Sales system and screening program

Abstract

To concentrate business resources on an object person who seems to have a pulse by performing screening without omission.SOLUTION: A call data storage unit 501 for storing call data in a call center for commodity sales for each call destination, and a marking unit 502 for marking the call data determined that the call destination is interested in the commodity or service; A sales system 1 includes an extraction model generation unit 503 that generates, by machine learning, a target extraction model in which an input is call data and an output is a target to be further sold, and a target extraction unit 504 that extracts a target from unmarked call data stored in a call data storage unit 501 by using the generated target extraction model.SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] The present invention relates to a system and program that performs efficient screening and enables sales resources to be focused on potential targets when, for example, making intensive sales calls to a large number of recipients via a call center to sell a specific product or provide a service. [Background technology]

[0002] Traditionally, when selling a company's new product through a call center, for example, a large number of operators would call based on a predetermined list or customer data to introduce the new product and conduct sales activities, i.e., sales. At this time, many customers would decline, stating that they were not interested, but some would listen to a brief explanation, and even a very small number would become interested and want to hear more in detail or request that materials be sent to them. In such cases, the operator would end the call by asking for permission to call back again to have a more specialized operator or technician provide further explanation, etc., as appropriate, and then connect the call to the next person. In other words, although the ratio is extremely small (especially as it approaches cold calling), operators will perform preliminary screening, such as marking the caller as "interested" in the product, and then hand over the list to specialized operators or sales representatives. This allows limited sales resources to be focused on callers who are interested, making sales activities more efficient and effective.

[0003] However, in reality, the reactions of the people on the other end of the line and the way the conversation flows vary, and the operators' perceptions also differ slightly. Therefore, even among those who are not marked, there may be people who are actually interested in the product, or who may potentially be interested. In other words, there is the problem that some people may be missed in the first screening.

[0004] The above are examples of operator-mediated screening for the purpose of selling products or providing services, but similar problems (oversights) also exist in the broader VOC (Voice of Customer) environment, including customer support after the sale of a product or the provision of a service. In other words, in the process of collecting feedback, opinions, requests, complaints, etc. from the other end of the line and using them to improve and further develop products and services, operators would mark cases as useful or as content that should be handled properly, and then pass them on to specialists, but there were cases where the content of the call or conversation that was not marked was missed. Regardless of whether an operator is present or not, oversights will occur to a greater or lesser extent depending on the personality and qualities of the individual making the judgment. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent Publication No. 2024-068536 [Patent Document 2] Patent Publication No. 2023-182380 [Patent Document 3] Special Publication 2012-513165 Summary of the Invention [Problem to be solved by the invention]

[0006] The present invention has been made in view of the above, and aims to build a system that, when making concentrated sales calls to a large number of parties, can screen them without missing any candidates and focus sales resources on targets who seem to have potential. More generally, the objective is to build a system that, when conducting sales activities with a large number of parties, can screen all candidates without missing any and focus sales resources on the target parties. [Means for solving the problem]

[0007] The sales system described in claim 1 is characterized by comprising: call data storage means for storing call data for each call recipient at a call center that conducts sales related to the sale of goods or the provision of services; call data marking means for marking call data that a call center operator has determined to indicate that the call recipient is interested in the product or service; focus target extraction model generation means for using the call data marked by the call data marking means as training data and generating by machine learning a focus target extraction model that takes call data as input and focuses targets who should be further marketed as output; and focus target extraction means for using the focus target extraction model generated by the focus target extraction model generation means to extract focus targets from unmarked call data stored in the call data storage means.

[0008] In other words, the invention of claim 1 substantially standardizes the judgments of call center operators, making it possible to efficiently extract target individuals who have the potential to lead to a sale from a large amount of unmarked call data, thereby enabling sales activities that do not miss any candidates. In other words, efficient and effective primary screening can be achieved, allowing the next stage of sales activities to be focused on key target individuals. The model can be updated as needed, but once the model is generated, it is possible to extract potential candidates regardless of whether the operator marks them or not (marking can be done automatically, and candidates who appear to be promising can also be extracted).

[0009] Sales is broadly defined, including the narrow definition of introducing products or services to call recipients, providing various explanations to encourage them to purchase products or enjoy services, as well as various inquiries (and responses to those inquiries) from call recipients and after-sales service. It also includes the so-called Voice of the Customer (VOC). While sales related to products or services refer to sales of the same product or service, it can also be broadened to include similar products, similar services, etc., as appropriate. In addition, in this application, sales includes activities that bring monetary benefits to the business or organization conducting the sales activity, activities that prevent monetary benefits, and even non-monetary activities that maintain and improve trust and brand power, or activities that prevent damage to trust and brand power. For example, sales also includes various efforts to improve the reputation of schools, hospitals, and local governments, and to increase the credibility of non-profit organizations. The call data may be either audio data or text data transcribed from the audio data. Audio data is preferable because information such as intonation, tone, and pauses are also the subject of learning. The call data storage means can also be called a call data recording means. Note that when transcribing audio data using software or the like, mistranslations may occur, but the data may be used as is, or data that has been proofread separately may be used. The call data marking means marks (sets a flag) the call data, but it is sufficient that information indicating that the operator has determined that the call data is interested is input, and this does not mean that the operator itself is the call data marking means. Also, it is possible for someone other than the operator to determine that the call data is interested. Note that the determination that the call data is interested is in a broad sense, and may also include a determination that the call data is likely to be interested (or is thought to be interested), as appropriate. During learning, marked call data is used as training data, but unmarked data can also be used as unmarked training data.

[0010] The sales system described in claim 2 is characterized by comprising: call data storage means for storing call data for each call recipient at a call center that conducts sales related to the sale of goods or the provision of services; focus target extraction model generation means for generating, by machine learning, a focus target extraction model that uses call data of call recipients that actually resulted in the sale of goods or the provision of services as training data and takes call data as input and focus targets who should be further marketed as output; and focus target extraction means for extracting focus targets from the call data sequentially stored by the call data storage means using the focus target extraction model generated by the focus target extraction model generation means.

[0011] In other words, the invention of claim 2 makes it possible to efficiently extract target individuals who have the potential to lead to a successful sale based on cases in which a product or service has actually been sold or a service has been provided, thereby enabling more efficient sales activities. In other words, efficient and effective primary screening can be achieved, allowing the next stage of sales activities to be focused on key target individuals. The model can be updated as needed, but once the model is generated, it is possible to extract potential candidates regardless of whether the operator marks them or not (marking can be done automatically, and candidates who appear to be promising can also be extracted).

[0012] Depending on the usage, a focus target extraction model may be generated by narrowing down the data of those who have concluded a deal among those who have been determined by a call center operator or the like to be interested in the product or service in question. In addition, a model can be constructed separately using machine learning to determine whether the call recipient is interested in the product or service in question, and the determination can be made based on this model. When learning, call data from parties that actually resulted in the sale of goods or the provision of services (leading to a contract) is used as training data, but data that did not result in the sale of goods or the provision of services can also be used as training data that did not result in a contract.

[0013] The sales system described in claim 3 is characterized by comprising: an interaction data storage means for recording interaction data for each counterparty in the process of conducting sales related to the sale of a product or the provision of a service by telephone, chat, email, social media, or other interactive communication means; an interaction data marking means for marking interaction data that is determined to indicate that the counterparty is interested in the product or service based on the content of the interaction; an attention target extraction model generation means for using the interaction data marked by the interaction data marking means as training data and generating by machine learning an attention target extraction model that takes the interaction data as input and specifies attention targets who should be further promoted as output; and a focus target extraction means for using the attention target extraction model generated by the focus target extraction model generation means to extract attention targets from the unmarked interaction data stored in the interaction data storage means.

[0014] In other words, the invention of claim 3 makes it possible to efficiently extract target individuals who are likely to lead to a sale from a large amount of conversation data obtained through telephone, chat, email, social media, and other interactive communication means, thereby enabling sales activities that do not miss any candidates. In other words, efficient and effective primary screening can be achieved, allowing the next stage of sales activities to be focused on key target individuals.

[0015] The determination of whether the other party is interested in the product or service based on the content of the dialogue can be made by an operator at a call center, or by an experienced sales representative who checks the content one by one. A determination model can also be constructed separately using machine learning, and the determination can be made based on the determination model. The dialogue data marking means may appropriately input a mark to dialogue data for which it has been determined that the other party is interested in the product or service based on the content of the dialogue. During learning, marked call data is used as training data, but unmarked data can also be used as unmarked training data.

[0016] The sales system described in claim 4 is characterized by comprising: an interaction data storage means for recording interaction data for each counterparty in the process of conducting sales related to the sale of goods or the provision of services by telephone, chat, email, social media, or other interactive communication means; a focus target extraction model generation means for generating, by machine learning, an interest target extraction model using the interaction data of the counterparty with whom the goods or services have actually been sold as training data, and taking the interaction data as input and focus targets who should be further marketed as output; and a focus target extraction means for extracting interest targets from the interaction data sequentially stored by the interaction data storage means, using the interest target extraction model generated by the focus target extraction model generation means.

[0017] In other words, the invention of claim 4 makes it possible to efficiently extract target individuals who have the potential to lead to a successful sale based on cases in which a product or service has actually been sold or a service has been provided, thereby enabling more efficient sales activities. In other words, efficient and effective primary screening can be achieved, allowing the next stage of sales activities to be focused on key target individuals. Depending on the mode of use, the focus target extraction model may be generated by narrowing down the data of the targets who have concluded a contract from among the appropriately marked counterparties. In addition, a model can be constructed separately using machine learning to determine whether or not the other party is interested in the product or service in question, and the determination can be made based on this model. When learning, call data from parties that actually resulted in the sale of goods or the provision of services (leading to a contract) is used as training data, but data that did not result in the sale of goods or the provision of services can also be used as training data that did not result in a contract.

[0018] The screening program described in claim 5 is a screening program for operating a sales system described in any one of claims 1 to 4, and is characterized in that it causes a computer that constructs the system to function as each of the means defined in the claim.

[0019] In other words, the invention of claim 5 realizes efficient and effective primary screening for sales of goods or services, and enables the next stage of sales activities to be focused on target individuals. The specific details of each means are as follows: <Program corresponding to claim 1> This is a program that builds a system that allows efficient screening when making sales calls to a large number of potential customers, and allows you to focus your sales capital on potential customers. Computer, A call data storage means for storing call data for each call destination at a call center that conducts sales related to product sales or service provision; a call data marking means for marking call data when a call center operator determines that the call recipient is interested in the product or service; a focus target extraction model generation means for generating a focus target extraction model by machine learning, using the call data marked by the call data marking means as training data, and taking the call data as input and focus targets who should be further marketed as output; and a focus target person extraction means for extracting a focus target person from unmarked call data stored in the call data storage means, using the focus target person extraction model generated by the focus target person extraction model generation means; A screening program characterized by functioning as a <Program corresponding to claim 2> This is a program that builds a system that allows efficient screening when making sales calls to a large number of potential customers, and allows you to focus your sales capital on potential customers. Computer, A call data storage means for storing call data for each call destination at a call center that conducts sales related to product sales or service provision; A means for generating a target person extraction model by machine learning, which uses call data of call destinations that actually resulted in product sales or service provision as training data, and generates a target person extraction model with the call data as input and target people who should be further marketed as output; and a focus target person extraction means for extracting a focus target person from the call data sequentially stored by the call data storage means, using the focus target person extraction model generated by the focus target person extraction model generation means; A screening program characterized by functioning as a <Program corresponding to claim 3> This is a program that builds a system that allows efficient screening when making sales calls to a large number of potential customers, and allows you to focus your sales capital on potential customers. Computer, Interaction data storage means for recording interaction data for each party in the process of conducting sales related to the sale of products or the provision of services via telephone, chat, email, social media, and other interactive communication means; a dialogue data marking means for marking dialogue data that indicates that the other party is interested in the product or service based on the content of the dialogue; a focus target extraction model generation means for generating a focus target extraction model by machine learning, using the dialogue data marked by the dialogue data marking means as training data, and taking the dialogue data as input and focus targets who should be further marketed as output; and a focus target person extraction means for extracting a focus target person from unmarked dialogue data stored in the dialogue data storage means, using the focus target person extraction model generated by the focus target person extraction model generation means; A screening program characterized by functioning as a <Program corresponding to claim 4> This is a program that builds a system that allows efficient screening when making sales calls to a large number of potential customers, and allows you to focus your sales capital on potential customers. Computer, Interaction data storage means for recording interaction data for each party in the process of conducting sales related to the sale of products or the provision of services via telephone, chat, email, social media, and other interactive communication means; A means for generating a focus target extraction model by machine learning, which uses dialogue data of a counterparty that has actually led to the sale of a product or provision of a service as training data, and generates a focus target extraction model in which the dialogue data is used as input and focus targets who should be further promoted as output; and a focus target person extraction means for extracting a focus target person from the dialogue data sequentially stored in the dialogue data storage means, using the focus target person extraction model generated by the focus target person extraction model generation means; A screening program characterized by functioning as a [Effects of the Invention]

[0020] According to the present invention, when making concentrated sales calls to a large number of people, it is possible to construct a system that can screen all candidates without missing any and focus sales resources on those who appear to be promising. More generally, when conducting sales activities with a large number of people, a system can be constructed that screens all candidates without missing any and allows sales resources to be focused on the target audience. [Brief explanation of the drawings]

[0021] [Figure 1] FIG. 1 is an explanatory diagram showing an example of the configuration of a sales system according to the present invention. [Figure 2] 1 is a diagram illustrating an example of a hardware configuration of a voice recording device 100. FIG. [Figure 3] 10 is a diagram illustrating an example of a screen configuration displayed on a liquid crystal monitor. [Figure 4]FIG. 2 is a diagram illustrating an example of a hardware configuration of an analysis device 200. [Figure 5] FIG. 2 is an explanatory diagram showing the functional configuration of a sales system. DETAILED DESCRIPTION OF THE INVENTION

[0022] A sales system in which the screening program of the present invention is introduced will now be described. Here we will explain the example of a printer manufacturer selling replacement ink and toner from other companies. The entry point is the sale of replacement ink and toner, but this is a business scheme in which the manufacturer uses this as a clue to gain customer trust and later sell or lease the manufacturer's own printers.

[0023] In this embodiment, replacement ink and toner are introduced and sold at a call center. <System configuration example> FIG. 1 is an explanatory diagram showing an example of the configuration of a sales system according to the present invention. The sales system 1 mainly comprises voice recording devices 100 (voice recording devices 100_1, voice recording devices 100_2, ...) that record call data, which is voice data of calls made by operators P (operator P1, operator P2, ...) at a call center C, an analysis device 200 that extracts call partners, i.e., focus targets, from the voice data via a model based on machine learning, as sales targets that should be focused on, and a network N that connects these.

[0024] <Appearance of the device> The voice recording device 100 and the analysis device 200 can be general computers, such as desktop PCs or server devices, and a description of their external configurations will be omitted.

[0025] <Hardware Configuration of the Audio Recording Device 100> The hardware configuration of the voice recording device 100 used by each operator P will be described. FIG. 2 is a diagram illustrating an example of the hardware configuration of the voice recording device 100. As shown in FIG. The audio recording device 100 has a hardware configuration including a CPU 101, a ROM 102, a RAM 103, a hard disk (HD) 104, a graphics board 105, an LCD monitor 106, a keyboard (K / B) 107, a mouse 108, a network interface 109, a sound board 110, and a headset 111.

[0026] The CPU 101, together with the OS, controls the entire voice recording device 100, sequentially saves voice data, and transmits it to the analysis device 200. Specifically, the CPU 101 converts, saves, transmits, links to the telephone number of the other party, marks, etc. of the voice data in accordance with a program stored on the hard disk 104. It also works in cooperation with other voice recording devices 100 and analysis devices 200 to transmit data in a manner that prevents voice data from being transmitted in large quantities at once (so as not to slow down the communication speed). In addition, the CPU 101 also controls the temporary storage of work data stored on the hard disk 104 in the RAM 103 .

[0027] The ROM 102 stores a boot program and the like. Depending on the mode of use, the ROM 102 may also store a control program for the audio recording device 100. The RAM 103 is used as a work area for the CPU 101. Specifically, it temporarily stores the contents of data read from the hard disk 104, the contents of programs, and the like.

[0028] The graphics board 105 sends an image signal to be output to the liquid crystal monitor 106. The graphics board 105 includes a GPU and an image output interface (image output I / F), and outputs an image processed by the GPU to the liquid crystal monitor 106.

[0029] 3 is an example of the screen configuration displayed on the LCD monitor 106. As shown in the figure, the name and phone number of the person on the other end of the call, a check box for marking them as a target, past product purchase history, past call history, a pull-down menu, a scrolling display within a frame, and a list of toner model numbers, compatible models, features, prices, etc. are displayed.

[0030] The network interface 109 connects the audio recording device 100 to a network N via a wired connection, Wi-Fi, or the like.

[0031] The hard disk 104 is made up of an application section 120 and a data storage section 130 .

[0032] The application section 120 includes an OS 121 that controls the entire voice recording device 100 , a voice data processing program 122 , and a voice data management program 123 .

[0033] The voice data processing program 222 processes the conversation between the operator and the other party input via the headset 111 on the sound board 110 to reduce the load on the CPU 101, and performs processing to appropriately reduce noise, compress the data, and store it on the hard disk 104 as voice data. The voice data management program 223 pairs the voice data in the hard disk 104 with the destination information (telephone number, name, past purchase history, etc.) and transmits them in a timely manner while checking the load and operating status of the analysis device 200. Note that all voice data is basically transmitted regardless of whether it is marked or not. To support the sales activities of the operator, the program displays product overviews and call recipient information on the LCD monitor 106, and also displays a check box that the operator can check if they feel that the call recipient is interested in the product being introduced. In addition, the voice data management program also stores and updates call recipient information on the hard disk 104.

[0034] The functional configurations described below are realized by the OS 111, the audio data processing program 222, and the audio data management program 223, either singly or in combination, and in some cases in cooperation with the data storage unit .

[0035] The data storage unit 130 includes a voice data storage unit 131 , a callee information storage unit 132 , and a product information storage unit 133 . The voice data storage unit 131 stores the voice data processed by the sound board 110. If a mark is added by the operator, the mark is incorporated as additional information of the voice data, and the updated voice data is stored. The callee information storage unit 132 stores callee information. This callee information consists of telephone number, company name, individual name (name of person in charge), name of department in charge, past purchase history and details (including history other than replacement ink and toner), etc. If necessary, capital, number of employees, operating profit and sales for the past three years, etc. are also loaded. This information is used by the operator to refer to as needed during the call and to assist in product purchases. The product information storage unit 133 stores various information related to the replacement ink / toner that is the product being introduced this time, such as compatible products, details on quality, price information, discount information, etc., for compatible ink / toner for other printers. Note that, since there may be inquiries about other products, such as printers, during a call, various information related to printers is also stored. In addition, various pieces of information are linked as appropriate, and the screen structure and screen transitions are constructed to improve operability for the operator.

[0036] <Hardware Configuration of Analysis Device 200> Next, the hardware configuration of analysis device 200 will be described. 4 is a diagram illustrating an example of the hardware configuration of analysis device 200. Analysis device 200 has, as its hardware configuration, CPU 201, ROM 202, RAM 203, hard disk (HD) 204, graphics board 205, monitor 206, keyboard (K / B) 207, mouse 208, network interface 209, and GPU 210 (different from that provided in graphics board 205). In the following, a description of the configuration similar to that of the audio recording device 100 will be omitted, and different hardware configurations will be mainly described.

[0037] The CPU 201, together with the OS, controls the entire analysis device 200, and extracts potential buyers from unmarked voice data, etc., based on the generated model. The CPU 201 also controls the temporary storage of work data stored on the hard disk 204 in the RAM 203.

[0038] The GPU 210 generates and updates a learning model for each sales case or sales field by machine learning, in cooperation with the RAM 203 and the HD 204. Note that a plurality of GPUs 210 may be provided as appropriate.

[0039] The hard disk 204 is made up of an application section 220 and a data storage section 230. Since audio data is collected from each audio recording device 100, a RAID configuration may be used as appropriate. The application section 220 is made up of an OS 221 that controls the entire analysis device 200, and an analysis program 222 that screens call partners and constructs a model for screening. The data storage unit 230 includes a voice data storage unit 231 , an extraction model storage unit 232 , a callee information storage unit 233 , and a product information storage unit 234 .

[0040] The analysis program 222 includes a model generation program 223 , a screening program 224 , and an information transmission update program 225 .

[0041] The model generation program 223 uses the marked voice data sent from the voice recording device 100 and stored in the call destination information storage unit 233 as training data, and generates a focus target extraction model through machine learning, with the voice data as input and focus targets who are further sales targets as output. The model generation algorithm may be a neural network, or may be a more reliable algorithm that will be developed in due course. Specifically, machine learning is used to learn call patterns when a person is interested based on word extraction, intonation, speaking time and speaking time ratio between the operator and the other party, silent time, etc. from the speech data. At this time, unmarked speech data can also be used for learning as appropriate. Because a model is built using machine learning, screening, which will be described later, can be performed effectively. Depending on the specifications, voice data of callers who actually purchased replacement ink or toner may be used as training data, going back to the caller information storage unit 233. In this case, the training data may be divided into three types: voice data that was not marked, and voice data of callers who were marked but did not ultimately purchase the product, and a model may be constructed by learning from these data.

[0042] The screening program 224 uses the extraction model for target persons generated by the model generation program 221 to extract target persons from unmarked call data stored in the voice data storage unit 231. This allows for selection of target persons even if an operator misses out, making it possible to perform thorough screening.

[0043] The information sending update program 225 sends and updates the latest versions of various data referenced by operators in each voice recording device 100. Information is managed centrally, and sales of replacement ink and toner can be realized based on the same information regardless of which voice recording device 100 an operator uses, regardless of which voice recording device 100.

[0044] The data storage unit 230 will now be described. The voice data storage unit 231 stores voice data transmitted from each voice recording device 100 at any time. Note that the analysis device 200 is not dedicated to replacement ink / toner orders, and therefore stores voice data for each type of order. The extraction model storage unit 232 stores the focus target person extraction model generated by the model generation program 221 for each case, in addition to the current case of replacement ink / toner. The callee information storage unit 233 stores callee information. This information is master data of callee information stored in each voice recording device 100. By centrally managing the information, the operator P can make calls while always viewing the latest callee information, without depending on the voice recording device 100. The product information storage unit 234 stores various information (price, compatible models, physical properties, comparison values ​​with other compatible inks and toners, delivery time, etc.) about the replacement ink / toner being introduced this time. This information is also master data of the product information stored in each voice recording device 100. Similarly, with centralized management, the operator P can make calls while always viewing product information that is updated as needed, regardless of the device.

[0045] <Functional configuration> Next, a description will be given of the functional configuration of the sales system 1. Fig. 5 is an explanatory diagram showing the functional configuration of the sales system 1. The sales system 1 has, as its functional configuration, a call data storage unit 501, a marking unit 502, an extraction model generation unit 503, and a focus target extraction unit 504.

[0046] The call data storage unit 501 stores call data for each call party at call center C, which sells replacement ink and toner. The call data is essentially audio data, and is stored after undergoing appropriate noise reduction and compression. While there are no particular limitations on the storage format, examples include mp3, AAC, and WMA. The functions of the call data storage unit 501 can be realized by, for example, the headset 111, the sound board 110, the audio data processing program 122, the audio data storage unit 131, the audio data storage unit 231, and the like.

[0047] Marking unit 502 marks the call data when operator P of call center C determines that the call recipient is interested in the target product, Dell replacement ink and toner. Specifically, when operator P marks a check box on the screen displayed on liquid crystal monitor 106, marking unit 502 adds the input information to the voice data. The marking unit 502 can realize its functions using, for example, the mouse 108, the liquid crystal monitor 106, the voice data management program 123, and the like.

[0048] The extraction model generation unit 503 uses the call data marked by the marking unit 502 as training data, and generates a focus target extraction model through machine learning, with the input being the call data and the output being focus targets, who should be further marketed to. The algorithm for generating the model is not particularly limited, and examples include neural networks, SVM (Support Vector Machine), decision trees, LSTM (Long Short Term Memory), etc. The algorithm may be changed depending on the business content, the industry and size of the call recipient, etc. As mentioned above, when learning, unmarked call data may also be used as training data, and further, voice data from the other party who actually purchased replacement ink or toner may be used for learning. The functions of the extraction model generation unit 503 can be realized by, for example, the GPU 210, the CPU 201, the RAM 203, the model generation program 223, the extraction model storage unit 232, and the like.

[0049] The focus target person extraction unit 504 uses the focus target person extraction model generated by the extraction model generation unit 503 to extract focus target persons from the unmarked call data stored by the call data storage means. The function of the attention target extraction unit 504 can be realized by the screening program 224, the extraction model storage unit 232, the voice data storage unit 231, the RAM 203, and the like.

[0050] As explained above, according to the present invention, target individuals can be extracted even from unmarked voice data using a model generated by machine learning, thereby eliminating the need for inefficient and unrealistic checking work such as manually checking each voice, and realizing effective screening to eliminate oversights and lead to a deal (purchase).

[0051] Although the above description is based on communication via voice calls, the present invention is not limited to voice calls. In other words, the present invention can be implemented in the process of conducting sales related to product sales or service provision via chat, email, social media, or other interactive communication means by recording dialogue data for each party.

[0052] In terms of functional configuration, the sales system can be configured to have a dialogue data storage unit, a dialogue data marking unit, a focus target person extraction model generation unit, and a focus target person extraction unit. Each functional unit performs the following processes.

[0053] The dialogue data storage unit records dialogue data for each party during the course of sales related to product sales or service provision, etc., via telephone, chat, email, social media, or other interactive communication means. The dialogue data marking unit marks dialogue data that is determined to indicate that the other party is interested in the product or service based on the content of the dialogue. The focus target extraction model generation unit uses the dialogue data marked by the dialogue data marking unit as training data, and generates a focus target extraction model through machine learning, with the dialogue data as input and focus targets, who should be further marketed, as output. The focus target extraction unit uses the focus target extraction model generated by the focus target extraction model generation unit to extract the focus target from the unmarked dialogue data stored in the dialogue data storage unit. When generating a model, dialogue data of the other party that actually led to the sale of a product or the provision of a service may be used as training data, or dialogue data that did not lead to the sale of a product or the provision of a service may also be used as training data as appropriate.

[0054] In addition, the present invention can also be constructed as the following business method. <Sales Method> 1. A method for a business model of screening call content, comprising the steps of: 1. An operator talks to a customer and assesses their interest in the business (sale of a product or provision of a service). 2. All calls are recorded, and if the operator deems the call data interesting, they flag it. 3. Create a machine learning model based on flagged call data to predict customer interests. 4. Using machine learning models, we extract calls that were not flagged and are predicted to be flagged. 5. A specialized operator or sales representative will call back the extracted cases, ensuring no sales opportunities are missed.

[0055] Furthermore, 6. The above method, further comprising the step of extracting features from the recorded data of the call content and training a machine learning model using the features. 6-1. Extract features such as customer responses, keywords, and speaking patterns from call recording data. 6-2. Feed the extracted features into a machine learning algorithm to create a model for predicting customer interests. 6-3. The created model identifies opportunities that are likely to be of interest from unflagged call data.

[0056] Furthermore, 7. The above method, further comprising a step of continuously adding new call data and updating the machine learning model to improve its accuracy. 7-1. As new call data is generated, add it to the training dataset for the machine learning model. 7-2. Evaluate the model's predictive accuracy and retrain the model if necessary. 7-3. Use the updated model to more accurately predict potential leads and maximize the effectiveness of your sales efforts. [Industrial Applicability]

[0057] According to the present invention, when making concentrated sales calls to a large number of parties, it is possible to construct a system that can screen all potential candidates without missing any and focus sales resources on those who appear to be promising. More generally, when making sales calls to a large number of parties, it is possible to construct a system that can screen all potential candidates without missing any and focus sales resources on those who should be focused on. This can be applied to product image surveys and other such purposes to explore customer preferences. [Explanation of symbols]

[0058] 1 Sales System 100 Audio Recording Device 101 CPU 103 RAM 104 Hard Disk 106 LCD monitor 110 Soundboard 111 Headset 120 Application Section 122 Audio data processing program 123 Voice Data Management Program 130 Data storage unit 131 Audio data storage unit 132 Call destination information storage unit 133 Product information storage section 200 Analysis equipment 201 CPU 203 RAM 204 Hard Disk 220 Application Department 221 Model Generation Program 222 Analysis Program 223 Model Generation Program 224 Screening Program 225 Information Transmission Update Program 230 Data storage unit 231 Audio data storage unit 232 Extraction model storage section 233 Call destination information storage unit 234 Product information storage section 501 Call Data Storage Unit 502 Marking section 503 Extraction Model Generation Unit 504 Focus Target Extraction Unit C. Call Center N Network P Operator

Claims

1. A call data storage means for storing call data for each call destination in a call center that sells products or provides services; a call data marking means for marking call data when a call center operator determines that the call recipient is interested in the product or service; a focus target extraction model generation means for generating a focus target extraction model by machine learning, using the call data marked by the call data marking means as training data, and taking the call data as input and focus targets who should be further marketed as output; a focus target person extraction means for extracting a focus target person from unmarked call data stored in the call data storage means, using the focus target person extraction model generated by the focus target person extraction model generation means; A sales system characterized by comprising:

2. A call data storage means for storing call data for each call destination in a call center that sells products or provides services; A focus target extraction model generation means for generating a focus target extraction model by machine learning, using call data of call destinations that actually resulted in product sales or service provision as training data, and taking the call data as input and focus targets who should be further marketed as output; a focus target person extraction means for extracting a focus target person from the call data sequentially stored by the call data storage means, using the focus target person extraction model generated by the focus target person extraction model generation means; A sales system characterized by comprising:

3. A conversation data storage means for recording conversation data for each counterparty in the process of conducting sales related to the sale of goods or provision of services by telephone, chat, email, social media, or other conversational communication means; a dialogue data marking means for marking dialogue data that indicates that the other party is interested in the product or service based on the content of the dialogue; a focus target extraction model generation means for generating a focus target extraction model by machine learning, using the dialogue data marked by the dialogue data marking means as training data, and taking the dialogue data as input and focus targets who should be further marketed as output; a focus target person extraction means for extracting a focus target person from unmarked dialogue data stored in the dialogue data storage means, using the focus target person extraction model generated by the focus target person extraction model generation means; A sales system characterized by comprising:

4. A conversation data storage means for recording conversation data for each counterparty in the process of conducting sales related to the sale of goods or provision of services by telephone, chat, email, social media, or other conversational communication means; A focus target extraction model generation means for generating a focus target extraction model by machine learning, using dialogue data of a counterparty that has actually led to the sale of a product or provision of a service as training data, and taking the dialogue data as input and focus targets who should be further marketed as output; a focus target person extraction means for extracting a focus target person from the dialogue data sequentially stored by the dialogue data storage means, using the focus target person extraction model generated by the focus target person extraction model generation means; A sales system characterized by comprising:

5. A screening program for operating the sales system according to any one of claims 1 to 4, A computer that constructs the system includes the means defined in the claims, A screening program characterized by functioning as a

Images