Communication data analysis method

Abstract

This invention provides a communication data analysis method that can analyze the level of relationships between multiple users. [Solution] Server 3 performs an acquisition process (STEP 1) to acquire multiple types of communication data transmitted between the first user M1 and the second user M2 via multiple types of communication services, and a relationship level analysis process (STEP 12) to estimate the relationship level, which represents the level of the relationship between the first user M1 and the second user M2, using a machine learning model based on the order of the communication data sent from the first user M1 to the second user M2 and the type of communication service in the communication data.

Description

Technical Field

[0006] , , ,

[0001] The present invention relates to a communication data analysis method for analyzing multiple types of communication data in multiple types of communication services.

Background Art

[0002] Conventionally, as an analysis method for analyzing communication data in communication services, the one described in Patent Document 1 is known. In this analysis method, by analyzing communication data between multiple terminals stored in multiple terminals and the like, the degree of relevance to a predetermined event (hereinafter referred to as "predetermined event") is evaluated. Specifically, the degree of relevance is evaluated based on the number of occurrences, the frequency of occurrence, or the importance of information related to the predetermined event in the communication data. Then, based on the evaluation result, the relationship between users of multiple terminals is displayed on the monitor.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] According to the above conventional analysis method, although the degree of relevance to a predetermined event in communication data can be analyzed, there is a problem that the level of the relationship (for example, intimacy, etc.) between multiple users cannot be analyzed regardless of the predetermined event.

[0005] The present invention has been made to solve the above problems, and an object thereof is to provide a communication data analysis method capable of analyzing the level of the relationship between multiple users.

Means for Solving the Problems

[0006] To achieve the above objective, the invention according to claim 1 is a communication data analysis method for analyzing multiple types of communication data transmitted between a first user and a second user via multiple types of communication services using a computing device, wherein the computing device performs an acquisition step of acquiring multiple types of communication data, and an analysis step of performing an analysis process that estimates a relationship level representing the level of relationship between the first user and the second user using a machine learning model, based on at least one of the order and communication date and time of the communication data transmitted between the first user and the second user and the type of communication service in the communication data.

[0007] According to this communication data analysis method, the processing unit acquires multiple types of communication data transmitted between the first user and the second user via multiple types of communication services. Of the multiple types of communication data acquired in this way, at least one of the order of the communication data transmitted between the first user and the second user, the date and time of the communication, and the type of communication service in the communication data can be considered to represent the progression and level of the relationship between the first user and the second user.

[0008] Therefore, by estimating the relationship level representing the level of the relationship between the first user and the second user using a machine learning model based on at least one of such order and the date and time of communication occurrence, and the type of communication service in the communication data, the level of the relationship between the first user and the second user can be accurately estimated. Furthermore, if the relationship between the first user and the second user is business-related, the level of the business relationship between the first user and the second user can be accurately estimated, thereby enabling the second user to effectively utilize the estimation results in their business. In this specification, "relationship between the first user and the second user" refers to the degree of intimacy, closeness, and emotional distance between the first user and the second user. Also, "order of communication data" refers to the order of the timing (occurrence time) of the communication data occurrence.

[0009] In the present invention, it is preferable that the arithmetic processing unit further performs a notification step to inform a second user of the relationship level estimation results.

[0010] According to this communication data analysis method, the estimated relationship level is communicated to the second user upon execution of the notification step. This allows the second user to obtain an objective assessment of the relationship level between the first user and the second user, even if the second user is unable to appropriately determine their own relationship with the first user.

[0011] In the present invention, it is preferable that the arithmetic processing unit further performs a proposal step to propose to the second user the use of at least one of a plurality of communication services as a means of communication when the second user contacts the first user, based on the relationship level.

[0012] According to this communication data analysis method, the execution of the proposed step suggests to the second user that they use at least one of several types of communication services as a means of communication when contacting the first user. This improves the convenience of the second user, as they can contact the first user using the suggested communication service even if they are unable to determine for themselves which communication service to use.

[0013] In the present invention, it is preferable that in the proposed step, the timing of use is further proposed to the second user in addition to the use of at least one communication service.

[0014] According to this communication data analysis method, the execution of the proposed step will result in the second user being offered the use and timing of at least one communication service. As a result, even if the second user is unable to determine the timing of using a communication service when contacting the first user, they can use the communication service at the suggested timing to contact the first user, thereby improving the convenience for the second user.

[0015] In the present invention, the first user is composed of multiple types of corporate users, and in the analysis process, it is preferable that the relationship level is estimated by a machine learning model based on the industry of each of the multiple types of corporate users, in addition to the order of the communication data and the date and time of communication occurrence and the type of communication service in the communication data.

[0016] According to this communication data analysis method, the relationship level is estimated based on the order of communication data, the date and time of communication occurrence, the type of communication service in the communication data, and the industry of each of the multiple types of corporate users, by performing the analysis process. This makes it possible to estimate the relationship level in accordance with the industry of the corporate users, thereby improving the accuracy of the relationship level estimation.

[0017] In the present invention, it is preferable that, in the analysis process, the relationship level is estimated by a machine learning model based on the content of the communication data, in addition to the order of the communication data, the date and time of the communication occurrence, and the type of communication service in the communication data.

[0018] According to this communication data analysis method, the relationship level is estimated based on the content of the communication data, in addition to the order of the communication data, the date and time of the communication, and the type of communication service in the communication data, by performing the analysis process. As a result, the relationship level can be estimated in accordance with the content of the communication data, in addition to the order of the communication data, the date and time of the communication, and the type of communication service in the communication data, thereby improving the accuracy of the relationship level estimation. In this specification, "content of communication data" means the content of the data obtained by transcribing the telephone conversation into text (written document) when the communication service is a telephone, and the content of the message when the communication service is an email or the like.

[0019] In the present invention, the first user is composed of multiple types of corporate users, and in the analysis process, it is preferable that the relationship level is estimated by a machine learning model based on the industry of each of the multiple types of corporate users, in addition to the order of the communication data and the date and time of communication occurrence and the type of communication service in the communication data.

[0020] According to this communication data analysis method, by executing the analysis process, in addition to at least one of the order of communication data and the communication date and time, and the type of communication service in the communication data, the relational level is estimated based on the business type of each of a plurality of types of corporate users. Thereby, the relational level can be estimated corresponding to the business type of the corporate user, and thereby, the estimation accuracy of the relational level can be improved.

[0021] In the present invention, the plurality of types of communication services are preferably a plurality of services among telephone services, SNS (Social Networking Service), SMS (Short Message Service), and email services.

[0022] According to this communication data analysis method, the communication data analysis method can be applied to common telephone services, SNS, SMS, and email services as communication services, thereby improving the versatility of the communication data analysis method.

Brief Description of Drawings

[0023] [Figure 1] It is a diagram showing the configuration of a communication system in which a communication data analysis method according to an embodiment of the present invention is executed. [Figure 2] It is a flowchart showing communication data processing. [Figure 3] It is a diagram showing an example of the acquisition result of communication data. [Figure 4] It is a flowchart showing communication data analysis processing. [Figure 5] It is a diagram showing an example of input data. [Figure 6] It is a diagram showing an example of proposed data. [Figure 7] It is a diagram showing another example of input data. [Figure 8] It is a diagram showing another example of the acquisition result of communication data.

Modes for Carrying Out the Invention

[0024] A communication data analysis method according to one embodiment of the present invention will be described below with reference to the drawings. The communication data analysis method of this embodiment is performed in the communication system 1 shown in Figure 1 as described later.

[0025] As shown in Figure 1, the communication system 1 comprises multiple first communication equipment sets 10 (only one set is shown), multiple second communication equipment sets 20, and a server 3. In this communication system 1, the multiple first communication equipment sets 10, the second communication equipment sets 20, and the server 3 are configured to communicate with each other via a communication network 4.

[0026] This communication network 4 is configured to include a fixed telephone network, a wireless telephone network, a wired LAN network, a wireless LAN network, and the Internet.

[0027] Multiple sets of first communication equipment 10 are used by multiple first users M1 (only one is shown in the figure), and in this embodiment, the case in which each of the multiple first users M1 uses each of the multiple sets of first communication equipment 10 to conduct commercial communication with a second user M2 will be described as an example.

[0028] Each first communication equipment set 10 consists of a landline telephone 11, a personal computer 12, and a smartphone 13. This landline telephone 11 is configured to communicate with the landline telephone 21 of the second communication equipment set 20 via the communication network 4.

[0029] Furthermore, the personal computer 12 is configured to communicate with the personal computer 22 or smartphone 23 of the second communication device set 20 via the communication network 4. In addition, the smartphone 13 is configured to communicate with the smartphone 23 or personal computer 22 of the second communication device set 20 via the communication network 4.

[0030] On the other hand, the second communication equipment set 20 is used by the second user M2, and in this embodiment, the case in which the second user M2 uses the second communication equipment set 20 to conduct communication with the first user M1 for commercial or other purposes will be explained as an example.

[0031] The second communication equipment set 20 consists of a landline telephone 21, a personal computer 22, and a smartphone 23. This landline telephone 21 is configured to communicate with the aforementioned landline telephone 11 of the first communication equipment set 10 via the communication network 4.

[0032] Furthermore, the personal computer 22 is configured to communicate with the aforementioned personal computer 12 or smartphone 13 of the first communication device set 10 via the communication network 4. In addition, the smartphone 23 is configured to communicate with the aforementioned smartphone 13 or personal computer 12 of the first communication device set 10 via the communication network 4.

[0033] With the above configuration, this communication system 1 is configured to enable communication between each first user M1 and second user M2 using the following communication services (hereinafter referred to as "communication services"). Specifically, it is configured to enable communication between fixed telephones 11 and 21 using fixed telephone services, and communication between smartphones 13 and 23 using mobile phone services and SMS (Short Message Service).

[0034] Furthermore, the system is configured to enable communication via email services and SNS (Social Networking Service) between PC 12 and PC 22 or smartphone 23, and between smartphone 13 and smartphone 23 or PC 22. In the following explanation, the use of LINE (registered trademark) as the SNS will be used as an example.

[0035] On the other hand, Server 3 is a physical server and is equipped with a processor, storage, I / O interface, and communication circuitry (none of which are shown). As will be described later, Server 3 is configured to acquire communication data from the various communication services described above between the first user M1 and the second user M2, and to analyze the communication data. In this embodiment, Server 3 corresponds to the arithmetic processing unit.

[0036] Server 3 stores a large language model as a machine learning model, and this large language model is used when performing the analysis processing of the communication data described above, as will be explained later. In this case, the large language model stored is one in which the model parameters have been sufficiently trained, and this training process is performed as described below.

[0037] Specifically, first, as a set of training data, a predetermined number of communication data (for example, tens to hundreds of times) are acquired between two users, starting from a predetermined communication occurrence timing (in this embodiment, the first communication occurrence timing). In this case, the communication data is transmitted between two communication device sets, similar to the first communication device set 10 and the second communication device set 20 described above.

[0038] Then, the type of communication service, the order of the communication data (communication sequence), and the date and time of communication are obtained from the acquired communication data. By adding a relationship level, which represents the relationship between one user and the other user at the time a predetermined number of communications have been completed, as a label to this data, training data is created.

[0039] In this case, the relationship level is set as an integer value within a predetermined range, and the higher the relationship between the two, the larger the value. In this embodiment, the predetermined range used is the range of values ​​1 to 5. However, the predetermined range is not limited to the range of values ​​1 to 5; a different range (for example, the range of values ​​1 to 3, 1 to 4, or 1 to 6) may also be used. Then, the model parameters of the large-scale language model are trained using this training data.

[0040] Similarly, by using a large number of training data sets (for example, several hundred to several thousand sets), the model parameters of the large-scale language model are trained, thereby generating a large-scale language model with sufficiently trained model parameters. In the case of such a large-scale language model, when the type of communication service, the order of the communication data, and the date and time of communication for a predetermined number of communication data between two users are input, the relationship level of one user to the other user will be output as a value within the predetermined range mentioned above.

[0041] Furthermore, Server 3 stores the user ID, username, landline phone number, mobile phone number, email address, and LINE account of each of the multiple first users M1, linked to each other. Additionally, Server 3 stores the user ID, username, landline phone number, mobile phone number, email address, and LINE account of the second user M2, linked to each other.

[0042] Furthermore, Server 3 stores a database for creating proposal data (see Figure 6), which will be described later, based on the relationship level.

[0043] Next, we will explain the various processes performed by Server 3. First, we will explain the communication data processing with reference to Figure 2. This communication data processing is performed when communication occurs between one of the multiple First Users M1 (hereinafter referred to as "Target First User M1") and Second User M2 using the various communication services mentioned above.

[0044] As shown in Figure 2, first, the communication data acquisition process is executed (Figure 2 / STEP 1). In this communication data acquisition process, when communication occurs between the target first user M1 and the second user M2, the communication data is acquired. Specifically, the communication data acquired includes the date and time of communication, the username of the target first user M1, the user ID, the person in charge ID (ID of the second user M2), and the type of service (see Figure 3). In this embodiment, the communication data acquisition process corresponds to the acquisition step.

[0045] Next, the communication data storage process is executed (Figure 2 / STEP 2). In this communication data storage process, the communication data acquired as described above is stored in Server 3. As the above process is repeatedly executed, a large amount of communication data is stored in Server 3, as shown in Figure 3.

[0046] Next, the communication data analysis process will be explained with reference to Figure 4. This communication data analysis process analyzes the communication data stored in Server 3, as described below.

[0047] As shown in Figure 4, first, it is determined whether or not the execution conditions for the communication data analysis process are met (Figure 4 / STEP10). Specifically, if the following condition (a) is met, it is determined that the execution conditions are met, and otherwise, it is determined that the execution conditions are not met.

[0048] (a) Among the multiple first users M1, there is a first user M1 that has not yet performed this communication data analysis process, and the number of communications with the second user M2 has reached a predetermined number, starting from a predetermined communication occurrence timing (in this embodiment, the first communication occurrence timing).

[0049] If this determination is negative (Figure 4 / STEP10...NO) and the execution condition is not met, the process ends immediately. On the other hand, if this determination is positive (Figure 4 / STEP10...YES) and the execution condition is met, the input data creation process is executed (Figure 4 / STEP11).

[0050] In this input data creation process, input data for input into the large-scale language model is created based on the communication data between the first user M1 and the second user M2, which are stored in Server 3. As shown in Figure 5, this input data is arranged in order of occurrence, communication date and time, and type of communication service.

[0051] Next, the relationship level acquisition process is executed (Figure 4 / STEP 12). In this relationship level acquisition process, the above input data is input to a large-scale language model, and the relationship level is acquired (estimated) as an output from the large-scale language model. This relationship level represents the level of relationship between the first user M1, who is the target of analysis, and the second user M2. In the following explanation, we will use the case where the relationship level is acquired as the value "3.1" as an example. Also, in this embodiment, the relationship level acquisition process corresponds to the analysis step.

[0052] Next, the proposal data creation process is executed (Figure 4 / STEP 13). In this proposal data creation process, proposal data is created by referring to the database in Server 3 based on the relationship levels obtained as described above. This proposal data is intended to propose to the second user M2 the timing of contact with the first user M1 being analyzed and the communication service to be used for that contact.

[0053] The database defines the relationship between the relationship level and the proposed content of the communication service to be used for contact, and the proposed data is created accordingly, for example, as shown in Figure 6. Specifically, the proposed data includes the relationship level value, the name of the first user M1 being analyzed as the contact recipient, the contact timing, and the type of communication service as the proposed content.

[0054] Next, the proposal data transmission process is executed (Figure 4 / STEP 14). In this proposal data transmission process, the aforementioned proposal data is sent to the second user M2's PC 22 and smartphone 23. As a result, the proposal data shown in Figure 6 is displayed on the displays of the PC 22 and smartphone 23. In this embodiment, the proposal data transmission process corresponds to the notification step and the proposal step.

[0055] As described above, according to the communication data analysis method of this embodiment, the server 3 acquires communication data from landline and mobile phone telephone services, SMS, email services, and SNS between the first user and the second user. If there is a first user M1 to be analyzed (the first user M1 whose number of communications with the second user M2 reaches a predetermined number, starting from a predetermined communication occurrence timing), the input data creation process (Figure 4 / STEP11) and the relationship level analysis process (Figure 4 / STEP12) are executed to estimate the relationship level, which represents the level of the relationship between the first user and the second user.

[0056] Here, the order of communication data exchanged between the first and second users, the date and time of communication, and the type of communication service in the communication data can be considered to represent the progression and level of the relationship between the first user and the second user. Therefore, by inputting the above input data into a machine learning model and estimating the relationship level between the first user and the second user, this can be estimated with high accuracy.

[0057] Furthermore, the proposed data is sent to the second user M2 when the proposed data creation process (Figure 4 / STEP 13) and the proposed data transmission process (Figure 4 / STEP 14) are executed. This proposed data includes the estimated relationship level, contact timing, and contact service, and is displayed on the first user M1's PC 12 and smartphone 13. As a result, even if the second user cannot appropriately judge their relationship with the first user M1 being analyzed, they can obtain an objective judgment result regarding the level of relationship between the second user and the first user M1 being analyzed.

[0058] Furthermore, even if the second user cannot determine when and which communication service to use when contacting the first user M1 being analyzed, they can contact the first user M1 being analyzed using the proposed communication service at the proposed communication timing, thereby improving convenience for the second user.

[0059] Furthermore, if multiple first users M1 consist of multiple types of corporate users, the system may be configured to estimate the relationship level using a machine learning model (e.g., a large-scale language model) based on the order of the communication data and the type of communication service in the communication data, as well as the industry of each of the multiple types of corporate users (e.g., finance, manufacturing, sales, etc.).

[0060] In that case, a machine learning model can be created by training the model parameters of the aforementioned machine learning model for each industry, storing it in server 3, and then, in STEP 12 of Figure 4, the relationship level estimation can be performed by using the machine learning model corresponding to the industry of the first user M1 being analyzed.

[0061] With this configuration, the relationship level can be estimated according to the industry of the corporate user, thereby improving the accuracy of the relationship level estimation.

[0062] Furthermore, a machine learning model may be used that outputs a relationship level when the content of the communication data (hereinafter referred to as "communication content") is input, in addition to the order of the communication data, the date and time of the communication, and the type of communication service in the communication data. For example, in STEP 12 of Figure 4, a machine learning model may be used that outputs a relationship level when input data such as that shown in Figure 7 is input. In the data shown in Figure 7, the communication content of landline and mobile phones has been converted into text by speech recognition processing on server 3.

[0063] Furthermore, when training the model parameters of a machine learning model, the training data used includes data with relationships labeled as levels, representing the order of communication data, the date and time of communication, the type of communication service in the communication data, and the content of the communication (for example, the content shown in Figure 8).

[0064] With the configuration described above, it is possible to estimate the relationship level in relation to the content of the communication, in addition to the order of the communication data and the type of communication service in the communication data, thereby improving the accuracy of the relationship level estimation.

[0065] In addition, if multiple primary users M1 consist of multiple types of corporate users, a machine learning model can be created by training the model parameters of the above machine learning model for each industry. In this case, the accuracy of relationship level estimation can be further improved.

[0066] Furthermore, although this embodiment uses a physical server as server 3, a virtual server may be used instead. In that case, the physical server constituting the virtual server corresponds to the processing unit.

[0067] Furthermore, although this embodiment uses Server 3 as the processing unit, alternatively, a personal computer may be used as the processing unit, multiple personal computers may be used in combination, a personal computer and a server may be used in combination, or a cloud computing system may be used.

[0068] On the other hand, while the embodiment uses a large-scale language model as the machine learning model, alternatively, a neural network, CNN (Convolutional Neural Network), or RNN (Recurrent Neural Network) may be used as the machine learning model, or a regression model such as a random forest may be used.

[0069] Furthermore, while the embodiment is an example in which the relationship level is estimated by a machine learning model based on the order of communication data communicated between the first user and the second user, the date and time of communication, and the type of communication service in the communication data, the machine learning model may also be configured to estimate the relationship level based on either the order of communication data or the date and time of communication, and the type of communication service in the communication data.

[0070] Furthermore, while the embodiment uses data communicated bidirectionally between the first and second users as the communication data communicated between them, alternatively, only the communication data communicated from the first user to the second user may be used as the communication data communicated between the first and second users.

[0071] Furthermore, while the embodiment uses telephone services via landline and mobile phones, SMS, email services, and SNS as communication services, two or more of these communication services may be used. In that case, a machine learning model may be used in which the communication data of these two or more communication services is used as training data to train the model parameters.

[0072] Furthermore, in the execution condition determination process (STEP 10) in Figure 4, instead of the aforementioned condition (a), it may be determined that the execution condition is met when the following condition (b) is met. (b) When the first user M1 who wants to perform the communication analysis processing is determined by the second user M2, and data indicating the user name or user ID of the first user M1 is transmitted from the second communication equipment set 20 and then received by the server 3.

[0073] Furthermore, while the embodiment shows an example where the predetermined communication occurrence timing is set as the first communication occurrence timing, the predetermined communication occurrence timing is not limited to this, and can be set to an appropriate timing such as the second or third time. In that case, the predetermined communication occurrence timing in the training data of the large-scale language model should also be set to the same timing.

[0074] Furthermore, while the proposed data in Figure 6 shows an example where the contact period is set to within 4 days, the contact period can also be set by date or by a predetermined period (for example, a period of 2 to 4 days). [Explanation of Symbols]

[0075] 1. Communication System 3. Server (Processing Unit) M1 First User M2 Second User

Claims

1. A communication data analysis method comprising analyzing multiple types of communication data transmitted between a first user and a second user via multiple types of communication services using a computing device, The aforementioned arithmetic processing unit is The acquisition step involves acquiring the aforementioned multiple types of communication data, An analysis step in which, from among the multiple types of communication data acquired, an analysis process is performed using a machine learning model to estimate a relationship level representing the level of the relationship between the first user and the second user, based on at least one of the order of the communication data communicated between the first user and the second user and the date and time of the communication, and the type of communication service in the communication data; A method for analyzing communication data, characterized by performing the following actions.

2. In the communication data analysis method described in claim 1, The aforementioned arithmetic processing unit is A communication data analysis method characterized by further performing a notification step to notify the second user of the estimation result of the relationship level.

3. In the communication data analysis method described in claim 1, The aforementioned arithmetic processing unit is A communication data analysis method characterized by further performing a proposal step to suggest to the second user the use of at least one of the plurality of communication services as a means of communication when the second user contacts the first user, based on the relationship level.

4. In the communication data analysis method described in claim 3, The communication data analysis method is characterized in that, in the proposed step, the timing of the use of the at least one communication service is proposed to the second user.

5. In the communication data analysis method described in claim 1, The aforementioned first user consists of multiple types of corporate users, The communication data analysis method is characterized in that, in the analysis process, the relationship level is estimated by the machine learning model based on at least one of the order of the communication data and the date and time of the communication, as well as the type of communication service in the communication data, and the industry of each of the multiple types of corporate users.

6. In the communication data analysis method described in claim 1, The communication data analysis method is characterized in that, in the analysis process, the relationship level is estimated by the machine learning model based on the content of the communication data, in addition to the order of the communication data, the date and time of the communication occurrence, and the type of communication service in the communication data.

7. In the communication data analysis method described in claim 6, The aforementioned first user consists of multiple types of corporate users, The communication data analysis method is characterized in that, in the analysis process, the relationship level is estimated by the machine learning model based on the order of the communication data, the date and time of the communication, the type of communication service in the communication data, the content of the communication data, and the industry of each of the multiple types of corporate users.

8. In the communication data analysis method according to any one of claims 1 to 7, A method for analyzing communication data, characterized in that the aforementioned multiple types of communication services are multiple services from among telephone services, SNS (Social Networking Service), SMS (Short Message Service), and email services.

Images