Voice data analysis device, voice data analysis method, and computer program

Abstract

To provide a voice data analysis apparatus configured to support conversational operating activities while suppressing increase in burden on a sales clerk.SOLUTION: A voice data analysis apparatus includes a conversation sound data acquisition unit, a conversation text generation unit, a conversation period determination unit, and a conversation content analysis unit. The conversation sound data acquisition unit acquires conversation sound data. The conversation text generation unit generates conversation text data. The conversation period determination unit determines, based on the conversation text data, a first period in a recording period of the conversation sound data. The conversation content analysis unit analyzes, using the conversation text data in the first period, the content of the conversation in the first period to generate analytical information.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to a voice data analysis device and a voice data analysis method. [Background technology]

[0002] 2. Description of the Related Art When a salesperson conducts sales activities through conversation with a customer, there are systems that support the salesperson's sales activities. For example, a sales proposal system has been proposed that accepts input of various information from salespeople during sales activities, creates a sales scenario based on the input information, and displays the sales scenario (Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6572354 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the above-mentioned business proposal system requires salespeople to input various pieces of information, which makes it time-consuming to use. In particular, in sales activities that involve conversation, it is troublesome and a heavy burden for a salesperson to perform input work while talking to a customer.

[0005] Therefore, it is desirable for one aspect of the present disclosure to provide a voice data analysis device and a voice data analysis method that can support sales activities through conversation while suppressing an increase in the burden on salespeople. [Means for solving the problem]

[0006] A voice data analysis device according to one aspect of the present disclosure includes a conversation sound data acquisition unit, a conversation text generation unit, a conversation period determination unit, and a conversation content analysis unit. The conversation sound data acquisition unit is configured to acquire conversation sound data. The conversation sound data is information in which the voice of a conversation is recorded. The conversation text generation unit is configured to generate conversation text data. The conversation text data is text data of the conversation content included in the conversation sound data.

[0007] The conversation period determination unit is configured to determine a first period of the recording period of the conversation sound data based on the conversation text data. The first period is a period of the recording period of the conversation sound data corresponding to a predetermined first conversation state. The conversation content analysis unit is configured to generate analysis information by analyzing the conversation content of the first period using the conversation text data of the first period.

[0008] This voice data analysis device can provide analysis information generated by analyzing the conversation content of the first period to a user of the voice data analysis device. Based on the analysis information, the user can determine areas for improvement in the conversation content in conversation-based sales activities and whether or not the conversation content is good. Therefore, by using the voice data analysis device, it is possible to improve the conversation skills of the person having the conversation and provide conversation content that can serve as a model for other conversations.

[0009] Therefore, the voice data analysis device can support conversational sales activities while suppressing an increase in the burden on salespeople during conversational sales activities. In the voice data analysis device of the present disclosure, the first conversation state may be a calling state in which a product salesperson calls out to people passing through a sales area. The conversation period determination unit may be configured to determine a first period within a recording period of the conversation sound data by using a machine learning model. The conversation period determination unit may determine that a period corresponding to the conversation text data determined to be the first period by the machine learning model is the first period. The machine learning model may be configured to determine whether or not the conversation text data is the first period in response to input of the conversation text data. The conversation content analysis unit may be configured to analyze at least one of the number of times the conversation is called out, the volume of the voice, whether or not a greeting is used, and whether or not a nickname is used in the conversation content of the first period.

[0010] Such a voice data analysis device can analyze the content of conversations that occur when a salesperson speaks to people passing by a sales area to sell products to them. The machine learning model may be trained in advance based on the teacher data. The teacher data may include first corresponding learning data, which is conversation text data corresponding to the first period, and first non-corresponding learning data, which is conversation text data not corresponding to the first period. Such a machine learning model is trained to determine that the first period is the first period in response to input of the first corresponding learning data, and machine learning is performed to determine that the first period is not the first period in response to input of the first non-corresponding learning data.

[0011] Next, in the voice data analysis device of the present disclosure, the first conversation state may be a chat state during a prearranged business negotiation. The chat state is a state in which casual conversation occurs during the business negotiation. Casual conversation is a conversation about a topic different from the main topic of the business negotiation.

[0012] The conversation period determination unit may be configured to determine a first period from a recording period of the conversation sound data by using a machine learning model. The conversation period determination unit may determine a period corresponding to conversation text data determined to be the first period by the machine learning model to be the first period. The machine learning model may be configured to determine whether or not the conversation text data is the first period in response to input of the conversation text data. The conversation content analysis unit may be configured to analyze at least one of the presence or absence of a topic related to the main topic and the number of times a topic related to the main topic is mentioned in the conversation content of the first period.

[0013] Such a voice data analysis device can analyze the content of casual conversation during a prearranged business meeting. The "pre-reserved business meeting" may be, for example, a business meeting in a conference room or a conference booth. The chat may include "conversation before the business meeting (ice breaker), (regular) chat, hearing, and conversation after the business meeting (elevator talk)."

[0014] Next, in the voice data analysis device of the present disclosure, the first conversation state may be a simple negotiation state in a conversation between a supplier and a customer representative. The simple negotiation state may be a conversation state regarding a product or service between a supplier and a customer representative during route sales or delivery.

[0015] The conversation period determination unit may be configured to determine a first period from a recording period of the conversation sound data by using a machine learning model. The conversation period determination unit may determine a period corresponding to conversation text data determined to be the first period by the machine learning model to be the first period. The machine learning model may be configured to determine whether or not the conversation text data is the first period in response to input of the conversation text data. The conversation content analysis unit may be configured to analyze at least one of the number of times a topic related to a product or service is discussed and the content of the topic related to the product or service in the conversation content of the first period.

[0016] Such a voice data analysis device can analyze the content of a simple business negotiation situation between a supplier and a person in charge at a business partner. An example of the "simple negotiation state" is a standing negotiation in which a deliverer and a person in charge at a business partner talk while standing when the deliverer visits the business partner.

[0017] Next, in the voice data analysis device of the present disclosure, the conversation period determination unit may be configured to determine a second period of the recording period of the conversation sound data based on the conversation text data. The second period may be a period of the recording period of the conversation sound data corresponding to a second conversation state different from the first conversation state. The conversation content analysis unit may be configured to generate analysis information by analyzing the conversation content of the second period using the conversation text data in the second period.

[0018] Such a voice data analysis device can provide a user of the voice data analysis device with analysis information generated by analyzing the conversation content not only in the first period but also in the second period. The user can determine areas for improvement in the conversation content and the presence or absence of good conversation content based on the analysis information in each of the first period and the second period.

[0019] Therefore, the voice data analysis device can improve the conversation skills of conversation participants in a wider range of situations and provide conversation content that can serve as a model for other conversation participants. Next, in the voice data analysis device of the present disclosure, the second conversation state may be a standing negotiation state in which a product salesperson is talking to a person standing in a sales area. The conversation period determination unit may be configured to determine the second period in the recording period of the conversation sound data by using a machine learning model. The conversation period determination unit may determine that a period corresponding to the conversation text data determined to be the second period by the machine learning model is the second period. The machine learning model may be configured to determine whether or not the conversation text data is the second period in response to input of the conversation text data. The conversation content analysis unit may be configured to analyze at least one of the following in the conversation content of the second period: voice volume, the number of standing negotiations, whether or not a topic is provided, and whether or not a standing negotiation action is performed.

[0020] Such a voice data analysis device can analyze the content of a conversation during a business negotiation while standing. The machine learning model may be trained in advance based on the teacher data. The teacher data may include second corresponding learning data, which is conversation text data corresponding to the second period, and second non-corresponding learning data, which is conversation text data not corresponding to the second period. Such a machine learning model is trained to determine that the second period is present in response to input of the second corresponding learning data, and machine learning is performed to determine that the second period is not present in response to input of the second non-corresponding learning data.

[0021] Next, in the voice data analysis device of the present disclosure, the second conversation state may be a main topic conversation state in which a conversation is about the main topic of a previously scheduled business meeting. The conversation period determination unit may be configured to determine the second period within the recording period of the conversation sound data by using a machine learning model. The conversation period determination unit may determine that a period corresponding to the conversation text data determined to be the second period by the machine learning model is the second period. The machine learning model may be configured to determine whether or not the period is the second period in response to input of the conversation text data. The conversation content analysis unit may be configured to analyze at least one of the presence or absence of a proposal related to the main topic and the number of proposals related to the main topic in the conversation content of the second period.

[0022] According to such a voice data analysis device, it is possible to analyze the contents of a conversation in a main topic conversation state. Next, in the voice data analysis device of the present disclosure, the conversation period determination unit may be configured to determine a third period of the recording period of the conversation sound data based on the conversation text data. The third period may be a period of the recording period of the conversation sound data that corresponds to a third conversation state different from the first conversation state and the second conversation state. The conversation content analysis unit may be configured to generate analysis information by analyzing the conversation content of the third period using the conversation text data in the third period.

[0023] Such a voice data analysis device can provide a user of the voice data analysis device with analysis information generated by analyzing the conversation content not only in the first and second periods but also in the third period. The user can determine areas for improvement in the conversation content and whether or not the conversation content is good based on the analysis information in each of the first, second, and third periods.

[0024] Therefore, the voice data analysis device can improve the conversation skills of conversation participants in a wider range of situations and provide conversation content that can serve as a model for other conversation participants. Next, in the voice data analysis device of the present disclosure, the third conversation state may be a seated negotiation state in which a product salesperson is conversing with a person seated in a seat in the sales area. The conversation period determination unit may be configured to determine the third period in the recording period of the conversation sound data by using a machine learning model. The conversation period determination unit may determine that a period corresponding to the conversation text data determined to be the third period by the machine learning model is the third period. The machine learning model may be configured to determine whether or not the period is the third period in response to input of the conversation text data. The conversation content analysis unit may be configured to analyze at least one of the number of seated negotiations and the volume of voice in the conversation content of the third period.

[0025] Such a voice data analysis device can analyze the number of times a salesperson sits down to negotiate. The machine learning model may be trained in advance based on the teacher data. The teacher data may include third corresponding learning data, which is conversation text data corresponding to the third period, and third non-corresponding learning data, which is conversation text data not corresponding to the third period. Such a machine learning model is trained to determine that the third period is present in response to input of the third corresponding learning data, and machine learning is performed to determine that the third period is not present in response to input of the third non-corresponding learning data.

[0026] Next, in the voice data analysis device of the present disclosure, the conversation sound data may include a time stamp indicating a correspondence relationship between the conversation sound and a recording time. The conversation period determination unit may be configured to determine that the time of the time stamp corresponding to the start time of a third period of the recording period of the conversation sound data is the start time of the third period based on not only the conversation text data but also start time information indicating the start time of the third period.

[0027] This voice data analysis device determines the start time of a third period of the recording period of the conversation sound data based on not only the conversation text data but also the start time information. This allows the voice data analysis device to improve the accuracy of determining the third period of the recording period of the conversation sound data. For example, the start time information may be set based on the time when the terminal device is operated by the user. The operation of the terminal device by the user may be an operation for recording a transition to a seated business negotiation state, or an operation for recording a transition from the seated business negotiation state to another state.

[0028] Next, the voice data analysis device of the present disclosure may include an analysis information display unit. The analysis information display unit may be configured to display at least one of the pieces of analysis information obtained by the conversation content analysis unit.

[0029] Such a voice data analysis device can provide the analysis information to a user of the voice data analysis device by displaying the analysis information on the analysis information display unit. Next, in the voice data analysis device of the present disclosure, the analysis information display unit may be configured to display the analysis information as time-series data.

[0030] Such a voice data analysis device can provide a user of the voice data analysis device with changes in the sales behavior of a product salesperson over time, based on time-series data of the analysis information. In the voice data analysis device of the present disclosure, the conversation content analysis unit may be configured to analyze the conversation content of each of the multiple product salespeople. The analysis information display unit may be configured to display the analysis information of each of the multiple product salespeople.

[0031] Such a voice data analysis device can provide a user of the voice data analysis device with analysis information on a plurality of product salespeople in a form that allows comparison with each other. Next, in the voice data analysis device of the present disclosure, the conversation sound data acquisition unit may include a voice data acquisition unit and a conversation sound data extraction unit. The voice data acquisition unit may be configured to acquire conversation sound data or voice data. The voice data may include conversation sound data and background sound data. The conversation sound data extraction unit may be configured to extract the conversation sound data from the voice data.

[0032] Thereby, when the conversation sound data acquisition unit acquires voice data, it can acquire the conversation sound data by extracting the conversation sound data from the voice data. The background sound data is information that records sounds that are different from conversation and occur in the background of a speaker.

[0033] A voice data analysis method according to another aspect of the present disclosure may include acquiring conversation voice data in which a voice of a conversation is recorded, generating conversation text data, determining a first period of a recording period of the conversation voice data based on the conversation text data, and analyzing the conversation content of the first period to generate analysis information. The conversation voice data is information in which the voice of a conversation is recorded. The conversation text data is text data of the conversation content included in the conversation voice data. The first period is a period of the recording period of the conversation voice data that corresponds to a predetermined first conversation state. When analyzing the conversation content of the first period to generate analysis information, the conversation text data of the first period is used.

[0034] This voice data analysis method can provide analysis information generated by analyzing the conversation content of the first period to a user of the voice data analysis device. Based on the analysis information, the user can determine areas for improvement in the conversation content in sales activities through conversation, whether the conversation content is good, etc. Therefore, by using the voice data analysis method, it is possible to improve the conversation skills of the person having the conversation, provide conversation content that can serve as a model for other conversation participants, etc.

[0035] Therefore, the voice data analysis method can support conversational sales activities while suppressing an increase in the burden on salespeople during conversational sales activities. In the above-mentioned audio data analysis method, obtaining conversation sound data may comprise obtaining conversation sound data or audio data, and extracting the conversation sound data from the audio data. The audio data includes conversation sound data and background sound data.

[0036] As a result, in the voice data analysis method, when voice data is acquired, conversation sound data can be acquired by extracting the conversation sound data from the voice data. The background sound data is information that records sounds that are different from conversation and occur in the background of a speaker.

[0037] In another aspect of the present disclosure, a computer program may be a computer program for causing a computer to function as at least one of the conversation sound data acquisition unit, the conversation text generation unit, the conversation period determination unit, and the conversation content analysis unit in the above-mentioned voice data analysis device.

[0038] The above voice data analysis method may be executed by a computer. A voice data analysis method to which the same technical idea as the above voice data analysis device is applied may be provided. A computer program having instructions for causing a computer to execute these voice data analysis methods may be provided. The computer program may be recorded on a non-transitory computer-readable recording medium. [Brief description of the drawings]

[0039] [Figure 1] FIG. 1 is a diagram showing a schematic configuration of a voice data analysis device. [Diagram 2] 13 is a flowchart showing the processing contents of a voice data analysis process. [Diagram 3]1 is an explanatory diagram in which the upper area typically shows speech periods and noise periods in a speech data waveform, and the lower area typically shows text data generated from the speech data waveform in the speech period. [Figure 4] 13 is a flowchart showing the processing contents of an analysis process. [Diagram 5] FIG. 2 is an explanatory diagram showing an example of conversation text data TD1 to which a time stamp is added. [Figure 6] An explanatory diagram showing an example of matching data in which seating start time, seating end time, and service genre are recorded. [Figure 7] FIG. 11 is an explanatory diagram showing an example of a period determination result obtained by the analysis process. [Figure 8] FIG. 11 is an explanatory diagram showing an example of analysis information generated by a conversation content analysis. [Figure 9] FIG. 13 is an explanatory diagram showing an example of a display screen showing performance data for each product salesperson. [Figure 10] FIG. 13 is an explanatory diagram showing an example of a display screen showing detailed performance data on the "number of calls" made by one product salesperson. [Figure 11] FIG. 13 is an explanatory diagram showing an example of a display screen showing detailed performance data of "voice volume" for one product salesperson. [Figure 12] FIG. 13 is an explanatory diagram showing an example of a display screen showing detailed performance data of "greetings" for one product salesperson. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0040] Exemplary embodiments of the present disclosure will now be described with reference to the drawings. It goes without saying that the present disclosure is in no way limited to the following embodiments, and various forms can be adopted as long as they fall within the technical scope of the present disclosure.

[0041] [1. First embodiment] [1-1. Overall composition] The voice data analysis device 1 of the first embodiment is a system for analyzing voices uttered by product salespeople (hereinafter also referred to as users). The voice data analysis device 1 is configured to analyze voices uttered by product salespeople during sales activities and support sales activities through conversation. An example of sales activities is when a product salesperson calls out to people passing by a mobile phone sales area in a shopping mall and explains the product to them in order to sell them mobile phones.

[0042] As shown in FIG. 1, the voice data analysis device 1 includes a recording device 11, a device attachment / detachment unit 13, a server device 21, a display device 41, and a mobile terminal device 43. Note that the voice data analysis device 1 is provided with a plurality of recording devices 11 corresponding to the number of product sales staff, but only one recording device 11 is shown in Fig. 1. The number of recording devices 11 corresponds to the number of product sales staff. Similarly, the voice data analysis device 1 is provided with a plurality of mobile terminal devices 43 corresponding to the number of product sales staff, but only one mobile terminal device 43 is shown in Fig. 1. The number of mobile terminal devices 43 corresponds to the number of product sales staff.

[0043] [1-2. Recording equipment] Recording device 11 is a portable device provided independently of device attachment / detachment unit 13 and server device 21. Recording device 11 includes microphone 11a, operation unit 11b, control unit 11c, memory unit 11d, charging terminal 11e, first communication terminal 11f, clip unit 11g, and housing 11h. Microphone 11a, operation unit 11b, charging terminal 11e, first communication terminal 11f, and clip unit 11g are provided on the outer surface of housing 11h. Control unit 11c and memory unit 11d are provided inside housing 11h.

[0044] The recording device 11 is configured so that it can be worn by a salesperson in a state in which the voice of the salesperson can be recorded. For example, the recording device 11 is configured so that it can be worn around the collar or breast pocket of the salesperson by using the clip part 11g. In detail, the recording device 11 can be worn by pinching the collar or breast pocket between the clip part 11g and the housing 11h.

[0045] The microphone 11a converts the user's voice into an electrical audio signal and outputs it to the control unit 11c. The operation unit 11b has one or more mechanical switches that can receive recording instructions and stop instructions from the user. The control unit 11c is configured to comprehensively control each unit of the recording device 11.

[0046] When a recording instruction is input from the user through the operation unit 11b, the control unit 11c converts the user's voice signal input from the microphone 11a from analog voice data to digital voice data and records it in the storage unit 11d until a stop instruction is input. The control unit 11c performs A / D conversion (analog-to-digital conversion) of the voice data and stores the converted voice data DA1 (digital data) in the storage unit 11d. The voice data DA1 may be, for example, digital data in binary format (bin format).

[0047] The storage unit 11d is an electrically rewritable semiconductor memory, such as a flash memory. Hereinafter, the audio data recorded in the storage unit 11d is also referred to as recorded data.

[0048] The charging terminal 11e is configured so as to be directly connectable to the device attachment / detachment unit 13 (more specifically, the power supply terminal 13a described later). The charging terminal 11e is configured so as to receive charging power from the connected device attachment / detachment unit 13.

[0049] The first communication terminal 11f is configured to be directly connected to the device attachment / detachment unit 13 (more specifically, the second communication terminal 13b described later) so as to be able to perform wired communication. The first communication terminal 11f is configured to configure a wired communication path together with the device attachment / detachment unit 13 to which it is connected.

[0050] The control unit 11c is configured to transmit and receive various information to and from the server device 21 by wired communication via the first communication terminal 11f and the device attachment / detachment unit 13. The control unit 11c is configured to transmit the recording data stored in the storage unit 11d to the server device 21 in response to a request from the server device 21.

[0051] [1-3. Equipment attachment / detachment unit and server device] The device attachment / detachment unit 13 is connected to the server device 21 via a wired cable 14. The device attachment / detachment unit 13 is configured so that the recording device 11 can be detachably attached thereto. The device attachment / detachment unit 13 includes a power supply terminal 13a and a second communication terminal 13b.

[0052] The power supply terminal 13a is configured to receive charging power supplied from the server device 21 via the wired cable 14. The power supply terminal 13a is configured to output the charging power supplied from the server device 21 to the recording device 11 by being directly connected to the charging terminal 11e of the recording device 11. The power supply terminal 13a, together with the connected charging terminal 11e, constitutes a part of a charging path for charging the recording device 11 by the server device 21. This charging path is a path for sending charging power from the server device 21 to the recording device 11.

[0053] The second communication terminal 13b is configured to be directly connected to the first communication terminal 11f of the recording device 11 for wired communication. The second communication terminal 13b is configured to be electrically connected to the server device 21 via a wired cable 14 for wired communication. The second communication terminal 13b, together with the connected first communication terminal 11f, constitutes a part of the wired communication path between the recording device 11 and the server device 21. This wired communication path is configured to transmit various information transmitted and received between the recording device 11 and the server device 21.

[0054] The server device 21 includes a control unit 23, a storage unit 25, and a communication unit 27. In detail, the server device 21 is configured by one or more computers. The control unit 23 includes a CPU 23a. The CPU 23a realizes various functions by executing processes according to computer programs stored in the storage unit 25. In the following, the processes executed by the CPU 23a may be described as processes executed by the control unit 23 or the server device 21.

[0055] The storage unit 25 stores computer programs executed by the CPU 23a and various data. The storage unit 25 includes one or more hard disk drives (HDD) and / or solid state drives (SSD). The communication unit 27 is configured to be able to communicate with the recording device 11 via the wired cable 14 and the device attachment / detachment unit 13. The communication unit 27 is configured to be able to communicate with the mobile terminal device 43 via wireless communication. Examples of wireless communication include mobile data communication, wireless LAN (WiFi (registered trademark), etc.), and short-range wireless communication (Bluetooth (registered trademark) communication, infrared communication, etc.).

[0056] Control unit 23 communicates with recording device 11 via communication unit 27, receives audio data from recording device 11, and executes various processes based on the audio data. Alternatively, control unit 23 communicates with mobile terminal device 43 via communication unit 27, receives audio data from mobile terminal device 43, and executes various processes based on the audio data.

[0057] The voice data includes at least conversational voice data and background voice data. The conversational voice data includes voice data uttered by a product salesperson to a customer during sales activities. The background voice data includes voice data other than the voice data of the product salesperson during sales activities (such as background music and surrounding noise).

[0058] For example, the control unit 23 executes a process of extracting conversation sound data from the received voice data. The control unit 23 also executes a process of generating text data of the conversation contents included in the conversation sound data (hereinafter also referred to as conversation text data). The control unit 23 also executes a process of determining "calling periods, standing negotiation periods, sitting periods, other periods" in the conversation sound data based on the conversation text data. The control unit 23 also executes a process of analyzing the conversation contents of each of the "calling periods, standing negotiation periods, sitting periods, other periods" by using the conversation text data for each of the "calling periods, standing negotiation periods, sitting periods, other periods" to generate analysis information.

[0059] The control unit 23 executes each process for each of the recording devices 11, thereby generating analysis information on the content of the conversation for each of the salespeople. [1-4. Display device and mobile terminal device] The display device 41 is connected to the server device 21 via a wired cable or wireless communication. The display device 41 is configured to display various information based on a display command from the server device 21. The display device 41 may be configured using, for example, a liquid crystal display, an organic EL display, or the like.

[0060] The mobile terminal device 43 is an electronic device terminal that can be carried by a product salesperson. The mobile terminal device 43 is configured to be able to communicate with the server device 21 (more specifically, the communication unit 27) via wireless communication. The mobile terminal device 43 may be configured using, for example, a smartphone, a tablet terminal, or the like.

[0061] [1-5. Audio data analysis processing] Next, the voice data analysis process executed by the server device 21 will be described. In response to the recording device 11 being attached to the device attachment / detachment section 13, the server device 21 executes the voice data analysis process shown in FIG.

[0062] Before being attached to the device attachment / detachment section 13, the recording device 11 is attached to the vicinity of the salesperson's collar, breast pocket, etc., and records audio data including the voice uttered by the salesperson during sales activities. As described above, in the recording device 11, the control section 11c performs A / D conversion (analog-to-digital conversion) of the audio data, and stores the converted audio data DA1 (digital data) in the storage section 11d.

[0063] The audio data DA1 includes conversational audio data DA2 and background audio data DA3, as shown typically in the upper region of Fig. 3. The conversational audio data DA2 includes audio data including the voice uttered by a product salesperson during sales activities. In Fig. 3, the conversational audio data DA2 is data for sections indicated as Conversation A, Conversation B, and Conversation C. The background audio data DA3 includes audio data that is not the audio data of a product salesperson during sales activities. In Fig. 3, the background audio data DA3 is data for a section indicated as Noise.

[0064] When the voice data analysis process starts, the server device 21 (more specifically, the control unit 23) acquires voice data DA1 from the recording device 11 and stores (saves) the voice data DA1 in the storage unit 25 (S110). The control unit 23 converts the acquired voice data DA1 into a predetermined data format (for example, WAV format, etc.) and stores it in the storage unit 25. The voice data DA1 acquired at this time may be the voice data DA1 recorded by the recording device 11 on the same day, or may be all of the voice data DA1 stored in the recording device 11 that has not yet been acquired by the server device 21.

[0065] The control unit 23 performs a process of extracting the conversation sound data DA2 from the voice data DA1 (S120). At this time, the control unit 23 extracts the conversation sound data DA2 using a known VAD (Voice Activity Detection) technique.

[0066] The control unit 23 performs a process of generating conversation text data TD1 corresponding to the conversation included in the conversation sound data DA2 (S130). At this time, the control unit 23 generates the conversation text data TD1 based on the conversation sound data DA2 using a known STT (Speech To Text) technique. The conversation text data TD1 is generated as text data corresponding to the conversations A, B, and C of the voice data DA1, as shown typically in the lower region of FIG.

[0067] The control unit 23 performs a process of adding a time stamp (time information) to the conversation text data TD1 (S140). For example, for each of the conversations A, B, and C in the conversation text data TD1 shown in Fig. 3, the control unit 23 adds the start time of the corresponding section in the conversation sound data DA2 as a time stamp.

[0068] 5 shows an example of conversation text data TD1 with a timestamp. It shows the date and time (timestamp) and the main text (text content) of each of the conversation text data TD1 from number 1 to number 50. In this example, the conversation text data TD1 is based on the speech of a product salesperson recorded by the recording device 11, and shows the conversation text data TD1 from "2023-06-08 10:43:44" to "2023-06-08_10:57:34" with some parts omitted.

[0069] The control unit 23 performs an analysis process of the conversation sound data DA2 based on the conversation text data TD1 (S150). As shown in Fig. 4, in the analysis process, the control unit 23 performs a period determination process using the conversation text data TD1 and a machine learning model (S210). The period determination process is a process of determining whether a determination target period in the conversation sound data DA2 corresponds to a "calling period, standing negotiation period, sitting period, other period." In other words, the period determination process is a process of determining a period in the recording period of the conversation sound data DA2 that corresponds to each of the "calling period, standing negotiation period, sitting period, other period."

[0070] The "calling period, standing negotiation period, sitting period, and other periods" are periods obtained by subdividing the sales activity period according to the customer service state. The calling period is a period corresponding to a calling state in which a salesperson calls out to people passing by. The standing negotiation period is a period corresponding to a standing negotiation state in which a person who was about to pass by stops in response to a call or other such action, and the salesperson has a conversation with the person who stopped. The sitting period is a period corresponding to a seated negotiation state in which a person who has stopped sits down in a chair in the negotiation area, and the salesperson has a conversation with the seated person.

[0071] Other periods are periods that do not fall into any of the "calling periods, standing negotiation periods, or sitting periods." Other periods include, for example, periods related to business-related conversations between product salespeople, and periods related to special "standing negotiations" that start not with a "calling" but with a "standing negotiation." Sales activities usually start with a "calling," but there are rare cases where they start with a "standing negotiation." Such irregular "standing negotiations" are classified as "other periods."

[0072] The machine learning model used in S210 is a model configured to output the period of a business negotiation corresponding to the conversation text data TD1 in response to the input of the conversation text data TD1. In other words, this machine learning model is a model for determining the period of a business negotiation corresponding to the input conversation text data TD1.

[0073] The machine learning model used in S210 is a model obtained by so-called "supervised learning." As supervising data for learning (in other words, training data), supervising data is used in which a given conversation sentence (text data) is associated with the period of the negotiation in which the conversation sentence was spoken (a calling period, a standing negotiation period, a seated period, or any other period).

[0074] The training data may be prepared, for example, for 100 days, for 50 salespeople, with 8 hours of speech recorded per day. In the learning process using this training data, a machine learning model may be constructed taking into account the characteristics of the speech sentences, the context, and the consistency of the transition order of each period. The machine learning model does not need to be limited to a form in which a model is constructed once and used as is indefinitely. The machine learning model may be configured to learn using new training data containing newly recorded conversation sentences (text data) after a predetermined period of time has passed, and the learning results may be updated. The machine learning model may be updated at any interval, such as once a month, every few months, or once a year.

[0075] When an unnatural period determination is found in the period determination result in S210, the control unit 23 performs a first correction process to correct the unnatural period determination (S220). For example, as described above, sales activities are usually started with a "call", but in rare cases, they may start with a "standing negotiation" without passing through a "call". Even for such irregular "standing negotiation", in the case of conversation text data TD1 including a phrase used in a "standing negotiation", the machine learning model determines the "standing negotiation period", making it difficult to distinguish it from a normal "standing negotiation period". Therefore, instead of determining based only on the conversation text data TD1, the determination result is corrected by also paying attention to the transition of the period of the sales activity. For example, for a period determined to be a "standing negotiation period" from the conversation text data TD1, in the case of a sales activity that started from a "standing negotiation period" without passing through a "call period", the period determination result is corrected by correcting the classification of the "standing negotiation period" to "other periods".

[0076] The control unit 23 performs a second correction process for correcting the period determination result in S210 using the comparison data (S230). As the comparison data, for example, a timestamp (time information) of the customer service record recorded in the mobile terminal device 43 used by the product salesperson may be used. The mobile terminal device 43 is equipped with a dedicated app for recording sales activities, and the product salesperson inputs the customer service record using the dedicated app after serving a customer. When inputting this customer service record, a timestamp including the seating start time and the seating end time is automatically recorded in the storage area of ​​the dedicated app of the mobile terminal device 43.

[0077] An example of the matching data is shown in Fig. 6. This matching data is data for the service category "sitting" with a sitting start time of "2023-06-08 10:45:45" and a sitting end time of "2023-06-08 10:55:15".

[0078] Control unit 23 acquires a timestamp of the customer service record from a dedicated application of mobile terminal device 43 at a predetermined time (for example, when executing S230 or S110). Control unit 23 uses the customer service time period during which the customer was seated, identified based on the timestamp of the customer service record, as matching data, and determines whether the period determination result in S210 corresponding to the seated time period of the matching data is a "seating period". If the period determination result in S210 is a "seating period", control unit 23 maintains the determination result, and if the period determination result in S210 is different from a "seating period", corrects the period determination result by correcting the period determination result to a "seating period".

[0079] FIG. 7 shows an example of a period determination result obtained in the analysis process (more specifically, the processes of S210 to S230). In Fig. 7, the results of period determination for each of the conversation text data TD1 numbers 1 to 50 are entered in the "Phase" column of the "Period Data" column. In addition, the start and end points of a series of sales activities for one customer (or one group of customers) are entered in the "Start / End" column of the "Period Data" column. Specifically, "Start" is entered in the conversation text data TD1 number 2, which corresponds to the start of the series of sales activities, and "End" is entered in the conversation text data TD1 number 49, which corresponds to the end.

[0080] Furthermore, FIG. 7 includes the following items: "Salesperson input," "Standing negotiation count," and "Sitting count." The "salesperson input" is information based on the above-mentioned verification data. The conversation text data TD1 for number 6 contains "start sitting down," and the conversation text data TD1 for number 49 contains "end sitting down."

[0081] The "standing negotiation count" is bit data for counting the number of standing negotiations performed by a product salesperson. "1" is set to the first conversation text data TD1 (conversation text data TD1 numbered 3 in FIG. 7) among a series of conversation text data TD1 determined to be "standing negotiations" (conversation text data TD1 numbered 3 to 5 in FIG. 7).

[0082] Fig. 7 shows a part of the conversation text data TD1 (and the period determination results) on the analysis target day, and there is conversation text data TD1 (and the period determination results) that is not shown in Fig. 7. The total number of times "1" is set in the "standing negotiation count" on the analysis target day represents the standing negotiation performance of the product salesperson on that day.

[0083] "Sitting Count" is bit data for counting the number of times a salesperson sat down during negotiations. Among a series of conversation text data determined to be "sitting down" (conversation text data numbered 6 to 49 in FIG. 7), the first conversation text data (conversation text data numbered 6 in FIG. 7) is set to "1." As with "Standing Negotiation Count," the total number of times "1" is set in "Sitting Count" on the day being analyzed represents the number of seated negotiations performed by the salesperson on that day.

[0084] The control unit 23 performs a content analysis process based on the conversation sound data DA2 and the conversation text data TD1 (S240). First, as a content analysis process, the control unit 23 performs a conversation content analysis regarding "voice volume, etc." based on the conversation sound data DA2, and also performs a conversation content analysis regarding "category, greetings, nicknames, empathy, etc." based on the conversation text data TD1.

[0085] FIG. 8 shows an example of the analysis information generated by the conversation content analysis. FIG. 8 shows examples of analysis information generated by conversation content analysis of the conversation sound data DA2 from number 1 to number 50 and the conversation text data TD1, including examples of the items "category, greeting, nickname, voice volume, and empathy."

[0086] "Category" is analysis information that indicates whether or not the conversation content contains terms related to categories such as "fees, terminal, radio waves, data capacity," and if so, the category name (in FIG. 8, "fees") is entered in the conversation text data TD1 (in FIG. 8, conversation text data TD1 numbered 3).

[0087] "Greetings" is analysis information that indicates whether greeting terms such as "hello" are included in the conversation content, and if they are included, "1" is set in the conversation text data TD1 (in Figure 8, conversation text data TD1 number 2).

[0088] "Nickname" is analysis information that indicates whether or not nicknames such as "Dad" and "Mom" are included in the conversation content. If they are included, "1" is set in the conversation text data TD1 (in FIG. 8, the conversation text data TD1 numbered 2).

[0089] "Voice volume" is analysis information that expresses the volume of the salesperson's voice as a numerical value, and a numerical value is recorded for each conversation text data TD1. "Sympathy" is analysis information that indicates whether or not a term that evokes sympathy from customers is included in the conversation content, and if it is included, "1" is set in the conversation text data TD1 (in FIG. 8, conversation text data TD1 with numbers 2 and 4). Examples of terms that evoke sympathy from customers include "please take it" and "are you interested?".

[0090] Furthermore, as a continuation of the content analysis process, the control unit 23 creates performance data for each product salesperson in which the analysis information generated by the above-mentioned conversation content analysis is associated with each period (calling out, standing while negotiating, sitting down).

[0091] An example of a display screen showing the performance data for each product salesperson is shown in Fig. 9. This display screen is displayed on the display device 41 or the mobile terminal device 43, for example. Figure 9 shows the performance data for three salespeople (clerks C, D, and J). Each item is marked as follows: "that number" if the result is a number, "◎" for extremely good results, "○" for good results, "R" for sufficient results, "△" for insufficient results, and (blank) if analysis is not possible. Note that if the analysis period (standing negotiations, sitting, etc.) has not been reached, "-" or "0" is marked for each item for that period (in Figure 9, each item is for the standing negotiation period for clerk C).

[0092] Furthermore, as a continuation of the content analysis process (S240), the control unit 23 creates detailed performance data for each item for one product salesperson. An example of a display screen showing detailed performance data on the "number of calls" made by one salesperson is shown in Fig. 10. This display screen is displayed on the display device 41 or the mobile terminal device 43, for example.

[0093] Figure 10 includes the actual data on the "number of calls" obtained in this analysis, including the total number of calls per day, the average value for each hour, and actual data shown in a bar graph for each hour. Figure 10 also lists (side by side) the actual data obtained in the previous analysis in addition to the actual data obtained in this analysis.

[0094] The performance data, which shows the number of calls per hour in a bar graph, is time-series data that shows the change in the number of calls per hour. Such time-series data shows the change in the sales behavior of the product salesperson over time.

[0095] 11 shows an example of a display screen showing detailed performance data of "voice volume" for one product salesperson. This display screen is displayed on the display device 41 or the mobile terminal device 43, for example.

[0096] Figure 11 includes the actual data for "voice volume" obtained in this analysis, including the daily average value, the maximum value, and the actual data shown in a bar graph for each hour. Figure 11 also lists the actual data obtained in the previous analysis in addition to the actual data obtained in this analysis.

[0097] The performance data, which shows the voice volume for each hour in a bar graph, is time-series data showing the change in voice volume for each hour. Such time-series data shows the change in the sales behavior of a product salesperson over time.

[0098] 12 shows an example of a display screen showing detailed performance data of "greetings" for one product salesperson. This display screen is displayed on the display device 41 or the mobile terminal device 43, for example.

[0099] Figure 12 includes the performance data on "greetings" obtained in this analysis, including the greeting implementation rate and performance data shown in a bar graph for each hour. The greeting implementation rate is the ratio of the number of times greetings were implemented to the number of times people were called out. Specifically, it is the value obtained by dividing the number of times greetings were implemented by the number of times people were called out. In addition to the performance data obtained in this analysis, Figure 12 also lists the performance data obtained in the previous analysis.

[0100] The performance data, which shows "greetings" every hour as a bar graph, is time-series data that shows the change in "greetings" every hour. Such time-series data shows the change in the sales behavior of the product salesperson over time.

[0101] [1-6. Effects] According to the embodiment described above in detail, the following effects can be obtained. (1a) In the voice data analysis device 1, when executing S210, the control unit 23 determines, based on the conversation text data TD1, which of the recording periods of the conversation sound data DA2 corresponds to, among "a calling period, a standing negotiation period, a seated period, or other period."

[0102] The voice data analysis device 1 can provide the analysis information to a user of the voice data analysis device 1 by displaying on the display device 41 (or the mobile terminal device 43) analysis information generated by analyzing the conversation content of each of the "calling period, standing negotiation period, sitting period, and other periods." Based on the analysis information, the user can determine areas for improvement in the conversation content in sales activities through conversation, whether the conversation content was good, and so on. Therefore, by using the voice data analysis device 1, it is possible to improve the conversation skills of the conversation participant and provide conversation content that can serve as a model for other conversation participants.

[0103] Therefore, the voice data analysis device 1 can support sales activities through conversation while suppressing an increase in the burden on salespeople in sales activities through conversation. (1b) In the voice data analysis device 1, the control unit 23 uses a machine learning model to determine the calling period within the recording period of the conversation sound data DA2 (S210). As shown in Fig. 9, the control unit 23 analyzes at least one of the number of times the conversation is called out, the volume of the voice, whether or not a greeting is used, and whether or not the person is called by a nickname in the conversation content during the calling period. With such a voice data analysis device 1, it is possible to analyze the conversation content during the calling state when a product salesperson sells products to people passing by a sales area.

[0104] The control unit 23 uses the machine learning model to determine the standing negotiation period in the recording period of the conversation sound data DA2 (S210). As shown in Fig. 9, the control unit 23 analyzes at least one of the following in the conversation content during the standing negotiation period: voice volume, the number of standing negotiations, whether or not a topic is provided, and whether or not a standing negotiation action is performed. According to such a voice data analysis device 1, it is possible to analyze the conversation content during a standing negotiation state.

[0105] The control unit 23 uses a machine learning model to determine the seated period in the recording period of the conversation sound data DA2 (S210). As shown in Fig. 9, the control unit 23 analyzes the number of times the person sat down to negotiate in the conversation content during the seated period. Alternatively, the control unit 23 may analyze the volume of the voice in the conversation content during the seated period. According to such a voice data analysis device 1, the conversation content during the seated negotiation state can be analyzed.

[0106] (1c) In the voice data analysis device 1, the control unit 23 corrects the period determination result (S210) determined based on the conversation text data TD1 using the matching data (S230). That is, the voice data analysis device 1 determines the start time of the sitting period within the recording period of the conversation sound data DA2 based not only on the conversation text data TD1 but also on the sitting start time information (in other words, start time information) included in the matching data. This enables the voice data analysis device 1 to improve the accuracy of determining the sitting period within the recording period of the conversation sound data DA2.

[0107] (1d) The voice data analysis device 1 includes a display device 41 (or a mobile terminal device 43) that displays various pieces of information based on a display command from the server device 21. The display device 41 (or the mobile terminal device 43) displays at least one piece of analysis information generated by the control unit 23, as shown in FIG.

[0108] Therefore, the voice data analysis device 1 can provide the analysis information to the user of the voice data analysis device 1 by displaying the analysis information on the display device 41 (or the mobile terminal device 43). (1e) In the voice data analysis device 1, the display device 41 (or the mobile terminal device 43) displays the analysis information as time-series data, as shown in Fig. 10 to Fig. 12. Therefore, the voice data analysis device 1 can provide the user of the voice data analysis device with the change in the sales behavior of the product salesperson over time, by the time-series data of the analysis information.

[0109] (1f) In the voice data analysis device 1, the control unit 23 analyzes the content of each of the conversations between the multiple product salespeople. The display device 41 (or the mobile terminal device 43) displays analysis information for each of the multiple product salespeople, as shown in Fig. 9. Thus, the voice data analysis device 1 can provide the user of the voice data analysis device 1 with the analysis information for the multiple product salespeople in a form that allows the analysis information to be compared with each other.

[0110] [1-7. Correspondence of terms] In this embodiment, the device attachment / detachment unit 13 and the control unit 23 correspond to an example of a conversation sound data acquisition unit or an example of a voice data acquisition unit in this disclosure. The control unit 23 corresponds to an example of a conversation sound data extraction unit in this disclosure. The control unit 23 corresponds to an example of a conversation text generation unit in this disclosure. The control unit 23 corresponds to an example of a conversation period determination unit in this disclosure. The control unit 23 corresponds to an example of a conversation content analysis unit in this disclosure.

[0111] In this embodiment, the calling period, the standing negotiation period, and the sitting period correspond to examples of the first period, the second period, and the third period in the present disclosure, respectively. The calling state, the standing negotiation state, and the sitting negotiation state correspond to examples of the first conversation state, the second conversation state, and the third conversation state in the present disclosure, respectively. The sitting start time information included in the matching data corresponds to an example of start time information indicating the start time of the third period in the present disclosure.

[0112] The display device 41 or the mobile terminal device 43 corresponds to an example of an analysis information display unit in the present disclosure. 2. Other embodiments Although the embodiments of the present disclosure have been described above, the present disclosure is not limited to the above-described embodiments and can be implemented in various modified forms.

[0113] (2a) In the above embodiment, a voice data analysis device is described that is configured to record the voice of a salesperson using recording device 11 and transmit voice data (recorded data) from recording device 11 to server device 21, but the voice data analysis device of the present disclosure is not limited to this configuration. For example, the voice data analysis device of the present disclosure may be configured to record the voice of a salesperson using mobile terminal device 43 instead of recording device 11 and transmit voice data from mobile terminal device 43 to server device 21. In this case, the method of transmitting the voice data from mobile terminal device 43 to server device 21 may be wireless communication.

[0114] Furthermore, the voice data analysis device of the present disclosure is not limited to a form in which voice data is temporarily recorded on the mobile terminal device 43, but may be configured so that the voice data collected by the mobile terminal device 43 is immediately transmitted to the server device 21.

[0115] (2b) In the above embodiment, a voice data analysis device configured to generate the items shown in Fig. 8 to Fig. 12 as analysis information has been described, but the analysis information is not limited to the items shown in Fig. 8 to Fig. 12. For example, the total value of the analysis information for all product salespeople belonging to one store may be displayed as the performance data of the store. For example, the total values ​​of the number of times all product salespeople call out to each other, the number of times they negotiate while standing, the number of times they negotiate while sitting, the number of times they make a deal, etc. may be displayed as the performance data of the store.

[0116] Predetermined analysis items during the calling period may include the number of calls, voice volume, greetings, and nicknames.Predetermined analysis items during the stand-up negotiation period may include the number of stand-up negotiations, topics (fees, terminals, signals, data capacity), and stand-up negotiation actions (interest generation, test closing, open questions).Predetermined analysis items during the seated period may include the number of seated negotiations.Predetermined analysis items during other periods may include the number of restricted negotiations.

[0117] (2c) In the above embodiment, the sales activity of a product salesperson analyzed using the voice data analysis device 1 of the present disclosure is the sale of mobile phones at a shopping mall, but the sales activity to be analyzed is not limited to the sale of mobile phones. For example, the sales activity to be analyzed may be the sale of study materials, the sale of home appliances, or membership information for a sports club.

[0118] (2d) In the above embodiment, the voice data analysis device 1 of the present disclosure is configured to acquire voice data DA1 (S110) and perform a process of extracting conversation sound data DA2 from the voice data DA1 (S120), but the voice data analysis device of the present disclosure is not limited to such a configuration. For example, if the recording device 11 stores conversation sound data DA2, the voice data analysis device may be configured to acquire the conversation sound data DA2 instead of the voice data DA1. In this case, the process of extracting the conversation sound data DA2 from the voice data DA1 (S120) can be omitted.

[0119] (2e) In the above embodiment, the analysis target using the voice data analysis device 1 of the present disclosure is the sales activities of a product salesperson, but the analysis target is not limited to the sales activities of a product salesperson. For example, the analysis target may be a business negotiation that has been scheduled in advance.

[0120] In this case, the first conversation state may be a chat state in which chatting other than the main topic is taking place during a pre-arranged business negotiation. The conversation content analysis unit may analyze at least one of "whether or not there is a topic related to the main topic" and "the number of times a topic related to the main topic has been discussed" in the conversation content of the first period. Such a voice data analysis device can analyze the chat content during a pre-arranged business negotiation.

[0121] In this case, the second conversation state may be a main topic conversation state in which a conversation is about the main topic of a previously scheduled business negotiation. The conversation content analysis unit may analyze at least one of "whether or not a proposal related to the main topic is made" and "the number of proposals related to the main topic" in the conversation content of the second period. Such a voice data analysis device can analyze the conversation content in the main topic conversation state.

[0122] (2f) In the above embodiment, the analysis target using the voice data analysis device 1 of the present disclosure is the sales activities of a product salesperson, but the analysis target is not limited to the sales activities of a product salesperson. For example, the analysis target may be a conversation between a supplier and a client.

[0123] In this case, the first conversation state may be a simple negotiation state in a conversation between the deliverer and the client's person in charge. The simple negotiation state may be a state in which a conversation between the deliverer and the client's person in route sales or delivery is about a product or service. The conversation content analysis unit may analyze at least one of the following in the conversation content for the first period: "the number of times a topic related to a product or service was discussed" and "the content of the topic related to a product or service."

[0124] Such a voice data analysis device can analyze the content of a conversation in a simplified business negotiation state between a supplier and a person in charge at a business partner. (2g) Multiple functions possessed by one component in the above embodiments may be realized by multiple components, or one function possessed by one component may be realized by multiple components. Also, multiple functions possessed by multiple components may be realized by one component, or one function realized by multiple components may be realized by one component. Also, part of the configuration of the above embodiments may be omitted. Also, at least part of the configuration of the above embodiments may be added to or substituted for the configuration of another of the above embodiments. [Explanation of symbols]

[0125] 1...voice data analysis device, 11...recording equipment, 11a...microphone, 11b...operation unit, 11c...control unit, 11d...memory unit, 11e...charging terminal, 11f...first communication terminal, 11g...clip unit, 11h...housing, 13...device attachment / detachment unit, 13a...power supply terminal, 13b...second communication terminal, 14...wired cable, 21...server device, 23...control unit, 23a...CPU, 25...memory unit, 27...communication unit, 41...display device, 43...mobile terminal device, DA1...voice data, DA2...conversational sound data, DA3...background sound data, TD1...conversational text data.

Claims

1. A conversation sound data acquisition unit configured to acquire conversation sound data in which a conversation sound is recorded; A conversation text generating unit configured to generate conversation text data which is text data of a conversation content included in the conversation sound data; a conversation period determination unit configured to determine a first period corresponding to a predetermined first conversation state within a recording period of the conversation sound data based on the conversation text data; A conversation content analysis unit configured to analyze the conversation content of the first time period by using the conversation text data for the first time period to generate analysis information; Equipped with The first conversation state is: A salesperson calls out to people passing by the sales area. During a pre-arranged business meeting, you may be chatting away from the main topic of the discussion. A simple sales negotiation state in which a conversation between a supplier and a customer regarding a product or service during route sales or delivery, At least one of the following conditions is satisfied: the conversation period determination unit is configured to determine, by using a machine learning model configured to determine whether or not the conversation text data is the first period in response to input of the conversation text data, a period corresponding to the conversation text data determined to be the first period by the machine learning model, among a recording period of the conversation sound data, as the first period; The conversation content analysis unit is configured to generate information regarding the first conversation state as the analysis information. Audio data analysis equipment.

2. 2. The voice data analysis apparatus according to claim 1, the first conversation state is the calling state, The conversation content analysis unit is configured to analyze at least one of a number of times of calling out to the user, a volume of a voice, whether or not a greeting is used, and whether or not a nickname is used in the conversation content during the first period. Audio data analysis equipment.

3. 2. The voice data analysis apparatus according to claim 1, the first conversation state is the chat state, The conversation content analysis unit is configured to analyze at least one of the presence or absence of a topic related to the main topic in the conversation content of the first period and the number of times the topic related to the main topic is mentioned. Audio data analysis equipment.

4. 2. The voice data analysis apparatus according to claim 1, The first conversation state is the simple negotiation state, The conversation content analysis unit is configured to analyze at least one of the number of times a topic related to the product or the service is discussed in the conversation content during the first period and the content of the topic related to the product or the service. Audio data analysis equipment.

5. 2. The voice data analysis apparatus according to claim 1, the conversation period determination unit is configured to determine, based on the conversation text data, a second period corresponding to a second conversation state different from the first conversation state within a recording period of the conversation sound data; the conversation content analysis unit is configured to analyze the conversation content in the second time period by using the conversation text data in the second time period to generate the analysis information; The second conversation state is A standing sales situation where a product salesperson is talking to people standing in the sales area. During a pre-arranged business meeting, you are in a conversation about the main topic. At least one of the following conditions is satisfied: the conversation period determination unit is configured to determine, by using a machine learning model configured to determine whether or not the conversation text data is the second period in response to input of the conversation text data, a period corresponding to the conversation text data determined to be the second period by the machine learning model, among a recording period of the conversation sound data, as the second period; The conversation content analysis unit is configured to generate information regarding the second conversation state as the analysis information. Audio data analysis equipment.

6. 6. The voice data analysis device according to claim 5, The second conversation state is the negotiation state, The conversation content analysis unit is configured to analyze at least one of a voice volume, a number of standing negotiations, whether or not a topic is provided, and whether or not a standing negotiation action is performed in the conversation content during the second period. Audio data analysis equipment.

7. 6. The voice data analysis device according to claim 5, the second conversation state is the main conversation state; The conversation content analysis unit is configured to analyze at least one of the presence or absence of a suggestion related to the main topic and the number of times the suggestion related to the main topic is made in the conversation content of the second period. Audio data analysis equipment.

8. 7. The voice data analysis device according to claim 6, the conversation period determination unit is configured to determine, based on the conversation text data, a third period corresponding to a third conversation state different from the first conversation state and the second conversation state, within a recording period of the conversation sound data; The conversation content analysis unit is configured to analyze the conversation content in the third time period by using the conversation text data in the third time period to generate the analysis information. Audio data analysis equipment.

9. 9. The voice data analysis device according to claim 8, The third conversation state is a seated negotiation state in which the product salesperson is talking to a person seated in a seat in the sales area, the conversation period determination unit is configured to determine, by using a machine learning model configured to determine whether or not the conversation text data is the third period in response to input of the conversation text data, a period corresponding to the conversation text data determined to be the third period by the machine learning model, among a recording period of the conversation sound data, as the third period; The conversation content analysis unit is configured to analyze at least one of the number of times of seated negotiations and the volume of a voice in the conversation content during the third period. Audio data analysis equipment.

10. 10. The voice data analysis device according to claim 9, the conversation sound data includes a time stamp indicating a correspondence between the conversation sound and a recording time, the conversation period determination unit is configured to determine, based on not only the conversation text data but also start time information indicating a start time of the third period, that the timestamp corresponding to the start time of the third period in the recording period of the conversation sound data is the start time of the third period. Audio data analysis equipment.

11. The voice data analysis device according to any one of claims 1 to 8, an analysis information display unit configured to display at least one of the analysis information by the conversation content analysis unit; Audio data analysis equipment.

12. The voice data analysis apparatus according to claim 11, The analysis information display unit is configured to display the analysis information as time-series data. Audio data analysis equipment.

13. The voice data analysis device according to claim 11, The conversation content analysis unit is configured to analyze the conversation content of each of a plurality of product salespeople; The analysis information display unit is configured to display the analysis information of each of the multiple product salespeople. Audio data analysis equipment.

14. The voice data analysis device according to claim 11, The conversation sound data acquisition unit includes: a voice data acquisition unit configured to acquire voice data including the conversation sound data or the conversation sound data and background sound data; a conversation sound data extraction unit configured to extract the conversation sound data from the voice data; Equipped with Audio data analysis equipment.

15. A computer-implemented method for analyzing speech data, comprising: Obtaining conversation sound data in which the voice of a conversation is recorded; generating conversation text data which is text data of the conversation content included in the conversation sound data; determining a first period corresponding to a predetermined first conversation state within a recording period of the conversation sound data based on the conversation text data; Analyzing the conversation content during the first time period using the conversation text data during the first time period to generate analysis information; Equipped with The first conversation state is: A salesperson calls out to people passing by the sales area. During a pre-arranged business meeting, you may be chatting away from the main topic of the discussion. A simple sales negotiation state in which a conversation between a supplier and a customer regarding a product or service during route sales or delivery, At least one of the following conditions is satisfied: Determining the first period includes using a machine learning model configured to determine whether the conversation text data is the first period in response to input of the conversation text data, and determining, as the first period, a period corresponding to the conversation text data determined to be the first period by the machine learning model, among a recording period of the conversation sound data; generating the analysis information includes generating information about the first conversation state as the analysis information; Methods for analyzing audio data.

16. 16. The method of analyzing speech data according to claim 15, The acquiring of the conversation sound data includes: acquiring the conversation sound data or audio data including the conversation sound data and background sound data; extracting the conversation sound data from the audio data; A voice data analysis method comprising:

17. A computer program for causing a computer to function as at least one of the conversation sound data acquisition unit, the conversation text generation unit, the conversation period determination unit, and the conversation content analysis unit in the voice data analysis device according to any one of claims 1 to 10.

Images