Conversation determination apparatus, conversation determination method, and program

Abstract

To allow a user to confirm whether explanation contents are suitable based on contents of a conversation in the explanation of a client upon contracting.SOLUTION: A conversation determination apparatus comprises: a voice information acquisition unit for acquiring voice information obtained by recording the voice in the conversation of a user; a conversation information output unit for outputting conversation information based on the voice information to a conversation analysis device; an analysis result acquisition unit for acquiring conversation analysis results by the conversation analysis device at least including determination results as to whether required items are spoken in the above conversation; and a presentation unit for presenting proposal contents based on the above determination results to the user.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a conversation determination device, a conversation determination method, and a program. [Background technology]

[0002] BACKGROUND ART Conventionally, techniques have been disclosed for obtaining appropriate explanation contents when explaining to customers at the time of signing a contract in the finance, insurance, real estate, and other industries (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-096053 Summary of the Invention [Problem to be solved by the invention]

[0004] Here, it is preferable that the person in charge of the contract (for example, a company sales representative) can check whether the explanation content is appropriate based on the content of the conversation that took place when explaining the contract to the customer. However, with the above-described conventional technology, it is not possible to check whether the contents of the explanation given to the customer at the time of signing the contract are appropriate based on the contents of the conversation.

[0005] The present invention has been made to solve the above problem, and its purpose is to provide a conversation determination device, a conversation determination method, and a program that can confirm whether the content of an explanation given to a customer at the time of signing a contract is appropriate based on the content of the conversation. [Means for solving the problem]

[0006] One embodiment of the present invention is a conversation determination device comprising: an audio information acquisition unit that acquires audio information obtained by capturing the audio of a user's conversation; a conversation information output unit that outputs conversation information based on the audio information to a conversation analysis device; an analysis result acquisition unit that acquires a conversation analysis result that is an analysis result by the conversation analysis device, the conversation analysis result including at least a determination result as to whether or not essential items in the conversation have been uttered; and a presentation unit that presents suggested content based on the determination result to the user.

[0007] One embodiment of the present invention is a conversation determination method that includes: a computer provided in a conversation determination device acquires audio information obtained by capturing the audio of a user's conversation; outputs conversation information including the audio information to a conversation analysis device; acquires an analysis result by the conversation analysis device that includes at least a determination result as to whether or not essential items in the conversation have been spoken; and presents suggestions based on the determination result to the user.

[0008] One embodiment of the present invention is a program for causing a computer to perform the following operations: acquire audio information obtained by capturing the audio of a user's conversation; output conversation information including the audio information to a conversation analysis device; acquire a conversation analysis result by the conversation analysis device that includes at least a determination result as to whether or not essential items in the conversation have been spoken; and present suggestions to the user based on the determination result. [Effects of the Invention]

[0009] According to this invention, it is possible to check whether the contents of the explanation are appropriate based on the contents of the conversation given to the customer when making a contract. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a diagram illustrating an example of a device configuration of a conversation determination system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a diagram illustrating an example of a functional configuration of the conversation determination device according to the present embodiment. [Figure 3]10 shows an example of an industry-specific template stored in a storage unit of the present embodiment. [Figure 4] FIG. 2 is a diagram showing an example of the flow of operations of the conversation determination device of the present embodiment. [Figure 5] FIG. 2 is a diagram illustrating an example of a conversation according to the present embodiment. [Figure 6] FIG. 10 is a diagram showing an example of an analysis result image presented by a presentation unit of the present embodiment. [Figure 7] FIG. 10 is a diagram showing another example of the analysis result image presented by the presentation unit of the present embodiment. [Figure 8] FIG. 10 is a diagram illustrating an example of a management image displayed on the management device of the present embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. 1 is a diagram showing an example of the device configuration of a conversation determination system 1 according to this embodiment. The conversation determination system 1 determines whether or not a conduct risk (hereinafter simply referred to as a risk) has occurred in a conversation, for example, when a company sales representative concludes a contract to provide a product or service to a customer.

[0012] [About conduct risk] Conduct risk is a term used in business industries such as finance, insurance, and real estate to refer to the responsibility of companies to behave fairly and appropriately toward their customers. Specifically, it refers to the risks associated with transactions with customers and the services they provide. Conduct risk can also refer to the risks arising from actions or practices that could cause harm or loss to customers. Conduct risk is a particularly important concept in the financial, insurance, and real estate industries. Conduct risk is considered to increase when inappropriate behavior or incorrect information provided by a company (especially its sales representatives) could have a negative impact on customers. Examples of conduct risk include proposing inappropriate products to customers, signing contracts without explaining their contents, and concealing important information. Properly managing conduct risk and treating customers appropriately is important to enhance the credibility of the industry and avoid legal trouble and reputational damage. By emphasizing compliance, ethics, and transparency, companies are required to minimize conduct risk and build trust with customers.

[0013] As an example, conduct risks in the insurance industry include: (1) Inappropriate provision of information For example, this can occur when the insurance plan and coverage details are not explained properly or incorrect information is provided, leading to misunderstandings about the payment conditions and excluded risks, leading to problems such as customers expecting insurance payments. (2) Insurability assessment For example, this can occur if there is a flaw in the process for verifying the compatibility of insurance products with customer needs. This can lead to problems such as an insurance product being proposed that does not match the customer's driving history, resulting in a decrease in satisfaction. (3) Clarity of the contract For example, this can occur when the contract contents and exclusions are difficult to understand. The limitations on the scope of coverage are unclear, which can lead to problems such as insurance claims being denied in the event of an accident. (4) Delays in processing claims For example, this can occur when there are delays in insurance payments and claims processing, or when there are delays in payment of car repair costs after an accident, forcing customers to bear the financial burden. (5) Privacy and Security For example, this occurs when customer information is not properly protected, and if customer information is leaked, there is a risk that personal information may be misused.

[0014] As described above, conduct risk can lead to a decrease in a company's sales and profits, damage to the brand image, and the like. For this reason, it is preferable to reduce conduct risk. The conversation determination system 1 of this embodiment reduces conduct risk by recording and analyzing conversations (e.g., the content of sales proposals) between a company's sales representative and a customer when they conclude a contract to provide a product or service. The configuration of the conversation determination system 1 will be described below.

[0015] The conversation determination system 1 includes a conversation determination device 10. The conversation determination device 10 is, for example, a personal computer, and is operated by, for example, a sales representative 2A. The conversation determination device 10 includes a sound collection unit 20 and a display unit 30. The sound collection unit 20 includes, for example, a microphone, and collects the conversation between the sales representative 2A and the contractor 3. Although the diagram shows an example in which the conversation determination device 10 and the sound collection unit 20 are separate devices, this is not limiting. The sound collection unit 20 may be built into the conversation determination device 10 (i.e., the conversation determination device 10 and the sound collection unit 20 are integrated). The display unit 30 includes, for example, a liquid crystal display, and displays various images.

[0016] The conversation determination device 10 is connected to a management device 11, a conversational voice recognition device 40, a transcription device 50, and a conversation analysis device 60 via a network NT (for example, the Internet). The management device 11 is, for example, a personal computer used by an administrator 2B (for example, a manager who is the superior of the sales representative 2A). The administrator 2B can use the management device 11 to grasp the results of the sales activities of the sales representative 2A. In the following description, the sales representative 2A and the manager 2B are collectively referred to as the user 2 of the conversation determination system 1 (or simply as the user 2).

[0017] The conversational speech recognition device 40, the transcription device 50, and the conversation analysis device 60 are all server devices (e.g., cloud servers). The conversation determination device 10 can exchange information with the conversational speech recognition device 40, the transcription device 50, and the conversation analysis device 60 via the network NT. The functional configuration and operational flow of conversation determination device 10 will be described below.

[0018] FIG. 2 is a diagram showing an example of the functional configuration of conversation determination device 10 of this embodiment. The conversation determination device 10 includes a calculation unit 110 and a storage unit 150. The calculation unit 110 includes, for example, a CPU (Central Processing Unit) and provides various functions based on programs and data stored in the storage unit 150. The storage unit 150 includes, for example, a nonvolatile semiconductor memory or a hard disk drive, and stores programs and data for operating the calculation unit 110. The storage unit 150 stores, for example, an industry-specific template 151.

[0019] 3 shows an example of an industry-specific template 151 stored in the storage unit 150 of this embodiment. The industry-specific template 151 is information in which, for each type of conversation, an explanation item 1511 is associated with a determination condition 1512. Here, the type of conversation is, for example, the type of business in which the conversation takes place (i.e., the type of business). The business types include, for example, finance, insurance, real estate, etc. In the following description, the business type template 151 is also simply referred to as a template.

[0020] The explanation items 1511 are items that should be explained in the conversation about the contract for the relevant industry. The judgment condition 1512 is a condition for judging whether the explanation items 1511 have been appropriately explained in the conversation about the contract for the relevant industry. The judgment condition 1512 includes at least one required item 1513, which is an item that must be explained in a conversation about a contract in the relevant industry. In the example shown in the same figure, the explanation item 1511 "(1) The entire important matters explanation document should be explained" includes two required items 1513: a first required item 1513-1 "The important matters explanation document should contain important matters related to the contract" and a second required item 1513-2 "The items in the important matters explanation document marked with 'contract summary' and 'warning information' contain particularly important content." In order to reduce conduct risk, the sales representative 2A is required to explain these essential items 1513 to the contractor 3 during the contract conversation.

[0021] Here, some of the factors that cause conduct risks differ depending on the type of business. Storage unit 150 may store industry-specific templates 151 for each type of business. That is, storage unit 150 may store explanatory items 1511, determination conditions 1512, and required items 1513 as templates that differ for each type of business. These may differ depending on the type of business. Conversation determination system 1 configured in this way can accommodate a variety of types of business.

[0022] In this embodiment, the classification of contracts for which different templates should be prepared is simply expressed as "industry," but this is just one example. It is sufficient to express the classification of contracts for which different templates should be prepared, and "industry" may also be interpreted as type of industry, type of customer, type of work, type of product or service, or type of contract. For example, if different templates are to be prepared for personal insurance and commercial insurance, "industry type" may be interpreted as type of customer or type of service.

[0023] That is, the determination conditions 1512 may be stored in advance as a template for each business type (that is, business-specific template 151). In addition, when the determination conditions 1512 are set as an industry-specific template, the contents of the explanatory items 1511 may differ for each industry. In this case, the determination conditions 1512 whose contents differ for each industry and the explanatory items 1511 whose contents differ for each industry may be associated with each other and stored in advance as the industry-specific template 151.

[0024] In one example of this embodiment, the business-specific template 151 (template) includes, as determination conditions 1512, explanatory items 1511 for the user 2 and at least one required item 1513 associated with the explanatory items 1511. The information to be presented to user 2 may include determination conditions 1512 in addition to explanation items 1511. That is, both explanation items 1511 and determination conditions 1512 may be prerequisite information to be presented to user 2, that is, information intended for user 2.

[0025] 2, the calculation unit 110 includes, as its functional units, a voice information acquisition unit 111, a recognition result acquisition unit 112, a character information acquisition unit 113, a conversation information output unit 114, an analysis result acquisition unit 115, and a presentation unit 116. The functions and operation flow of the calculation unit 110 will be described below with reference to FIG. 4 in addition to FIG. 2.

[0026] FIG. 4 is a diagram showing an example of the flow of operations of conversation determination device 10 of this embodiment. (Step S10) The sound collection unit 20 collects the conversation between the sales representative 2A (user 2) and the contractor 3. The sound collection unit 20 outputs the collected voice of the conversation as voice information 201 to the calculation unit 110. The voice information acquisition unit 111 acquires the voice information 201, which is the collected voice of the conversation of the user 2.

[0027] FIG. 5 is a diagram showing an example of a conversation in this embodiment. The figure shows an example of a conversation between a sales representative 2A, who is a user 2 of the conversation determination system 1, and his / her customer (e.g., a policyholder 3). The example shown in the figure shows a series of conversations 202 between the sales representative 2A and the policyholder 3 when concluding a new insurance contract. Here, the series of conversations 202 refers to a group of multiple conversations from the start of the explanation of the contract contents to the end (so-called closing). In this example, the conversations in the series of conversations 202 take place between time t1 and time tn.

[0028] The sound collection unit 20 collects the sounds of this series of conversations 202 in time series (that is, in order from time t1 to time tn). That is, the sound information acquisition unit 111 sequentially acquires the sounds of the series of conversations 202 in time series.

[0029] For example, the speech information acquisition unit 111 detects conversation breaks (for example, timing of periods, silent sections, etc.) from the acquired speech. The speech information acquisition unit 111 executes the processes from step S20 onwards for each detected conversation break. In the example shown in Fig. 5, the speech information acquisition unit 111 breaks the conversation at each time t from time t1 to time tn. In this case, the voice information acquisition unit 111 executes the processes from step S20 onwards at each break in the conversation, rather than waiting until the end of a series of conversations 202. That is, the conversation determination device 10 sequentially processes the voice information 201 in a real-time format or a streaming format (hereinafter also referred to as a real-time streaming format).

[0030] [Speaker identification function] Conversation determination device 10 may have a function of identifying a speaker. The following describes a case where conversation determination device 10 has a function of identifying a speaker.

[0031] The speech information 201 acquired by the speech information acquisition unit 111 includes speech from multiple speakers. In one example of this embodiment, the speech information 201 includes speech from two speakers: a sales representative 2A and a contractor 3. By including the identification result acquisition unit 112, the conversation determination device 10 can distinguish and handle the speech of a specific speaker (e.g., sales representative 2A) from the speech of another speaker (e.g., contractor 3) from among the speech information 201 that includes speech from multiple speakers.

[0032] In this case, the voice information acquiring unit 111 outputs the acquired voice information 201 (for example, information in which a series of conversations 202 are sequentially picked up in time series) to the identification result acquiring unit 112.

[0033] (Step S20) Returning to FIG. 4, the recognition result acquisition unit 112 transmits the result to the conversational voice recognition device 40.

[0034] (Step S410) Based on a known algorithm, the conversational voice recognition device 40 identifies the speaker of the conversational voice included in the voice information 201. Specifically, the conversational voice recognition device 40 identifies whether the conversational voice included in the voice information 201 is from the sales representative 2A or the contractor 3. (Step S420) Conversational speech identification device 40 transmits speaker identification result 401 to conversation determination device 10.

[0035] (Step S30) The identification result acquisition unit 112 acquires the identification result 401 (i.e., the result of identifying the speaker of the conversation). The identification result acquisition unit 112 associates the identification result 401 with the voice information 201, thereby making it possible to distinguish the speaker of the voice, as shown in FIG. The identification result acquisition unit 112 outputs the identification result 401 to the presentation unit 116 .

[0036] In the above example, the conversational speech recognition device 40 is configured as a cloud server or the like and is an external device to the conversation determination device 10, but the present invention is not limited to this. The conversational speech recognition device 40 may be configured as a part of the built-in functions of the conversation determination device 10. Furthermore, in the conversation determination device 10, the calculation unit 110 has been described as including, as its functional units, voice information acquisition unit 111, identification result acquisition unit 112, character information acquisition unit 113, conversation information output unit 114, analysis result acquisition unit 115, and presentation unit 116, but this is not limiting. For example, the calculation unit 110 may only include the functions of voice information acquisition unit 111 and presentation unit 116 (e.g., a web browser function), and each of the functions of identification result acquisition unit 112, character information acquisition unit 113, conversation information output unit 114, analysis result acquisition unit 115, etc. may be implemented on a cloud server (not shown).

[0037] Furthermore, the above-described speaker identification function is not essential for the conversation determination system 1, and the conversation determination system 1 does not necessarily have to have the speaker identification function. In the case where the conversation determination system 1 does not have the speaker identification function, the conversation determination system 1 may be configured to determine the content of the conversation without distinguishing the speakers of the audio information 201. In this case, the operations of steps S20 to S30 are omitted.

[0038] [Conversational audio transcription function] Conversation determination device 10 may have a function of transcribing conversational voices. Hereinafter, a case where conversation determination device 10 has a function of transcribing conversational voices will be described. Speech information acquisition unit 111 outputs speech information 201 acquired in step S10 described above to text information acquisition unit 113. Furthermore, if conversation determination device 10 has a function for identifying speakers, identification result acquisition unit 112 may output speech information 201 in which identification result 401 is associated with speech information 201 to text information acquisition unit 113.

[0039] (Step S40) Returning to FIG. 4, the text information acquisition unit 113 transmits the audio information 201 to the transcription device 50.

[0040] (Step S510) Based on a known algorithm, the transcription device 50 transcribes the conversation included in the audio information 201. Specifically, the transcription device 50 converts the audio of the conversation included in the audio information 201 into text in Japanese (or a predetermined language such as English) (i.e., converts it into text data). (Step S520) The transcription device 50 transmits the transcribed result to the conversation determination device 10 as text information 501.

[0041] (Step S50) Text information acquisition unit 113 acquires text information 501 (i.e., information obtained by transcribing the conversation included in audio information 201). Text information acquisition unit 113 outputs acquired text information 501 to conversation information output unit 114 and presentation unit 116.

[0042] In the above example, the transcription device 50 is configured as a cloud server or the like and is described as an external device to the conversation determination device 10, but this is not limiting. The transcription device 50 may also be configured as part of the built-in functions of the conversation determination device 10.

[0043] Furthermore, the conversation determination system 1 does not necessarily have to have the conversation transcription function, and the conversation determination system 1 does not necessarily have to have the conversation transcription function. In the case where the conversation determination system 1 does not have the conversation transcription function, the conversation determination system 1 may be configured to determine the content of the conversation without transcribing the audio information 201 (i.e., using the data format of the audio information 201 as is). In this case, the operations of steps S40 to S50 are omitted.

[0044] [Conversation analysis function] Next, the conversation analysis function provided by the conversation determination device 10 will be described.

[0045] (Step S60) The conversation information output unit 114 transmits information about a conversation based on the sound of the sound information 201 collected by the sound collection unit 20 to the conversation analysis device 60 as conversation information 101.

[0046] As described above, the conversation information 101 includes the conversation determination condition 1512 in addition to the voice information 201. The conversation information output unit 114 outputs the type of conversation to the conversation analysis device 60 as the determination condition 1512.

[0047] Here, if the conversation determination system 1 has a function of identifying speakers, the speaker identification result 401 may be included in the conversation information 101 and transmitted to the conversation analysis device 60. That is, in this case, the conversation information 101 includes the speech information 201 of the speech information 201 and the identification result 401 in which the speaker of the conversation included in the speech information 201 is identified.

[0048] Furthermore, if the conversation determination system 1 has a function of transcribing conversational voice, the text information 501 obtained by transcribing the voice information 201 may be transmitted to the conversation analysis device 60 as the conversation information 101. In this case, the conversation information 101 includes the text information 501.

[0049] That is, the conversation information output unit 114 outputs the text information 501 to the conversation analysis device 60 as the conversation information 101 .

[0050] In either case, it can be said that the conversation information 101 is generated based on the voice information 201. That is, the conversation information output unit 114 outputs the conversation information 101 based on the voice information 201 to the conversation analysis device 60.

[0051] [Conversational audio transcription function]

[0052] (Step S610) The conversation analysis device 60 determines, based on a known algorithm, whether or not the conversation information 101 includes an utterance corresponding to the required item 1513. The conversation analysis device 60 generates the result of the determination as to whether or not the utterance corresponding to the required item 1513 is included as the determination result 602. (Step S620) The conversation analysis device 60 includes the generated determination result 602 in the conversation analysis result 601 and transmits it to the conversation determination device 10.

[0053] The conversation analysis device 60 of this embodiment is configured by so-called generative artificial intelligence (AI). Generative AI is software that, when given a question using text data or the like, generates a response based on a trained model for the domain to which the question belongs. The conversation analysis device 60 of this embodiment determines whether essential items 1513 of the conversation are missing based on a trained model that has learned the correspondence between speech information 201 (or text information 501 obtained by transcribing the speech information 201) and predetermined determination conditions 1512.

[0054] As described above, the conversation determination device 10 includes the explanatory items 1511, the predetermined determination conditions 1512, and the required items 1513 in addition to the audio information 201 in the conversation information 101, and provides the conversation analysis device 60 with the conversation information 101. The explanatory items 1511, the predetermined determination conditions 1512, and the required items 1513 function as part of a prompt for the conversation analysis device 60. A prompt is a question given to the generation AI. The generation AI generates an analysis result based on the content of the prompt. In general, if the content of the prompt contains sufficient information about the domain to which the question belongs, the accuracy of the analysis result generated by the generation AI can be improved (for example, the desired result can be obtained from the perspective of user 2).

[0055] Here, the domain to which a question belongs refers to a combination of customers (or customer segments), customer needs (or customer functions), and the things (or resources) that a company can provide to customers. The conversation determination device 10 of this embodiment expresses this domain by the above-mentioned "industry type." As mentioned above, by providing the appropriate question to the generation AI (i.e., by providing the appropriate prompt), it is possible to elicit an appropriate answer from the generation AI. For example, by providing the generation AI with information about the domain to which the question belongs along with the question, the generation AI will provide an answer that is appropriate to the domain to which the question belongs.

[0056] The conversation information output unit 114 outputs to the conversation analysis device 60 the conversation information 101 including the determination condition 1512 obtained by selecting the business-specific template 151 (template) based on the business type in which the conversation takes place.

[0057] The conversation determination device 10 provides the domain information expressed by the industry-specific template 151 together with the conversation information 101 to the conversation analysis device 60 (i.e., the generation AI), thereby obtaining a conversation analysis result 601 that is in line with the domain (i.e., industry) to which the conversation information 101 belongs.

[0058] In addition, the conversation analysis device 60 may generate a conversation analysis result 601 with improved accuracy of the inference result by using the inference result of conversation analysis using an existing language-trained trained model and a sentence generation method using RAG (Retrieval-augmented Generation) using audio information 201 and judgment conditions 1512. That is, the conversation analysis device 60 may use a sentence generation method using RAG, which searches an existing conversation information database and generates answer sentences based on the conversation information obtained as a result of the search. That is, the conversation analysis device 60 may be configured to generate the conversation analysis result 601 based on a search of an existing conversation information database, a RAG-based sentence generation method that generates answer sentences based on the conversation information obtained as a result of the search, and a trained model that has learned the correspondence between the audio information 201 and the judgment condition 1512. The trained model, conversation information database, and RAG algorithm can be existing ones. According to the conversation analysis device 60 configured in this way, it is possible to improve the inference accuracy of the inference results of conversation analysis using an existing trained model that has undergone language training. In the above example, the conversation analysis result 601 is described as a reply sentence, but is not limited to this. The conversation analysis result 601 may be information indicating to what extent the content of the conversation to be judged satisfies a predetermined standard (for example, a numerical value on a three-level scale such as 0.0, 0.5, and 1.0).

[0059] The conversation determination device 10 may provide auxiliary information to the conversation analysis device 60 in addition to the determination conditions 1512 based on the industry-specific template 151. For example, the conversation determination device 10 may provide the conversation analysis device 60 with mandatory explanatory matters stipulated by laws, ordinances, etc. in the domain to which the conversation information 101 belongs (for example, the contents of important information sheets, contracts, pamphlets, etc. in insurance contracts or real estate contracts) as auxiliary information. Furthermore, for example, the conversation determination device 10 may provide exemplary conversation examples (good conversation examples) or typical failure examples (bad conversation examples) in the domain to which the conversation information 101 belongs as auxiliary information to the conversation analysis device 60. Such auxiliary information may be included as determination conditions 1512 in the industry-specific template 151.

[0060] That is, the business-specific template 151 (template) includes examples of good conversation or bad conversation for each explanation item 1511 as the judgment conditions 1512 . Incidentally, being included in the business template 151 also includes including in the business template 151 a link to a web page that contains information on good conversation examples or bad conversation examples.

[0061] (Step S70) The analysis result acquisition unit 115 acquires the conversation analysis result 601 transmitted by the conversation analysis device 60. As described above, the conversation analysis result 601 includes a determination result 602 as to whether or not an utterance corresponding to the required item 1513 is included.

[0062] That is, the analysis result acquisition unit 115 acquires a conversation analysis result 601 which is the analysis result by the conversation analysis device 60 and includes at least a determination result 602 as to whether or not a required item 1513 in the conversation has been uttered. The analysis result acquisition unit 115 outputs the acquired conversation analysis result 601 to the presentation unit 116.

[0063] As described above, the conversation analysis device 60 generates a conversation analysis result 601 with improved accuracy of the inference result by using the inference result of conversation analysis by an existing trained model that has undergone language learning and a sentence generation method by RAG that uses the speech information 201 and the judgment condition 1512. In this case, it can be said that the analysis result acquisition unit 115 acquires the conversation analysis result 601 obtained as a result of providing the speech information 201 and the judgment condition 1512 to the trained model.

[0064] In the above description, the conversation analysis device 60 is described as being equipped with a generation AI, but this is not limited to this. The conversation analysis device 60 may be configured to output, as the conversation analysis result 601, an answer selected based on a correspondence between a predefined question sentence and an answer sentence.

[0065] Furthermore, in the above description, it is assumed that the conversation analysis device 60 is a text-based generation AI (i.e., a generation AI in which the question format is text data), but this is not limited to this. The conversation analysis device 60 may also be configured as a voice-based generation AI (e.g., a generation AI that can directly interpret the voice information 201).

[0066] (Step S80) The presentation unit 116 acquires the conversation analysis result 601. The presentation unit 116 presents the proposed content 102 based on the determination result 602 included in the conversation analysis result 601 to the user 2. As an example, the presentation unit 116 displays the analysis result image P10 shown in FIG. 6 on the display unit 30 as the proposed content 102.

[0067] 6 is a diagram showing an example of an analysis result image P10 presented by the presentation unit 116 of this embodiment. The analysis result image P10 includes a business negotiation item image P1, a conversation content image P2, an explanation item image P3, a determination result image P4, and an end operation image P5. The analysis result image P10 shown in the figure shows a business negotiation item image P1, a conversation content image P2, an explanation item image P3, a determination result image P4, and a termination operation image P5 when it is determined that the business negotiation (contract procedure) cannot be completed.

[0068] The business negotiation item image P1 is an image showing information about the conversation (business negotiation) item. The business negotiation item image P1 displays the name of the customer (contractor 3), the name of the product (service), the name of the sales representative 2A, and the like. The conversation content image P2 is an image showing the content of the conversation between the sales representative 2A and the contractor 3. The explanatory item image P3 is an image that lists items that the sales representative 2A should explain to the contractor 3 in a series of conversations 202 (for example, conversations during business negotiations). The explanatory item image P3 displays, for example, the contents of the explanatory items 1511 and the determination conditions 1512 of the industry-specific template 151. Note that, depending on the settings of the industry-specific template 151, the configuration may be such that the contents of the explanatory items 1511 can be displayed without displaying the contents of the determination conditions 1512.

[0069] The explanation item image P3 may include a check image P31. The check image P31 is an image that distinguishes between explanation items 1511 that have been explained and explanation items 1511 that have not been explained in the series of conversations 202. For example, in the check image P31, an explanation item 1511 that has been explained is displayed with a check icon indicating that it has been explained. Also, in the check image P31, an explanation item 1511 that has not been explained is displayed with no check icon.

[0070] A check icon may be displayed on this check image P31 by the sales representative 2A himself / herself performing a check operation. In this case, the check image P31 functions as an image for a self-check (self-assessment) by the sales representative 2A to reduce conduct risk.

[0071] In this case, the check image P31 functions as a suggestion 102 from the conversation determination device 10 to reduce the conduct risk.

[0072] Furthermore, the check image P31 may display a check icon based on the judgment result 602 generated by the conversation analysis device 60 as a result of analyzing the utterances of the sales representative 2A. In this case, the check image P31 functions as an image for an objective judgment (third-party judgment) by the conversation analysis device 60 to reduce conduct risk.

[0073] In this case as well, the check image P31 functions as a suggestion 102 from the conversation determination device 10 to reduce the conduct risk.

[0074] As described above, the check image P31 is an image that distinguishes between explanation items 1511 that have been explained and explanation items 1511 that have not been explained in a series of conversations 202, and functions as the proposal content 102 from the conversation determination device 10 to the sales representative 2A. That is, the proposal content 102 includes, for each explanation item 1511, a determination result 602 as to whether or not the required item 1513 associated with the explanation item 1511 is included in the conversation.

[0075] The determination result image P4 is an image showing items for which explanation is insufficient based on the determination result 602 by the conversation analysis device 60. For example, the presentation unit 116 causes the display unit 30 to display an image showing that explanation is insufficient for the required items 1513 determined by the conversation analysis device 60 as not being uttered, among the required items 1513 defined in the industry-specific template 151.

[0076] If the conversation determination device 10 has a transcription function, the result of transcription by the transcription device 50 is displayed in the conversation content image P2.

[0077] That is, the presentation unit 116 presents text information 501, which is a transcription of a conversation by the user 2, in association with the determination result 602 of the conversation.

[0078] Furthermore, if conversation determination device 10 has a function for identifying speakers, conversation content image P2 displays the content of the conversation separated for each identified speaker. In this case, the presentation unit 116 presents the determination result 602 for the voice of the sales representative 2A out of the voice of the sales representative 2A included in the voice information 201 and the voice of the contractor 3. Here, the sales representative 2A is also referred to as the specific speaker to be determined.

[0079] That is, the presentation unit 116 presents the determination result 602 for the voice of the predetermined speaker to be determined from the collected conversation.

[0080] The end operation image P5 is a button image that is operated by the sales representative 2A when the series of conversations 202 has ended. The end operation image P5 includes an end impossible image P51 and an end possible image P52. The termination impossible image P51 is an image that is displayed when it is determined that the negotiation (contract procedure) cannot be terminated because any of the required items 1513 has not been explained in the series of conversations 202. The termination impossible image P51 displays, for example, "Negotiation cannot be terminated," and the sales representative 2A is placed in a state where no operation can be performed. The end possible image P52 is an image that is displayed when it is determined that all of the required items 1513 have been explained in the series of conversations 202 and that the negotiations (contract procedures) can be ended. The end possible image P52 displays, for example, "Negotiations Ended" and can be operated by the sales representative 2A.

[0081] (Step S90) Returning to Fig. 4, the conversation determination device 10 determines whether or not the series of conversations 202 has ended. For example, when the endable image P52 is operated, the conversation determination device 10 determines that the series of conversations 202 has ended.

[0082] The example in the figure shows a case where the conversation analysis device 60 has determined that one of the required items 1513 (for example, "disclaimer") has not been explained. In this case, the termination impossible image P51 is displayed, and the termination possible image P52 cannot be operated. In this case, the conversation determination device 10 determines that the series of conversations 202 has not ended, and returns the process to step S10.

[0083] In addition, the conversation determination device 10 may determine whether the series of conversations 202 has ended based on the presence or absence of conversations at the time of contract completion, such as "Thank you for signing the contract" at time tn-1 and "Thank you for your cooperation" at time tn shown in Figure 5. Alternatively, the conversation analysis device 60 may determine whether the series of conversations 202 has ended using, for example, a generation AI, and the conversation determination device 10 may receive the result of the determination from the conversation analysis device 60 .

[0084] That is, when conversation determination device 10 determines that the series of conversations 202 has ended (step S90; YES), it ends the series of processes. On the other hand, when conversation determination device 10 determines that the series of conversations 202 has not ended (step S90; NO), it returns the process to step S10.

[0085] In this way, the conversation determination device 10 repeatedly executes steps S10 to S90 while the conversation between the sales representative 2A and the contractor 3 is ongoing (i.e., before the series of conversations 202 ends). The conversation determination device 10 presents the proposal content 102 (for example, displays the analysis result image P10 in step S80) before the series of conversations 202 ends.

[0086] That is, the presenting unit 116 presents the proposed content 102 for at least a part of the series of conversations 202 before the series of conversations 202 ends.

[0087] FIG. 7 shows an example of the proposal content 102 when the sales representative 2A confirms the proposal content 102, advances the conversation further, and adds an explanation about the required item 1513 (for example, "disclaimer").

[0088] 7 is a diagram showing another example of an analysis result image presented by the presentation unit 116 of this embodiment. The diagram shows an analysis result image P101 in which an explanation is given about the required items 1513 (for example, "disclaimers") and it is determined that the business negotiation (contract procedure) can be completed.

[0089] That is, the presentation unit 116 presents the proposed content 102 based on the determination result 602 regarding the re-conversation, which is the conversation after the proposed content 102 has been presented to the user 2.

[0090] In the example shown in the same figure, a termination image P52 is displayed as the termination operation image P5. The termination image P52 is an image that is displayed when all of the required items 1513 have been explained in the series of conversations 202 and it has been determined that the negotiations (contract procedures) can be terminated. The termination image P52 displays, for example, "Negotiations Ended" and can be operated by the sales representative 2A.

[0091] Note that the conversation information output unit 114 selects an industry-specific template 151 before the start of the series of conversations 202, and does not change to another type of industry-specific template 151 between the start and end of the series of conversations 202. With this configuration, even when the speech information 201 is processed sequentially, the conversation determination device 10 does not change information such as the determination condition 1512 to be provided to the conversation analysis device 60, and therefore a stable conversation analysis result 601 can be obtained.

[0092] 8 is a diagram showing an example of a management image P11 displayed on the management device 11 of this embodiment. The management image P11 includes a result list image P6 and a risk level image P7. The result list image P6 is an image that displays, in a list format, the status of conversations (e.g., business negotiations) between the sales representatives 2A managed by the administrator 2B and the contractors 3. The risk level image P7 is an image that shows the status of conduct risk occurrence for each sales representative 2A. By presenting management image P11 to manager 2B, conversation determination device 10 makes it easier for manager 2B to identify sales representative 2A who is likely to cause conduct risk. This allows manager 2B to provide appropriate guidance to sales representative 2A who is likely to cause conduct risk in order to reduce the conduct risk, for example. In other words, conversation determination device 10 can reduce conduct risk.

[0093] As described above, the conversation determination device 10 of this embodiment has a prompt function that adds information about the domain to which the subject of the conversation belongs to the content of a conversation by the user 2 and provides the added information to the conversation analysis device 60. For example, when the user 2 (e.g., sales representative 2A) negotiates an insurance contract, the conversation determination device 10 provides information about essential items in the explanation of the insurance contract to the conversation analysis device 60. The conversation determination device 10 configured in this way can use a general-purpose conversation analysis device 60 that stores information about various domains, and extract a conversation analysis result 601 according to the domain from the conversation analysis device 60.

[0094] Furthermore, since the conversation determination device 10 of this embodiment processes the audio information 201 sequentially in a real-time streaming format, it is possible to notify the sales representative 2A of the presence or absence of a conduct risk and to make suggestions to reduce the conduct risk before the end of the series of conversations 202.

[0095] In the above description, the conversation determination device 10 sequentially processes the audio information 201 in a real-time streaming format, but this is not limiting. The conversation determination device 10 may present the proposal content 102 processed in a batch format using the audio information 201 of a series of conversations 202 stored in advance. According to the conversation determination device 10 configured in this way, for example, the manager 2B can check the conversation results of the sales representative 2A at a later date (for example, offline).

[0096] Although the embodiments of the present invention have been described in detail above with reference to the drawings, the specific configuration is not limited to this embodiment and can be appropriately modified without departing from the spirit of the present invention. The configurations described in the above-described embodiments may be combined.

[0097] Each unit included in each device in the above-described embodiments may be realized by dedicated hardware, or may be realized by a memory and a microprocessor.

[0098] In addition, each part of each device may be composed of a memory and a CPU (central processing unit), and the functions of each part of each device may be realized by loading a program into memory and executing it.

[0099] In addition, a program for realizing the functions of each unit of each device may be recorded on a computer-readable recording medium, and the program recorded on the recording medium may be read into a computer system and executed to perform processing by each unit of the control unit. Note that the term "computer system" here includes hardware such as an OS and peripheral devices.

[0100] Furthermore, if a WWW system is used, the "computer system" also includes the homepage provision environment (or display environment). "Computer-readable recording media" refers to portable media such as flexible disks, optical magnetic disks, ROMs, and CD-ROMs, as well as storage devices such as hard disks built into computer systems. Furthermore, "computer-readable recording media" also includes devices that dynamically store programs for a short period of time, such as communication lines used when transmitting programs over networks like the Internet or over communication lines like telephone lines, and devices that store programs for a fixed period of time, such as volatile memory within computer systems that serve as servers or clients. Furthermore, the programs may be programs that implement some of the aforementioned functions, or may be programs that can realize the aforementioned functions in combination with programs already stored in the computer system. [Explanation of symbols]

[0101] 1...conversation determination system, 2...user, 2A...sales representative, 2B...administrator, 3...contractor, 10...conversation determination device, 110...calculation unit, 111...voice information acquisition unit, 112...identification result acquisition unit, 113...text information acquisition unit, 114...conversation information output unit, 115...analysis result acquisition unit, 116...presentation unit, 150...storage unit, 11...management device, 20...sound collection unit, 30...display unit, 40...conversation voice recognition device, 50...transcription device, 60...conversation analysis device, 101...conversation information, 102...proposal content, 151...industry-specific template Plate, 1511...explanatory items, 1512...judgment conditions, 1513...required items, 201...audio information, 202...series of conversation, 401...recognition results, 501...text information, 601...conversation analysis results, 602...judgment results, P10...analysis result image, P11...management image, P1...business negotiation image, P2...conversation content image, P21...re-conversation image, P3...explanatory item image, P4...judgment result image, P5...termination operation image, P51...cannot be terminated image, P52...can be terminated image, P6...result list image, P7...risk level image

Claims

1. a voice information acquisition unit that acquires voice information obtained by collecting voices of conversations between users; a conversation information output unit that outputs conversation information based on the voice information to a conversation analysis device; an analysis result acquisition unit that acquires a conversation analysis result obtained by the conversation analysis device, the conversation analysis result including at least a determination result as to whether or not a required item in the conversation has been uttered; a presentation unit that presents a proposal based on the determination result to a user; A conversation determination device comprising:

2. the voice information acquisition unit sequentially acquires a series of conversation voices in time series, The presentation unit presents the proposed content for at least a part of the series of conversations before the series of conversations ends. The conversation determination device according to claim 1 .

3. the conversation information includes, in addition to the voice information, a conversation determination condition; The conversation information output unit outputs the type of conversation as the determination condition to the conversation analysis device. The conversation determination device according to claim 1 .

4. the conversation analysis device generates a conversation analysis result in which the accuracy of the inference result is improved by using an inference result of conversation analysis using an existing trained model that has been language-trained, and a sentence generation method using RAG (Retrieval-augmented Generation) that uses the speech information and the judgment conditions; The analysis result acquisition unit acquires the conversation analysis result obtained as a result of providing the speech information and the determination condition to the trained model. The conversation determination device according to claim 3 .

5. The type of conversation is the type of business in which the conversation takes place, The determination conditions are stored in advance as templates for each of the business types, The conversation information output unit outputs the conversation information, including the determination condition obtained by selecting the template based on the business type in which the conversation takes place, to the conversation analysis device. The conversation determination device according to claim 3 .

6. The conversation information output unit selects the template before a series of conversations starts, and does not change to another type of template from the start to the end of the series of conversations. The conversation determination device according to claim 5 .

7. The template includes, as the determination conditions, explanatory items for the user and at least one required item associated with the explanatory items. The conversation determination device according to claim 5 .

8. The proposed content includes, for each of the explanation items, a determination result as to whether or not the required item associated with the explanation item is included in the conversation. The conversation determination device according to claim 7 .

9. The template includes examples of good conversation or bad conversation for each of the explanation items as the judgment criteria. The conversation determination device according to claim 7 .

10. a character information acquiring unit that acquires character information obtained by converting a conversation included in the voice information into text; Furthermore, The conversation information output unit outputs the character information to the conversation analysis device as the conversation information. The conversation determination device according to claim 1 .

11. The presentation unit presents the character information, which is a transcription of a conversation between users, in association with a determination result of the conversation. The conversation determination device according to claim 10.

12. an identification result acquisition unit that acquires an identification result that identifies the speaker of the conversation; Furthermore, The presentation unit presents a result of the determination of the voice of a predetermined speaker to be determined from the collected conversation. The conversation determination device according to claim 1 .

13. The presentation unit presents a proposal based on a determination result regarding a re-conversation, which is a conversation after the proposal is presented to the user. The conversation determination device according to claim 1 .

14. A computer included in the conversation determination device Acquires audio information that captures the user's conversation, outputting conversation information including the voice information to a conversation analysis device; acquiring a conversation analysis result obtained by the conversation analysis device, the conversation analysis result including at least a determination result as to whether or not a required item in the conversation has been uttered; Presenting the user with a proposal based on the determination result; A conversation determination method comprising:

15. On the computer, Acquiring audio information obtained by capturing the audio of a user's conversation; outputting conversation information including the voice information to a conversation analysis device; acquiring a conversation analysis result obtained by the conversation analysis device, the conversation analysis result including at least a determination result as to whether or not a required item in the conversation has been uttered; presenting a proposal based on the determination result to the user; A program to execute.

Images