Interview support device, interview support method, and program

Abstract

To support an interview by providing effective information in the interview.SOLUTION: An interview support device is provided with a control part for performing acquisition processing for acquiring information of a final selection result by a guest about an interview performed between a host and the guest, extraction processing for extracting only utterance content of the guest from the utterance content between the host and the guest in the interview, and learning processing for setting input data for learning from the utterance content of the guest extracted by the extraction processing to learn a correlation between the final selection result and the utterance content.SELECTED DRAWING: Figure 8

Description

[Technical field]

[0001] The present invention relates to an interview support device, an interview support method, and a program, and in particular to a technique for supporting an interview. [Background technology]

[0002] Interviews are conducted in various fields such as business transactions, education, entertainment, recruitment, and more. For example, in business transactions, interviews (business negotiations) are conducted between sales representatives and customers. In addition to face-to-face interviews, online interviews via networks such as the Internet and telephone lines have also become common in recent years.

[0003] The following Patent Document 1 discloses a technique for managing the correlation between up-sales talk spoken by a sales clerk and actual sales performance, thereby improving sales. The following Patent Document 2 describes a method of selecting a specific term from a customer's comment, and selecting a response corresponding to that term from among responses prepared in advance. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2016-200851 A [Patent Document 2] Patent No. 6572354 Summary of the Invention [Problem to be solved by the invention]

[0005] However, it is difficult to clearly understand what kind of products a customer wants from the statements of salespeople such as store clerks and sales representatives during interviews. This is because store clerks and sales representatives also tend to recommend the products they want to sell. For this reason, it is difficult to accurately learn the relationship between a customer's latent desires and sales performance from conversations during interviews.

[0006] Therefore, the present invention provides a technique for enabling accurate estimation of products, etc., that a customer consciously or unconsciously desires. [Means for solving the problem]

[0007] The interview support device of the present invention includes a control unit that performs an acquisition process for acquiring information on the final selection result by the guest in an interview held between a host and a guest, an extraction process for extracting only the guest's utterance content from the utterance content of the host and guest in the interview, and a learning process for setting learning input data from the guest's utterance content extracted in the extraction process and learning the correlation between the final selection result and the utterance content. For example, a host and a guest are a sales representative and a customer in a business negotiation. The final selection result is the thing or object that the guest finally selects, and in the case of a business negotiation, it would be the product that the guest purchases. The interview support device excludes the contents of the host's speech and extracts the contents of the guest's speech, and learns the correlation between the contents of the guest's speech and the final selection result.

[0008] In addition, the interview support device of the present invention is equipped with a control unit that performs an acquisition process to acquire conversation audio of an interview between a host and a guest, an extraction process to extract only the guest's speech content from the conversation audio, a determination process to determine a final selection result related to the speech content extracted in the extraction process by referring to learning data generated in a learning process that learns the correlation between the final selection results of past interviews and the guest's speech content, and a process to send proposed information to a terminal device based on the determination result of the determination process. The system acquires the guest's speech during the interview, refers to the learning data, and determines the final selection results that are highly relevant to the guest's speech.Then, it makes a proposal based on the result of the judgment, such as a recommendation of a product. Effect of the Invention

[0009] According to the present invention, by excluding the utterances of the interview host, it is possible to avoid using business utterances, such as utterances that the host side recommends to the guest to lead them to a final selection result, in learning, and to set learning input data that is useful for estimating the customer's intentions. This makes it possible to build learning data that is useful for business negotiations, etc. Furthermore, according to the present invention, by estimating the guest's potential desires using learning data during or after a business negotiation, the host is able to suggest desirable final choices to the guest. [Brief description of the drawings]

[0010] [Figure 1] 1 is a block diagram of an example of a system configuration including an interview support device according to an embodiment. [Diagram 2] FIG. 13 is an explanatory diagram of an interview screen according to an embodiment. [Diagram 3] 1 is a block diagram of an information processing device serving as an interview support device according to an embodiment of the present invention. [Figure 4] 11 is an explanatory diagram of examples of proposed products, final purchased products, and customer utterances for each customer. [Diagram 5] FIG. 1 is an explanatory diagram of a proposed example of an embodiment. [Figure 6] FIG. 4 is an explanatory diagram of learning data according to the embodiment. [Figure 7] FIG. 2 is an explanatory diagram of general conversation content and specific conversation content according to an embodiment. [Figure 8] 11 is a flowchart of an example of a learning process according to the first embodiment. [Figure 9] 13 is a flowchart of an example of a process for updating a general conversation content and a specific conversation content according to an embodiment; [Figure 10] 13 is a flowchart of an example of a proposal process during business negotiation according to the first embodiment. [Figure 11] 13 is a flowchart of an example of a learning process according to the second embodiment. [Figure 12] 13 is a flowchart of an example of a proposal process during business negotiation according to the second embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] The embodiments will be described below in the following order. <1. System configuration> <2. Configuration of information processing device as interview support device> <3. First embodiment> <4. Second embodiment> <5. Effects of the embodiment and examples of modifications>

[0012] In this disclosure, an interview includes a face-to-face meeting, a face-to-face meeting via remote communication, and even a verbal or written meeting via telephone, chat, etc. In this disclosure, a written meeting is also included in the term "utterance."

[0013] In the embodiment, a "business negotiation" between a sales representative and a customer is given as an example of an "interview." The sales representative is an example of a host, and the customer is an example of a guest. However, the host is not limited to a sales representative, and the guest is not limited to a customer. It is sufficient that one party in various interviews is the host and the other is the guest.

[0014] <1. System configuration> 1 is a block diagram showing an example of a system configuration including an interview support device 1 according to an embodiment and its peripheral devices. A person in charge terminal 2, a customer terminal 5, etc. are connected to the interview support device 1 via a network 7 so as to be able to communicate data with each other. Also shown are a microphone 4 and a transmitter 6.

[0015] The interview support device 1, staff terminal 2, and customer terminal 5 are configured with information processing devices equipped with a processor such as a CPU (Central Processing Unit), and a microcomputer having a ROM (Read Only Memory) and RAM (Random Access Memory). Specifically, the staff terminal 2 and the customer terminal 5 are assumed to be personal computers, smartphones, tablet devices, etc. The interview support device 1 is assumed to be a computer device such as a personal computer or a workstation, or a computer device serving as a cloud server.

[0016] The network 7 is a network such as the Internet or a telephone network. The communication transmission path of the network 7 is assumed to be a wired transmission path, a wireless transmission path, or a combination of these.

[0017] The staff member terminal 2 is a terminal device used by, for example, a salesperson who negotiates business with a customer. The salesperson can use the staff member terminal 2 to meet face-to-face with the customer's terminal device (customer terminal 5) via communication over the network 7 and negotiate business. Although only one pair of the person in charge terminal 2 and the customer terminal 5 is shown in the figure, negotiations may be conducted simultaneously between multiple pairs of the person in charge terminal 2 and the customer terminal 5.

[0018] Information related to the business negotiation is transmitted from the person in charge terminal 2 to the interview support device 1. Specifically, this information is image data of a business negotiation screen for a remote business negotiation, and audio data of the business negotiation. The image data includes video and still images. It is assumed that the image data includes video of the faces of the customer and the person in charge. The audio data of the business negotiation is audio data of a conversation between the person in charge and the customer.

[0019] 2A and 2B show examples of display contents that are synchronously displayed to the person in charge and the customer during a remote business negotiation. Fig. 2A shows the screen of the customer terminal 5 (customer screen 30), and Fig. 2B shows the screen of the person in charge terminal 2 (person in charge screen 31). Here, the display contents are exemplified when the images provided from the customer terminal 5 and the staff terminal 2 are captured images of the customer and the staff, respectively.

[0020] In this case, on the customer screen 30, as shown in Fig. 2A, a captured image 32 of the person in charge as a communication partner (image provided from the person in charge terminal 2) is displayed large, and a captured image 33 of the customer himself is displayed small. Similarly, on the person in charge screen 31, as shown in Fig. 2B, a captured image 33 of the customer as a communication partner (image provided from the customer terminal 5) is displayed large, and a captured image 32 of the person in charge himself is displayed small. For example, a business negotiation can be carried out while viewing each other's images in this way.

[0021] 2A and 2B show an example of synchronous display of display contents in which captured images 32 and 33 are displayed synchronously on each terminal, but the images to be displayed synchronously are not limited to captured images 32 and 33. For example, a document may be shared between a customer and a person in charge, in which case the image to be displayed synchronously is an image showing the document. In this case, the synchronous display may be such that, for example, when one of the users who provides the document, either the customer or the person in charge, performs a page forward operation on the document, the page forward operation is also performed on the other user. Furthermore, if necessary, the customer terminal 5 is also controlled by the interview support device 1. The interview support device 1, staff terminal 2, and customer terminal 5 can communicate with each other via a network 7, enabling the interview support device 1 to perform synchronous display control for presenting various types of information to the customer terminal 5.

[0022] Also, a message 35 may be displayed on the salesperson screen 31 as shown in Fig. 2B. For example, in the case of this embodiment, the interview support device 1 estimates the potential demand of the customer, and displays a message suggesting to the salesperson to recommend a certain product. This allows the salesperson to know the product to be proposed to the customer during the business negotiation without the customer being aware of it.

[0023] 1 shows a microphone 4. The microphone 4 is assumed to be placed, for example, at a place where face-to-face business negotiations are taking place. The microphone 4 collects the voices of the person in charge and the customer during the business negotiation. The voice data obtained by the microphone 4 is transmitted as business negotiation information to the interview support device 1 by the transmitter 6. Communication between the transmitter 6 and the interview support device 1 is possible via network communication using a wired transmission path or a wireless transmission path, or dedicated line communication.

[0024] In actual business negotiations, the person in charge uses a terminal device to display documents and other information, present them to the customer, and keep records of the negotiations. In other words, it is common to use a tablet terminal or a personal computer. Therefore, the microphone 4 and the transmitter 6 referred to here may be considered to be a microphone and a communication device section built into the person in charge terminal 2.

[0025] The interview support device 1 stores voice data and image data for each business negotiation in a database 8 as business negotiation information. The interview support device 1 also performs a learning process based on the business negotiation information of each business negotiation stored in the database 8, and generates learning data. As will be described in detail later, the learning data is data indicating the relationship between the content of the customer's utterance and the customer's final selection result. In the case of this embodiment, the final selection result is information regarding what product the customer purchased after the business negotiation. The interview support device 1 also acquires the customer's speech during the ongoing interview, and refers to the learning data to estimate what kind of product the customer wants, and proposes a product to the salesperson, for example, in the form of the message 35 described above.

[0026] <2. Configuration of information processing device as interview support device> FIG. 3 is a block diagram showing an example of the hardware configuration of an information processing device that can function as the interview support device 1.

[0027] As shown in Fig. 3, the information processing device includes a control unit 11, a ROM 12, and a RAM 13. The control unit 11 is composed of, for example, a CPU, and functions as an arithmetic processing unit that performs various processes, and executes various processes according to a program stored in the ROM 12 or a program loaded from the storage unit 19 to the RAM 13. The RAM 13 also stores data necessary for the CPU to execute various processes as appropriate. The control unit 11 may be composed of multiple CPUs, an AI (Artificial Intelligence) processor, or the like.

[0028] The control unit 11, the ROM 12, and the RAM 13 are interconnected via a bus 14. To this bus 14, an input / output interface 15 is also connected.

[0029] An input unit 16 consisting of operators and operation devices is connected to the input / output interface 15. For example, the input unit 16 may be various operators and operation devices such as a keyboard, a mouse, a key, a touch panel, a touch pad, a dial, a remote controller, or the like. An operation is detected by the input unit 16, and a signal corresponding to the detected operation is interpreted by the control unit 11.

[0030] Further, the input / output interface 15 is connected, either integrally or separately, to a display unit 17 formed of an LCD (Liquid Crystal Display) or an organic EL (Electro-Luminescence) panel, or the like, and an audio output unit 18 formed of a speaker, or the like. The display unit 17 is used to display various types of information, and is configured, for example, by a display device provided in the housing of the information processing device, or a separate display device connected thereto.

[0031] The display unit 17 executes display of images for various image processing, moving images to be processed, etc. on the display screen based on instructions from the control unit 11. Furthermore, the display unit 17 performs display of various operation menus, icons, messages, etc., that is, as a GUI (Graphical User Interface), based on instructions from the control unit 11.

[0032] A storage unit 19 and a communication unit 20 may be connected to the input / output interface 15 . The storage unit 19 is configured with a hard disk drive (HDD) or a solid state drive (SSD), and stores various types of information. A database may be constructed in the storage unit 19. For example, the database 8 required for processing as the interview support device 1 may be constructed in the storage unit 19.

[0033] The communication unit 20 performs communication processing via a transmission path such as the Internet, and communication with various devices via wired / wireless communication, bus communication, and the like.

[0034] A drive 22 is also connected to the input / output interface 15 as required, and a removable recording medium 21 such as a memory card or an optical disk is appropriately attached thereto.

[0035] The drive 22 allows data files such as programs used in each process to be read from the removable recording medium 21. The read data files are stored in the storage unit 19, and images and sounds contained in the data files are output on the display unit 17 and the sound output unit 18. In addition, the computer programs and the like read from the removable recording medium 21 are installed in the storage unit 19 as necessary.

[0036] In an information processing device having the above hardware configuration, for example, software for processing as the interview support device 1 of this embodiment can be installed via network communication by the communication unit 20 or via the removable recording medium 21. Alternatively, the software may be stored in advance in the ROM 12, the storage unit 19, etc. The control unit 11 performs processing operations based on various programs, thereby executing the necessary information processing and communication processing of the interview support device 1.

[0037] Although the above describes an example of an information processing device that functions as the interview support device 1, the staff terminal 2 and the customer terminal 5 can also be realized by information processing devices having a similar configuration. The microphone 4 and the transmitter 6 may also be configured as built-in devices of an information processing device having a similar configuration.

[0038] Furthermore, the interview support device 1 is not limited to being configured with a single information processing device as shown in Fig. 3, but may be configured with a system of multiple information processing devices. The multiple information processing devices may be systemized using a LAN (Local Area Network) or the like, or may be located in a remote location using a VPN (Virtual Private Network) using the Internet or the like. The multiple information processing devices may include information processing devices serving as a server group (cloud) available through a cloud computing service.

[0039] Assuming that the interview support device 1 is an information processing device, the control unit 11 has functions as a learning processing unit 11a and a proposal processing unit 11b as shown in the figure. For example, these functions are realized in the control unit 11 by a program installed in the information processing device.

[0040] The learning processing unit 11a has the function of performing an acquisition process to acquire information on the purchased product, which is the final selection result by the customer in a past interview, an extraction process to extract only the customer's speech content from the speech content during the interview, and a learning process to set learning input data from the extracted customer's speech content and learn the correlation and affinity between the purchased product and the speech content.

[0041] The proposal processing unit 11b has functions of performing an acquisition process for acquiring conversational voice of an ongoing interview, an extraction process for extracting only the customer's speech content from the conversational voice, a determination process for determining a product related to the extracted customer's speech content by referring to the learning data generated / updated by the learning processing unit 11a, and a process for transmitting proposal information to, for example, the staff terminal 2 during the ongoing interview based on the determination result. The proposal information is information on a product that the customer is potentially looking for.

[0042] In addition, there are cases where the interview support device 1 is provided with the learning processing unit 11a and not provided with the proposal processing unit 11b. Conversely, there are cases where the interview support device 1 is provided with the proposal processing unit 11b and not provided with the learning processing unit 11a. For example, there are cases where the interview support device 1 having the learning processing unit 11a and the interview support device 1 having the proposal processing unit 11b are separate devices.

[0043] <3. First embodiment> An example of processing performed by the interview support device 1 according to the first embodiment will be described. First, an overview will be given. The processing in this embodiment is roughly divided into a learning processing by the learning processing unit 11a and a proposal processing by the proposal processing unit 11b.

[0044] The learning process learns from the information of multiple past business negotiations, and generates or updates learning data on the relationship between the conversation in the business negotiation and the final selection result of the customer as a result of the business negotiation, i.e., the purchased product. In this case, learning is performed by extracting the content of the customer's utterances from the conversation between the sales representative and the customer in the business negotiation.

[0045] Generally, a salesperson will negotiate with a customer to sell a certain product, and during the negotiation, they will have various conversations and explanations. As a result, the customer may or may not end up buying the product, but there are also cases where the customer purchases a product different from the one the salesperson initially recommended.

[0046] Figure 4 shows the products that sales representatives initially proposed to various customers A, B, C, etc., and the products that the customers ultimately purchased. The initially proposed products and the final purchased products may or may not be the same. Information from sales negotiations in various cases like these is used for learning. By paying particular attention to the content of customer utterances, learning data can be generated that can determine the customer's latent requests and desires.

[0047] For example, in a business negotiation with Customer A, utterances such as "I'm interested in trends in the world" and "I want to support the younger generation" are extracted from the customer's statements. In this case, the relationship between these utterances and the final purchased product, "stocks," is learned. Similarly, in negotiations with Customer B, the system learns the relationship between the customer's speech, such as "I want to support the younger generation," and the final purchased product, "stocks." Similarly, in negotiations with Customer C, the system learns the relationship between speech such as "I'm worried about the future" and "I don't know when I'll be unable to work" and the final product purchased, "insurance."

[0048] For example, by repeatedly learning from past sales negotiations in this way, learning data is constructed that shows the affinity and correlation between the final purchased product and various customer utterances. For example, Figure 6 shows an example of learning data showing affinity scores for various utterances SP1, SP2, SP3, etc. and various products M1, M2, M3, etc. For example, the affinity score, with 100 being the maximum and 0 being the minimum, can be said to be a score that shows the degree to which each utterance appeared in a sales negotiation with respect to the purchased product.

[0049] FIG. 6 is merely an example of learning data shown diagrammatically, but the learning process involves generating such learning data and successively inputting new business negotiation information to update the learning data. The above-mentioned "stocks" and "insurance" are merely examples of product types, and more detailed classifications, such as "life insurance," "earthquake insurance," "fire insurance," etc., may be used for insurance. Needless to say, the actual learning data will be learned for each product type in accordance with the products of the company using the interview support device 1.

[0050] Next, the proposal process is a process in which, while a business negotiation is being carried out, the learning data is referred to and a product to be recommended to the customer is proposed, for example, to a sales representative (or directly to the customer). By referencing the learning data, the system can predict what products the customer is looking for based on the customer's current speech. The system then communicates the predicted products to the sales representative, encouraging them to recommend the products to the customer.

[0051] For example, as shown in Fig. 5, suppose that the contents of a customer's speech during a business negotiation include "I'm worried about the future" and "I don't know when I will be unable to work." In this case, by referring to the learning data, it is determined that "insurance" is a product that has a high affinity with these contents of the speech. In this case, the interview support device 1 presents a message 35 such as "It would be good to recommend insurance" to the salesperson. Also, suppose that the speech content of a customer during a business negotiation is recognized as "I want to support the younger generation." In this case, by referring to the learning data, it is determined that "stocks" are a product that has a high affinity with the speech content. In this case, the interview support device 1 presents the sales representative with a message 35, for example, "It would be good to recommend stocks."

[0052] In the interview support device 1, the learning process and the proposal process are carried out as described above. In the present embodiment, the input data for the learning process (learning input data) is selected from the following viewpoints (1) and (2).

[0053] (1) Extract only the customer's utterances from past sales negotiations in which a certain product was sold. In other words, actively exclude the sales representative's utterances from the learning input data. (2) For each product, a type of conversation content that is not correlated or compatible with that product is set as a “general conversation content.” For past sales negotiations, customer utterances that correspond to the general conversation content for the product that was sold are not used as learning input data.

[0054] Figure 7 shows examples of general conversation content according to the product. In Figure 7, general conversation contents are illustrated for each product, such as real estate investment, stock investment, insurance purchase, car purchase, etc. For real estate investment, general conversation contents include, for example, talk about seasons and weather, pastimes and hobbies, health, and food. Usually, conversations between salespeople and customers in sales negotiations include topics that are completely unrelated to the product. Therefore, the content of customer utterances on such topics is excluded from the learning target. For example, the topic of the weather is often discussed in sales negotiations, so it is usually a general conversation content that is unrelated to the product.

[0055] However, what is considered general conversation content for many products may be important for some products. For example, the topic of weather may be highly relevant for products such as disaster insurance and home insurance that deal with extreme weather. Therefore, general conversation content is set for each product, as shown in Figure 7.

[0056] The interview support device 1 stores, for example, setting information on such general conversation contents and uses it to select learning input data.

[0057] Figure 7 also shows examples of "specific conversation content" for each product. The specific conversation content is a type of conversation content that is determined to have a high correlation with the purchase of that product. For example, as a result of the learning process, the specific conversation content for each product may be determined and stored. For example, among the learning data in FIG. 6, the utterance content having an affinity score value equal to or greater than a predetermined value may be stored as the specific conversation content. Information on the specific conversation content according to this product may also be considered as part of the learning data. In particular, it is desirable to select the specific conversation content based on a criterion that the conversation content is spoken in a certain commercial product but is not often spoken in other commercial products. Such specific conversation content information can be used during the proposal process.

[0058] It is considered that the interview support device 1 successively updates the types of conversation content corresponding to general conversation content and specific conversation content as shown in FIG. 7 in accordance with the learning process.

[0059] An example of processing performed by the control unit 11 of the interview support device 1 will now be described. FIG. 8 shows an example of learning processing by the function of the learning processing unit 11a of the control unit 11.

[0060] As a premise, each time a business negotiation is held, the interview support device 1 stores business negotiation data of that business negotiation in the database 8. The business negotiation data includes conversation data that took place during the business negotiation. The conversation data is voice data of the conversation voice during the business negotiation, or text data obtained by converting the voice data into text, or both. The business negotiation data also includes information related to the business negotiation, such as the date and time, the name of the person in charge, the progress status of the business negotiation, the progress stage, the customer name, the customer attributes, and the customer's final selection result (for example, the final product purchased).

[0061] As a learning process, the control unit 11 of the interview support device 1 acquires the business negotiation information for each past business negotiation, and performs the process of Fig. 8. For example, the process of Fig. 8 may be performed as a batch process once a day.

[0062] In step S101, the control unit 11 selects, from the past negotiation data stored in the database 8, unprocessed negotiation data for learning processing.

[0063] After selecting certain business negotiation data, the control unit 11 determines in step S102 whether the business negotiation has currently resulted in a contract (whether the customer has purchased any merchandise). If the deal has not been concluded, the control unit 11 proceeds from step S102 to step S111 to determine whether the business negotiation has already been concluded. Some business negotiations are concluded in one go, while others are conducted several times before reaching a deal. Therefore, even if a business negotiation has been conducted in the past, it may not have been concluded.

[0064] For example, the process proceeds to step S111 when the negotiation data selected in step S101 is either for a negotiation that ended in failure or for a negotiation that has not yet ended. If the negotiation data selected in step S101 is for a negotiation that ended without a contract, the control unit 11 proceeds to step S110, sets the negotiation data as processed in terms of the learning process, and proceeds to step S112. If the selected negotiation data is for a negotiation that has not yet ended, the control unit 11 proceeds to step S112, does not set the negotiation data as processed in terms of the learning process, and causes the data to be treated as unprocessed the next time the process of Fig. 8 is executed. This is to avoid excluding the negotiation data from the learning process because there is a possibility that the negotiation will be concluded at the next time or later.

[0065] If the business negotiation data selected in step S101 is related to a business negotiation that has led to a contract, the control unit 11 proceeds from step S102 to step S103, and obtains information on the product purchased by the customer from the business negotiation data.

[0066] In step S104, the control unit 11 acquires conversation data from the business negotiation data. If the conversation data consists of only voice data, the control unit 11 may convert the conversation data into text data at this stage.

[0067] In step S105, the control unit 11 separates the conversation data into the sales representative's speech portion and the customer's speech portion, and sets the voice or text data of only the customer's speech portion as the target of learning processing.

[0068] In step S106, the control unit 11 acquires general conversation content and specific conversation content according to the purchased product from the customer's utterance content. That is, the control unit 11 acquires information on the type of conversation content that corresponds to the general conversation content and the specific conversation content set as shown in FIG. 7 according to the information on the purchased product acquired in step S103.

[0069] In step S107, the control unit 11 performs an analysis process on the contents of the customer's utterances extracted in step S105. That is, the control unit 11 determines the contents of each of the customer's utterances during the interview and categorizes them to some extent.

[0070] The categorization will be explained. For example, suppose a customer during a business negotiation utters the following: "But I care about young people and I want to support them in some way." "Are there any programs that can support entrepreneurs in their 20s?" "There are not many young people because of the aging population, so I want them to do their best." Although spoken language is diverse, the meaning of these utterances is determined through analysis and categorized into sentence meanings such as "I want to support the younger generation."

[0071] Similarly, for example, What should I do in the future if my only income is my pension? "When I think about what will happen in 10 years' time, I don't feel very reassured." For example, utterances such as the above are categorized into sentences with the meaning "I am worried about the future."

[0072] In step S108, the control unit 11 sets learning input data from the business negotiation to be processed. For example, assume that the final purchase item in the business negotiation data to be processed is real estate investment. In this case, in the example of Fig. 7, talk about seasons and weather, pastimes and hobbies, health, food, etc. are considered to be general conversation content. Therefore, if the utterance content categorized as described above falls within the category of these general conversation contents, the utterance itself or the categorized utterance content is not included in the learning input data. Among the customer's utterance contents, such utterance contents corresponding to general conversation contents are excluded, and the remaining utterance contents (or categorized utterance contents) are used as learning input data.

[0073] In step S109, the control unit 11 inputs the utterance content that has been set as the learning input data into the learning process as utterance content that may be related to the purchased merchandise acquired in step S103. The utterance content as the learning input data may be the utterance content itself (the words spoken by the customer) or the utterance content categorized as described above. Considering the efficiency of the learning process, it is preferable to use a pair of the categorized utterance content and the purchased merchandise as the learning input data.

[0074] Then, the learning input data is used to generate or update learning data. There are various possible specific learning processes, but one idea is to input a large amount of information on combinations of categorized utterance content and purchased products and have the AI ​​processor perform machine learning. As a result, an affinity score for each combination, as shown in Figure 6, is generated and updated. In addition, each speech content (SP1, SP2, . . . ) in the learning data as shown in FIG. 6 is categorized, which is suitable for improving the efficiency of the proposal process described later.

[0075] The learning process algorithm is not particularly limited. In this embodiment, it is sufficient that learning data is generated and updated by some learning algorithm using learning input data obtained under the above conditions (1) and (2).

[0076] In step S110, the control unit 11 sets the negotiation data currently being processed as having been processed in terms of the learning process, and then the control unit 11 proceeds to step S112.

[0077] In step S112, the control unit 11 determines whether the current learning process is complete. For example, when the selection in step S101 is completed for all unprocessed negotiation data stored in the database 8, the current learning process is completed. If not completed, the process returns from step S112 to step S101, and other unprocessed negotiation data is selected and the process from step S102 onwards is performed.

[0078] The interview support device 1 accumulates sales negotiation data on various sales negotiations, and by performing the above learning process, for example, periodically, the affinity score is updated for each combination of the customer's speech content and the product, thereby improving the learning accuracy.

[0079] In the above learning process, the information on general conversation contents shown in Fig. 7 is used to select the learning input data. In addition, information on specific conversation contents is obtained as the learning result. It is desirable to update this information on general conversation contents and specific conversation contents for each product on a regular basis.

[0080] For example, conversation content that has a similar affinity score for a certain number of products through the learning process can be added to the general conversation content for those products. Also, conversation content that was considered to be general conversation content for a wide range of products but has a high affinity score for only one product due to updates to the learning data can be removed from the general conversation content.

[0081] In addition, in the learning data updated by the learning process, if a conversation content has a high affinity score only for a certain product but is not considered to be a specific conversation content, it can be added to the specific conversation content for that product. Also, if a conversation content was considered to be a specific conversation content for a certain product but has a high affinity score for many other products, it can be removed from the specific conversation content for the product.

[0082] Therefore, the control unit 11 sequentially performs the process shown in FIG. In step S301, the control unit 11 checks the general conversation content. For example, the control unit 11 checks the affinity score of the utterance content corresponding to the conversation type listed as the general conversation content for each product in the learning data at that time. As a result, it extracts the content that is determined not to be the general conversation content. In addition, for the type of utterance content that is not listed as the general conversation content, it refers to the learning data and judges whether it should be evaluated as the general conversation content for each product.

[0083] In step S302, the control unit 11 checks the specific conversation content. For example, the control unit 11 checks the affinity score of the utterance content corresponding to the conversation type listed as the specific conversation content for each product in the learning data at that time. As a result, it extracts those that are determined not to be the specific conversation content. In addition, for the type of utterance content not listed as the specific conversation content, it refers to the learning data and determines whether or not it should be evaluated as the specific conversation content for each product.

[0084] In step S303, the control unit 11 judges whether or not an update should be performed based on the results of steps S301 and S302. For example, if there is a conversation type to be removed from the general conversation content for one or more commercial products, or if there is a conversation type to be added to the general conversation content, it is judged that the general conversation content should be updated. Also, if there is a conversation type to be removed from the specific conversation content for one or more commercial products, or if there is a conversation type to be added to the specific conversation content, it is judged that the specific conversation content should be updated.

[0085] If it is determined that an update is to be performed, the control unit 11 proceeds to step S304 and performs the necessary update processing.

[0086] By successively updating the general conversation content in this manner, it is possible to improve the accuracy of the selection of learning input data in the learning process of FIG. Furthermore, by updating the specific conversation content, it is possible to improve the accuracy of the proposal process, which will be described next.

[0087] The proposal process will be explained with reference to FIG. When a certain business negotiation is taking place, the control unit 11 of the interview support device 1 carries out the process of FIG. 10 in response to the business negotiation in real time.

[0088] When the business negotiation starts, the control unit 11 starts inputting conversational voice in step S201. After that, the control unit 11 repeats the process of steps S202 to S207 for the input conversational voice until the business negotiation is completed.

[0089] In step S202, the control unit 11 performs conversation separation. This is a process of separating the input conversation voice (voice data or text data obtained by converting voice into text) into whether it is a speech of a salesperson or a customer. Separation can be performed by determining the speaker based on voiceprints, the voice input path (whether the voice was collected at the staff terminal 2 or the customer terminal 5, etc.), the positioning of the voice collected by microphone 4, or the context.

[0090] In step S203, the control unit 11 analyzes the customer's utterance content extracted by the conversation separation, and then refers to the learning data based on the utterance content obtained as a result of the analysis to determine whether or not there is a highly correlated product. For example, the control unit 11 analyzes the contents of the extracted customer's utterances and performs the above-mentioned categorization. Then, it determines whether the utterance contents are stored as utterance contents SP1, SP2, etc. in the learning data. It also determines whether the current utterance contents are set as specific conversation contents. Then, if it is determined that the utterance contents have high affinity and correlation with a certain product, it determines that the corresponding product is the product to be proposed.

[0091] For example, it may be determined that the analyzed utterance content has an extremely high affinity score with a certain product, or that the utterance content corresponds to a specific conversation content. It may be determined that an utterance with a high affinity score with a certain product has appeared a predetermined number of times or a predetermined number of types, rather than just that one utterance content has a high correlation with a certain product. It may be determined that a utterance corresponding to a certain type or number of times of a specific conversation content has appeared. The affinity score may be accumulated after each utterance content, and a product with an affinity score equal to or greater than a predetermined value may be selected as a product to be proposed.

[0092] In either case, the control unit 11 returns to step S202 via steps S204 and S208 and repeats the analysis of the customer's utterance content and the determination of the product to be proposed until it is determined that there is a product to be proposed. Therefore, the control unit 11 refers to the learning data for each utterance of the customer during the business negotiation, and repeats the determination based on the utterance content using a predetermined determination algorithm.

[0093] The determination algorithm is not particularly limited. In the present embodiment, it is sufficient that the algorithm is capable of determining products with high correlation and affinity with the contents of customer utterances extracted during a business negotiation by using the above-mentioned learning data.

[0094] When the product to be proposed is determined in the process of step S203 at a certain point in time, the control unit 11 proceeds from step S204 to step S205, and transmits proposal information indicating the product to be proposed. For example, the control unit 11 performs a process of displaying a message 35 indicating the content that should be proposed in the business negotiation on the screen of the salesperson terminal 2. This allows the salesperson in the business negotiation to recognize the product to be recommended to the customer. The salesperson can see this and direct the business negotiation policy toward selling the proposed product. The control unit 11 may also cause the proposed product to be displayed on the customer terminal 5.

[0095] In step S206, the control unit 11 enables documents related to the proposed merchandise to be shared and presented on the screens of the staff member terminal 2 and the customer terminal 5. For example, the control unit 11 displays a button or the like that serves as a guide for displaying document sharing on the screen of the staff member terminal 2, so that the documents of the proposed merchandise can be displayed on the customer terminal 5 during negotiations at the discretion of the sales staff member. This allows the sales staff member to smoothly move on to explaining the proposed product.

[0096] In step S207, the control unit 11 performs a process of adding the information on the proposed product and the corresponding keywords, that is, the contents of the customer's utterance, to the minutes data of the business negotiation. For example, the interview support device 1 can automatically create minutes of a business negotiation based on information such as the conversation voice of the business negotiation. In this case, the proposed product and one or more utterances that became keywords in the decision to make the proposal are recorded in the minutes, so that the sales representative and his / her superior can refer to them later.

[0097] During the period when the business negotiation is being carried out, the control unit 11 continues to execute the above-mentioned processing. When the current business negotiation is completed, the control unit 11 ends the processing of FIG. The above proposal process makes it possible to provide interview support by providing useful information to sales representatives during business negotiations.

[0098] In the above processing example, the proposed product is determined using the customer's speech during the business negotiation, and is transmitted to and displayed on the staff member terminal 2 or the customer terminal 5. However, information on the proposed product may be transmitted to the staff member terminal 2 after the business negotiation. For example, after a business negotiation is completed, the control unit 11 performs the process of Fig. 10 using the business negotiation information (voice data of the business negotiation or its text data). Then, when the product to be proposed is determined by referring to the learning data, the proposed information is transmitted to the salesperson terminal 2. This allows the salesperson to know the product that should be proposed in the next business negotiation with the customer.

[0099] Further, the control unit 11 may execute the process in FIG. 10 up to the determination of the product to be proposed during the business negotiation, and transmit the proposal information (step S205) after the business negotiation.

[0100] <4. Second embodiment> An example of processing by the control unit 11 according to the second embodiment will be described with reference to FIGS. Fig. 11 shows an example of the learning process of the control unit 11. Note that the same processes as those in Fig. 8 are given the same step numbers to avoid redundant explanation.

[0101] The following describes a case where unprocessed negotiation data is selected in step S101 and the selected data has led to a contract, that is, the process proceeds from step S102 to step S103. The process proceeds from step S102 to step S111 in the same manner as in FIG.

[0102] When the negotiation data that has led to a contract is selected as the processing target, the control unit 11 acquires information on the product purchased by the customer from the negotiation data in step S103.

[0103] Furthermore, in step S120, the control unit 11 acquires attribute information of the customer from the business negotiation data. The attribute information here is assumed to be age, sex, organization (company, department, etc.), residential area, product purchase record, etc.

[0104] In step S104, the control unit 11 acquires conversation data from the business negotiation data. In step S105A, control unit 11 excludes the conversation during the ice-breaking period from the conversation data, separates the conversation, and extracts the speech portion of the customer. During business negotiations, there is a time period called ice-breaker, where conversations that are not directly related to the sale of the product are held. These conversations are held to ease the minds of both parties and deepen relationships.

[0105] The ice breaker is performed at the beginning of the business negotiation or at an appropriate timing, and the control unit 11 can determine the time period for the ice breaker from the content of the conversation. Also, if the salesperson inputs the progress of the business negotiation or marks it afterwards, the control unit 11 can detect the ice breaker time period based on that information. The control unit 11 excludes the ice-breaker conversation as being not directly related, so that it is not reflected in the learning input data.

[0106] In steps S107 and S108, the control unit 11 acquires general conversation content according to the product and analyzes the customer's utterance content. Then, in step S109A, the control unit 11 generates or updates learning data. In this case, the control unit 11 performs a learning process that associates the product with the utterance content as well as customer attributes. For example, as a result, processing is performed in which learning data as shown in FIG. 6 is generated / updated for each customer attribute. The attributes to be used may be appropriately selected according to the product. For example, learning data may be generated for each age group or gender, or learning data may be generated according to the organization to which the customer belongs.

[0107] In this way, learning data is generated according to the attributes of the customers. In addition, by not using ice-breaker conversations in the learning process, it is possible to avoid excessive input to the learning process or learning from speech content that is not originally relevant.

[0108] An example of the proposed process is shown in Fig. 12. The same processes as those in Fig. 10 are given the same step numbers.

[0109] In step S210, the control unit 11 acquires attributes of the customer of the business negotiation to be carried out. For example, the customer attributes may be acquired as information from a sales representative, or attribute information may be acquired from the past business negotiation data of the customer. Then, when the business negotiation starts, the control unit 11 starts inputting the conversation voice in step S201.

[0110] In step S211, the control unit 11 determines whether or not the ice-breaking is currently in progress. The control unit 11 may determine whether or not the ice-breaking is currently in progress by analyzing the content of the conversation, or may determine the same based on an operation on the sales representative's terminal 2. When the ice-breaking is in progress, the control unit 11 waits without proceeding from step S211 to step S202.

[0111] If it is determined that the ice-breaking session is not in progress, the control unit 11 proceeds to step S202 and subsequent steps. In step S202, the control unit 11 separates the speech of the salesperson and the speech of the customer. In step S203A, the control unit 11 analyzes the customer's speech content extracted by conversation separation. Then, based on the speech content from the analysis result, the control unit 11 refers to the learning data and determines whether or not there is a highly correlated product. In this case, the control unit 11 refers to the learning data according to the attributes of the customer. For example, when learning data is generated for each attribute, the control unit 11 selects and refers to the learning data corresponding to the attributes of the current customer and determines whether or not there is a highly correlated product.

[0112] The processes from step S204 to step S208 are similar to those in FIG. According to the process in Fig. 12, no product suggestions are made based on the conversation during the ice breaker. In addition, the products to be suggested are selected taking into account the attributes of the customer.

[0113] <5. Effects and Modifications of the Embodiments> According to the interview support device 1 explained above as the first and second embodiments, the following effects can be obtained.

[0114] The interview support device 1 of the embodiment is equipped with a control unit 11 (learning processing unit 11a) that performs an acquisition process to acquire information on the final purchased product (final selection result) by the customer during an interview between a sales representative and a customer, an extraction process to extract only the customer's speech content from the speech content between the sales representative and the customer during the interview, and a learning process to set learning input data from the customer's speech content extracted in the extraction process and learn the correlation between the purchased product and the speech content. By learning the customer's speech and the products purchased, it is believed that it is possible to infer the customer's latent desires from the customer's speech during a sales negotiation. In this case, by excluding the speech of the salesperson, it is possible to avoid using sales-related speech, such as speech in which the salesperson recommends a product that they are trying to sell, in the learning process, and it is possible to set learning input data that is useful for inferring the customer's intentions. This makes it possible to build learning data that is useful during sales negotiations, etc.

[0115] In the embodiment, an example has been given in which the control unit 11, in the learning process, excludes general conversation content corresponding to the product from the extracted customer utterance content and sets the learning input data. By excluding general conversational content that is unrelated to the final purchased product from the customer utterances contained in the sales negotiation data from the learning input data, the efficiency and reliability of the learning process can be improved.

[0116] In the embodiment, the type of general conversation content is set for each type of final purchased product. It is believed that what kind of conversation is considered to be general conversation content unrelated to the product varies depending on the type of product. Therefore, general conversation content is set for each type of product, and whether or not the customer's utterance is general conversation content is determined depending on the type of product as the final selection result. This can further improve the efficiency and reliability of the learning process.

[0117] In the embodiment, an example is given in which the control unit 11, in the learning process, sets learning input data by excluding the content of utterances during a specific time period during the interview, such as the ice-breaking time period, from the extracted customer utterances. By excluding general conversational content unrelated to the product from the learning input data, the efficiency and reliability of the learning process can be improved. The specific time period to be excluded is not limited to ice-breaking. For example, utterances during the time period after a contract is concluded, during the time period when contract documents are being prepared, etc. may be excluded from the learning input data.

[0118] In the embodiment, an example has been given in which the control unit 11 learns the correlation between the last purchased product, the utterance content, and the guest attributes in the learning process. It is expected that the relevance between the content of a conversation and a product will vary depending on the customer's attributes. Therefore, by adding customer attributes to the correlation learning, more reliable learning data can be generated.

[0119] The interview support device 1 of the embodiment is equipped with a control unit 11 (proposal processing unit 11b) that performs an acquisition process to acquire conversation audio of an interview between a sales representative and a customer (for example, conversation audio of a current interview between a sales representative and a customer), an extraction process to extract only the customer's speech content from the conversation audio, a determination process to determine a product related to the speech content extracted in the extraction process by referring to learning data generated in a learning process that learns the correlation between the final purchased product (final selection result) of a past interview and the customer's speech content, and a process to send proposal information to a terminal device based on the determination result. This will make it possible to recommend products based on learning data from what the customer says during a sales negotiation.

[0120] In the embodiment, an example has been given in which the control unit 11 performs a process of transmitting proposal information to the salesperson terminal 2 during negotiation. This makes it possible to provide effective support in real time to the salesperson during negotiation. It has also been mentioned that the control unit 11 transmits the proposal information to the salesperson terminal 2 after the business negotiation is completed. This makes it possible to provide the salesperson with appropriate information to prepare for the next business negotiation.

[0121] In the embodiment, it has been described that the control unit 11 may transmit the proposal information to the customer terminal 5. During or after the business negotiation, directly proposing to the customer products that the customer may want can also be an aid for making the business negotiation proceed more smoothly. The process of transmitting proposal information to the client terminal 5 to propose the product may be performed during negotiation, but may also be performed in a form in which materials for the proposed product are transmitted after the negotiation is completed.

[0122] The program of the embodiment is a program that causes the control unit 11 to execute an acquisition process for acquiring information on the final purchased item by the customer from an interview between a sales representative and the customer, an extraction process for extracting only the customer's utterance content from the conversation content between the sales representative and the customer during the interview, and a learning process for setting learning input data from the customer's utterance content extracted in the extraction process and learning the correlation between the final purchased item and the conversation content. For example, the program causes the control unit 11 to execute the processes of Figs. 8 and 11.

[0123] The program of the embodiment is a program that causes the control unit 11 to execute an acquisition process for acquiring a conversation voice of an interview between a sales representative and a customer, an extraction process for extracting only the customer's utterance content from the conversation voice, a determination process for determining a final purchased product related to the utterance content extracted in the extraction process by referring to learning data generated in a learning process for learning the correlation between the final purchased product of a past interview and the customer's utterance content, and a process for transmitting proposal information to a terminal device (such as the sales representative terminal 2 during the interview or the customer terminal 5) based on the determination result. For example, the program causes the control unit 11 to execute the processes of Figs. 10 and 12.

[0124] Such programs can realize an information processing device that executes all or part of the functions of the interview support device 1 described in the embodiment. For example, by downloading a program to a personal computer, a portable information processing device, or the like, the personal computer or the like can be made to function as the interview support device 1.

[0125] Such a program can be recorded on a recording medium. Examples of recording media include a HDD built into a device such as a computer device, a ROM in a microcomputer, a non-volatile memory, etc., and the program can be recorded in advance on these recording media. In addition, the program can be provided by temporarily or permanently recording it on a removable recording medium such as a memory card or an optical disk. In addition, such a program can be downloaded from a download site via a network such as a LAN (Local Area Network) or the Internet.

[0126] The interview support device 1 of the embodiment performs learning processing and proposal processing to support business negotiations between a person in charge and a customer, but the technology of the interview support device 1 is not limited to so-called business negotiations, and can be applied to estimating the intentions of a guest during various interviews in the fields of medicine, education, consulting, counseling, etc. For example, a case can be considered in which the potential wishes of a client who is the guest are estimated and presented.

[0127] Furthermore, the final choice made by a guest as a result of an interview, such as the final product to be purchased, is not necessarily limited to one. For example, a customer may purchase product A and product B as a result of a business negotiation. This technology is effective even in such cases. For example, the system can learn the relationship between the customer's speech during the business negotiation and product A, and then learn the relationship between the customer's speech during the business negotiation and product B. This makes it possible to carry out appropriate learning even when there are multiple final selection results (final purchased products). Furthermore, based on the learning data generated in this way, it becomes possible to propose products to the customer during and after the interview. [Explanation of symbols]

[0128] 1 Interview support device 2. Personnel terminal 4 Microphones 5. Customer terminal 6 Transmitter 7 Network 8 Database 11 Control section 11a Learning processing unit 11b Proposal processing section

Claims

1. An acquisition process for acquiring information on the final selection result by the guest regarding the interview between the host and the guest; An extraction process for extracting only the guest's speech content from the speech content of the host and the guest during the interview; a learning process for setting learning input data from the guest speech content extracted in the extraction process and learning a correlation between the final selection result and the speech content; A control unit for performing Interview support device.

2. In the learning process, the control unit sets learning input data by excluding general conversation content that is not a target for determining the correlation with the final selection result from among the speech content of the guest extracted in the extraction process. The interview support device according to claim 1.

3. The type of the general conversation content is set for each type of the final selection result. The interview support device according to claim 2.

4. In the learning process, the control unit sets learning input data by excluding utterance contents in a specific time period in the interview from the utterance contents of the guest extracted in the extraction process. The interview support device according to any one of claims 1 to 3.

5. In the learning process, the control unit The correlation between the final selection result, the speech content, and the guest's attributes is learned. The interview support device according to any one of claims 1 to 3.

6. An acquisition process for acquiring information on the final selection result by the guest regarding the interview between the host and the guest; An extraction process for extracting only the guest's utterance content from the utterance content of the host and the guest during the interview; a learning process for setting learning input data from the guest speech content extracted in the extraction process and learning a correlation between the final selection result and the speech content; The interview support method is executed by an information processing device.

7. An acquisition process for acquiring information on the final selection result by the guest regarding the interview between the host and the guest; An extraction process for extracting only the guest's speech content from the speech content of the host and the guest during the interview; a learning process for setting learning input data from the guest speech content extracted in the extraction process and learning a correlation between the final selection result and the speech content; A program for causing an information processing device to execute the above.

8. An acquisition process for acquiring a conversation voice of an interview between a host and a guest; An extraction process of extracting only the guest's speech content from the conversation voice; A determination process for determining a final selection result related to the utterance content extracted in the extraction process by referring to learning data generated in a learning process that learns the correlation between the final selection result of the past interview and the guest's utterance content; A process of transmitting proposal information to a terminal device based on a determination result in the determination process; A control unit for performing Interview support device.

9. The control unit is The proposed information based on the result of the judgment process is transmitted to the terminal device of the interview host during or after the interview. The interview support device according to claim 8.

10. The control unit is A process of transmitting proposal information based on the determination result in the determination process to the terminal device of the guest is performed. The interview support device according to claim 8 or 9.

11. An acquisition process for acquiring a conversation voice of an interview between a host and a guest; An extraction process of extracting only the guest's speech content from the conversation voice; A determination process for determining a final selection result related to the utterance content extracted in the extraction process by referring to learning data generated in a learning process that learns the correlation between the final selection result of the past interview and the guest's utterance content; A process of transmitting proposal information to a terminal device based on a determination result in the determination process; The interview support method is executed by an information processing device.

12. An acquisition process for acquiring a conversation voice of an interview between a host and a guest; An extraction process of extracting only the guest's speech content from the conversation voice; A determination process for determining a final selection result related to the utterance content extracted in the extraction process by referring to learning data generated in a learning process that learns the correlation between the final selection result of the past interview and the guest's utterance content; A process of transmitting proposal information to a terminal device based on a determination result in the determination process; A program for causing an information processing device to execute the above.

Images