teacher data; a response information modifying step of modifying the response information by the extracted corrected response information portion; 2. The call center program according to claim 1, for causing a computer to execute the above steps.

3. moreover, a re-inquiry information receiving step of receiving re-inquiry information from the user terminal after transmitting the response information to the user terminal; a re-response information step of transmitting re-response information created by an administrator to the user terminal after receiving the re-inquiry information; 2. The call center program according to claim 1, for causing a computer to execute the above steps.

4. An inquiry receiving means for receiving inquiry information from a user terminal; A query information transmitting means for transmitting the query information to a large-scale language model; a response information receiving means for receiving response information to the query information from the large-scale language model; a response information transmitting means for transmitting the response information to the user terminal; A call center device comprising:

5. moreover, a corrected response information part estimation extraction means for estimating and extracting a corrected response information part by inputting the response information using a learned response information part learning control model with an inappropriate response information part as input teacher data and a corrected response part of the inappropriate response information part as output teacher data; a response information correcting means for correcting the response information by the extracted corrected response information portion; 5. The call center device according to claim 4, further comprising:

6. moreover, a re-inquiry information receiving means for receiving re-inquiry information from the user terminal after transmitting the response information to the user terminal; a re-response information means for transmitting re-response information created by an administrator to the user terminal after receiving the re-inquiry information; 5. The call center device according to claim 4, further comprising:

Images