JP7465030B1 - Information processing system, information processing method, and program - Google Patents

Abstract

[Problem] To evaluate and understand before sending a message created by a user whether or not the message appropriately expresses positive feelings toward the recipient user. [Solution] An information processing system has a control unit. The control unit inputs a message that has been created by the user using a business communication tool at a sender terminal and has not yet been sent to a recipient to a learning model that has undergone machine learning to learn information regarding appropriate expressions of positive feelings toward the recipient contained in the message. The control unit then receives, from the learning model, advice information regarding the appropriate expressions in the input message, and transmits the advice information to the sender terminal for display. [Selected Figure] Figure 4

Description

The present invention relates to an information processing system, an information processing method, and a program that can assist in creating messages to be sent in a business communication tool.

Conventionally, when an employee sends a message to another employee within a company using a business communication tool, there is technology that analyzes the message sent and evaluates the user.

For example, the following Patent Document 1 describes a method for diagnosing activity levels and judging the risk of quitting a company based on the number of thank-you messages sent or received by each user within a specified period of time.

Patent No. 6379270

However, conventional technologies such as those described in Patent Document 1 do not analyze the contents of messages, and depending on the content of the message sent, the sender's emotions may not be properly conveyed to the recipient.

In view of the above circumstances, the object of the present invention is to provide an information processing system, information processing method, and program that can evaluate and allow a user to determine before sending a message whether the message is appropriately expressing positive feelings toward the recipient user.

To achieve the above object, an information processing system according to one embodiment of the present invention has a control unit. The control unit inputs a message that has been created by a user using a business communication tool at a sender terminal and has not yet been sent to a recipient into a learning model that has been machine-learned to learn information regarding appropriate expressions of positive feelings toward a recipient that are included in the message. The control unit then receives advice information regarding the appropriate expressions in the input message from the learning model, and transmits the advice information to the sender terminal for display.

This allows the information processing system to evaluate and determine before sending whether a message created by a user appropriately expresses positive feelings toward the recipient user, allowing the system to send a high-quality message. The above learning model is, for example, a generative AI such as ChatGPT based on a large-scale language model (LLM), but is not limited to this.

The control unit may receive evaluation value information indicating an evaluation value regarding the expression of emotions in the input message from the learning model, and transmit the evaluation value information to the sender terminal for display.

This allows the information processing system to allow the user to objectively understand how appropriately emotions are expressed in the messages they create.

The control unit may calculate the evaluation value based on a predetermined index related to expressions of gratitude or praise contained in the message, and may train the learning model to generate information related to the index related to the calculation of the evaluation value as the advice information.

This allows the information processing system to generate the evaluation value as an objective index, and allows the user to understand information about the index (the basis on which the evaluation value was calculated) and improve the message.

The information processing system may further include a storage unit that stores user information about a user of the business communication tool in association with advice request right information about the right of the user to request the advice information from the learning model. In this case, when the control unit receives an advice request for the advice information from the business communication tool used by the user, the control unit may determine whether or not the right exists based on the stored advice request right information, and if it determines that the right exists, may request the advice information from the learning model together with inputting the message.

As a result, when the use of the learning model incurs a usage cost, for example, on a pay-as-you-go basis, if a user uses the learning model indefinitely, this would put a significant strain on the user's company budget. In this way, the information processing system can manage usage per user so that the amount of usage is appropriate for the budget, whereas if a user uses the learning model indefinitely, this would put a significant strain on the user's company budget.

The information processing system may further include a storage unit that stores the user's behavior history information accumulated in a predetermined business tool. In this case, the control unit may calculate a business relationship value indicating the level of the business relationship between the user and other users based on the stored behavior history information, train the learning model to extract other users with a high business relationship value, receive information on other users with a high business relationship value from the learning model, calculate a statistical value by calculating the evaluation value in the advice information created so far for each user to whom the message is to be sent, train the learning model to extract other users with a low statistical value, receive information on other users with a low statistical value from the learning model, and transmit recommendation information to the sender terminal that recommends other users with a high business relationship value and a low statistical value as a new destination to which the message is to be sent.

This allows the information processing system to prompt the user to send a message to other users who have a strong business relationship with the user but who have not been able to adequately convey feelings of gratitude or praise in their messages. The above statistical values are, for example, average values, maximum values, minimum values, etc., but may also be values that have undergone other mathematical processing.

In response to a request from the sender terminal, the control unit may display, as the recommendation information, a list of other users who have a high business relationship value and a low statistical value on the sender terminal together with a user interface for displaying a creation screen for a message addressed to the other user.

This allows the information processing system to display a list of other users to whom messages should be sent along with an interface for creating messages, allowing the user to intuitively perform tasks from understanding the need to send a message to creating the message.

The control unit may train the learning model to respond with information on the basis when the user asks the control unit about the basis for determining that the other user's business relationship value is high and the statistical value is low, and when the control unit receives the question from the user from the sender terminal via a chatbot, the control unit may receive the information on the basis from the learning model and display it on the sender terminal via the chatbot.

This allows the information processing system to allow the user to intuitively understand, via the chatbot, objective reasons why they should send a message to another user.

The control unit may display, on the sender terminal, a list of other users having a high business relationship value, together with a user interface for requesting the statistical values of the other users, and when any other user is selected through the user interface, may display, on the sender terminal, information indicating the statistical values of the other users, together with a user interface for displaying a creation screen for a message addressed to the other user.

In this way, the information processing system allows the user to select a user of interest from among other users to whom to send messages, and by allowing the user to understand the evaluation values of messages sent to that user to date, the information processing system can better impress upon the user the need to send messages to that other user.

An information processing method according to another aspect of the present invention includes:
A message that has been created by a user using a business communication tool on a sender terminal and has not yet been sent to a recipient is input to a learning model that has been machine-learned to learn information regarding appropriate expressions of positive emotions toward a recipient that are included in the message, and
receiving, from the learning model, advisory information regarding the appropriate expression in the input message;
and transmitting the advice information to the sender terminal for display.

According to yet another aspect of the present invention, there is provided a program for an information processing device,
A step of inputting a message that has been created by a user using a business communication tool at a sender terminal and has not yet been sent to a recipient into a learning model that has been machine-learned to learn information regarding appropriate expressions of positive feelings toward a recipient contained in the message;
receiving, from the learning model, advisory information regarding the appropriate expression in the input message;
and transmitting the advice information to the sender terminal and displaying the advice information.

As described above, according to the present invention, a user can evaluate and understand before sending a message whether the message he or she created appropriately expresses positive feelings toward the recipient user. However, this effect does not limit the present invention.

1 is a diagram showing a configuration of a message creation support system according to a first embodiment of the present invention. FIG. 2 is a diagram showing a hardware configuration of a mediation server according to the first embodiment of the present invention. 2 is a diagram showing a configuration of a database included in the intermediation server according to the first embodiment of the present invention; FIG. 4 is a sequence diagram showing a flow of advice information transmission processing by the message creation support system according to the first embodiment of the present invention. FIG. 4 is a diagram showing an example of a user interface of advice information that the intermediation server according to the first embodiment of the present invention causes a user terminal to display. FIG. 4 is a sequence diagram showing a flow of a process for extracting recipients of a message by the message creation support system according to the first embodiment of the present invention. FIG. 10 is a diagram showing an example of a user interface of message recipient recommendation information displayed on a user terminal by the intermediation server according to the first embodiment of the present invention; FIG. 10 is a diagram showing an example of a user interface of message recipient recommendation information displayed on a user terminal by the intermediation server according to the first embodiment of the present invention; FIG. 10 is a diagram showing an example of a user interface of message recipient recommendation information displayed on a user terminal by the intermediation server according to the first embodiment of the present invention; FIG. FIG. 11 is a sequence diagram showing the flow of a process for extracting recipients of a message by the message creation support system according to the second embodiment of the present invention. 13 is a diagram showing an example of a user interface of message recipient recommendation information displayed on a user terminal by an intermediation server according to the second embodiment of the present invention. FIG. 13 is a diagram showing an example of a user interface of message recipient recommendation information displayed on a user terminal by an intermediation server according to the second embodiment of the present invention. FIG. FIG. 13 is a sequence diagram showing the flow of a process for extracting recipients of a message by the message creation support system according to the third embodiment of the present invention. FIG. 13 is a diagram showing an example of a user interface for message recipient recommendation information displayed on a user terminal by an intermediation server according to the third embodiment of the present invention. FIG. 13 is a diagram showing an example of a user interface for message recipient recommendation information displayed on a user terminal by an intermediation server according to the third embodiment of the present invention. FIG. 13 is a sequence diagram showing the flow of a process for extracting recipients of a message by the message creation support system according to the fourth embodiment of the present invention. FIG. 13 is a sequence diagram showing the flow of a process for extracting recipients of a message by the message creation support system according to the fifth embodiment of the present invention.

The following describes an embodiment of the present invention with reference to the drawings.

First Embodiment
First, a first embodiment of the present invention will be described.

[System Configuration]
FIG. 1 is a diagram showing the configuration of a message creation support system according to this embodiment.

As shown in the figure, this system is configured by connecting two intermediation servers 100A and 100B, a generation system AI 300, and multiple user terminals 200 of each company via a network such as the Internet 50.

The user terminal 200 is a terminal used by a user who is an employee of a company, and is, for example, a notebook PC (Personal Computer), a desktop PC, a smartphone, a tablet PC, etc.

A business communication tool 20 (hereinafter simply referred to as business tool 20) is installed on the user terminal 200 and is used by employees of each company for business-related communication.

Examples of business communication tools 20 include email clients such as Outlook (registered trademark), groupware such as Microsoft Teams (registered trademark) and Slack (registered trademark), messaging apps, group chat apps, business SNS (Social Networking Services), and also tools specialized for emotional communication between employees (expressing gratitude and praise to colleagues within the company) (hereinafter also referred to as "emotional communication tools").

The intermediary server 100A is a server that mediates communication between user terminals 200 that use common business tools such as Outlook (registered trademark) and Microsoft Teams (registered trademark) among the business tools.

The intermediary server 100B is a server that mediates communication between user terminals 200 using the emotional communication tools between employees among the business tools.

The generative AI 300 is a learning model based on a large-scale language model (LLM), such as ChatGPT, and is connected to the intermediary servers 100A and 100B via an API.

In this embodiment, the intermediary servers 100A and 100B execute a process to assist in message creation when a user (employee) of the user terminal 200 sends a message for the above-mentioned emotional communication (a message of gratitude or praise) to another employee.

Specifically, the intermediary server 100B can have the generation AI 300 evaluate whether the message created by the user appropriately expresses positive feelings toward the recipient user before the message is sent, generate advice information, and provide the advice information to the user terminal 200.

Furthermore, in this embodiment, the intermediary servers 100A and 100B can induce the user to create a message by having the generative AI 300 extract other employees to whom a thank you/praise message should be sent based on the user's past exchanges of thank you/praise messages with other employees and business interactions using business tools, and providing information about the employees to whom the message should be sent to the user terminal 200.

[Hardware configuration of the mediation server]
2 is a diagram showing the hardware configuration of the intermediation server 100B. As shown in the figure, the intermediation server 100B includes a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, an input/output interface 15, and a bus 14 connecting these components to each other.

The CPU 11 accesses the RAM 13 etc. as necessary and performs various calculation processes while controlling each block of the intermediation server 100B as a whole. The ROM 12 is a non-volatile memory in which firmware such as the OS, programs and various parameters to be executed by the CPU 11 are permanently stored. The RAM 13 is used as a working area for the CPU 11 and temporarily stores the OS, various applications being executed, and various data being processed.

The input/output interface 15 is connected to a display unit 16, an operation reception unit 17, a memory unit 18, a communication unit 19, etc.

The display unit 16 is a display device that uses, for example, an LCD (Liquid Crystal Display), an OELD (Organic ElectroLuminescence Display), a CRT (Cathode Ray Tube), etc.

The operation reception unit 17 is, for example, a pointing device such as a mouse, a keyboard, a touch panel, or other input device. If the operation reception unit 17 is a touch panel, the touch panel may be integrated with the display unit 16.

The storage unit 18 is a non-volatile memory such as a hard disk drive (HDD), a flash memory (SSD; solid state drive), or other solid-state memory. The storage unit 18 stores the OS, various applications, and various data.

In particular, in this embodiment, the storage unit 18 stores user information of the emotional communication tool, information regarding thank-you and praise messages exchanged between users using the tool, and the like, and also stores applications and other programs that allow the intermediation server 100B to execute message creation support processing using this data. As will be described later, the storage unit 18 has databases containing such data, including a user information database, an advice request right information database, and an emotional score information database.

The communication unit 19 is, for example, a NIC (Network Interface Card) for Ethernet or various modules for wireless communication such as wireless LAN, and is responsible for communication processing between the user terminal 200, the intermediation server 100A, and the generation system AI 300.

Although not shown, the hardware configuration of the intermediation server 100A and the hardware configuration of the user terminal 200 are similar to the hardware configuration of the intermediation server 100B.

[Mediation server database configuration]
FIG. 3 is a diagram showing the configuration of a database included in the intermediate server 100B.

As shown in the figure, the intermediation server 100B has a user information database 31, an advice request right information database 32, and an emotional score information database 33 in the memory unit 18.

The user information database 31 stores the user information of each employee, i.e., user name, user ID, job title, department, date of joining the company, age, gender, address, profile image, etc.

The advice request right information database 32 stores information regarding the user's right to request advice on thank-you/praise messages from the generation AI 300. This is because, if the generation AI 300 is a pay-per-use system, and the user uses the generation AI 300 endlessly, this puts a significant strain on the user's company's budget, so the amount of usage of the generation AI 300 per user is managed so that the amount of usage is appropriate in light of the budget.

Specifically, the intermediation server 100B manages the user's right to make an advice request to the generation system AI 300 (hereinafter also referred to as the "advice request right") by using one or a combination of the following four management methods (1) to (4). Note that the following detailed conditions are examples and are not limited to these.

(1) Managing the number of times advice requests can be made: Each user is granted the right to make advice requests five times per month, for example, and the remaining number is managed.
(2) Management by points: 1,000 points are given to users per month, and a control is set up so that 200 points are consumed per advice request, and the remaining points are managed.
(3) Managing the time between a previous advice request and the next one: Setting a control requiring a gap of at least two days between a previous advice request and the next advice request, the interval between advice requests is managed by time.
(4) Managing actions from the previous advice request to the next: When an advice request is made, the advice request is locked, and the lock is released by the action of sending a thank you/praise message. This control is set up, and management is done based on whether or not an action of sending a thank you/praise message is taken between advice requests.

The advice request right information database 32 manages information such as the remaining number of advice requests per month, remaining points per month, the time elapsed since the last advice request, and whether or not the advice request is locked for each user, depending on the management methods (1) to (4) above.

The emotional score information database 33 stores, for each user and for each message, information indicating an evaluation value (emotional score) that the generative AI 300 generates by evaluating expressions of gratitude or praise contained in a message when the user requests advice from the generative AI 300.

Although not shown, the intermediary server 100A accumulates behavioral history data of each user in business tools such as the above-mentioned Outlook (registered trademark) and Microsoft Teams (registered trademark). Examples of behavioral history information include, but are not limited to, the number of times chat messages were sent, the time spent in online meetings, the number of documents shared, the number of mentions, the number of emails sent and received, and the sending and receiving of reactions (likes, emoji stamps).

These databases are used by mutual reference as necessary in the message creation support process by intermediary servers 100A and 100B, which will be described later.

[System Operation]
Next, the operation of the message creation support system configured as above will be described. The operation is executed by the cooperation of hardware such as the CPU 11 and communication unit 19 of the transfer server 100B and the CPU and communication unit of the transfer server 100A, and the storage unit 18 of the transfer server 100B and the software stored in the storage unit of the transfer server 100A. In the following description, for convenience, the CPUs of the transfer servers 100A and 100B are regarded as the operating subjects.

[Advice Information Providing Process]
First, the process of providing advice information for the thank-you/praise message by the message creation support system will be described.

Figure 4 is a sequence diagram showing the flow of advice information transmission processing by the system. Also, Figure 5 is a diagram showing an example of a user interface for advice information that the intermediary server 100B displays on the user terminal 200. Figure 5 (A) is an example of the display when the user has the right to request advice and the message satisfies the advice availability condition, and Figure 5 (B) is an example of the display when the user does not have the right to request advice or the message does not satisfy the advice availability condition.

As shown in FIG. 4, first, the user terminal 200 creates a thank you/praise message by accepting input from the user in the message input field 51 of the thank you/praise message creation screen on the emotional communication tool as shown in FIG. 5 (step 41).

Then, the user terminal 200 accepts a request for advice regarding the created thank-you/praise message via the advice request button 52 located, for example, at the bottom right of the message input field 51 on the thank-you/praise message creation screen (step 42), and transmits the advice request to the intermediary server 100B (step 43).

When the CPU 11 of the intermediation server 100B receives the advice request (step 44), it refers to the advice request right information database 32 and determines whether the user who made the advice request has the right to request advice (step 45).

If it is determined that the user does not have the right to request advice, the CPU 11 of the intermediation server 100B displays information to that effect in the advice display field 53 of the thank you/praise message creation screen, as shown in FIG. 5(B). Alternatively, in this case, the CPU 11 may control the advice request button 52 itself to be inactive (to be inoperable).

If it is determined that the user has the right to request advice, the CPU 11 of the intermediation server 100B determines whether the created message satisfies the advice availability conditions (step 46).

Specifically, the CPU 11 determines that a message that satisfies the following two conditions satisfies the advice availability condition, but the determination conditions are not limited to these.
・The number of characters must be 50 or more.
- Words of gratitude such as "thank you" or "thank you very much" or words of praise such as "amazing,""wonderful," or "amazing" are used at least once.

If it is determined that the message does not satisfy the advice availability condition, the CPU 11 of the intermediation server 100B displays information to that effect in the advice display field 53 of the thank you/praise message creation screen, as shown in FIG. 5(B). Alternatively, in this case, the CPU 11 may control the advice request button 52 itself to be inactive (to be inoperable).

If it is determined that the message satisfies the advice availability condition, the CPU 11 of the intermediation server 100B sends an advice generation request together with the message body to the generation system AI 300 (step 47).

When the generative AI 300 receives the advice generation request (step 48), it analyzes the message, calculates an emotional score based on predetermined indicators related to expressions of gratitude and praise contained in the message, and generates advice information including information indicating areas for improvement (step 49).

In order for the generative AI 300 to execute this process, the CPU 11 of the intermediation server 100B trains the generative AI 300 via an API to calculate the emotional score of a message based on, for example, the following four indicators (1) to (4) and judgment conditions. The indicators and judgment conditions are not limited to these.

(1) Repeatedly expressing gratitude or praise (maximum 10 points): If this is done once, it is calculated as 5 points, and if it is done twice, it is calculated as 10 points.
(2) The reasons for gratitude or praise are mentioned in detail (maximum 40 points): The amount of detail and appropriateness of the explanation of what the other person did, in what situation, and how that contributed is judged based on the number of expressions used and matching with pre-registered expressions.
(3) Expressing gratitude or praise to the other person (as expected, thanks to..., etc.) (maximum 40 points): A score is calculated based on whether or not the speaker was able to pay attention to the other person's characteristics (skills, mindset, personality, etc.) and express them appropriately.
(4) Promising to reciprocate (maximum 10 points): If no promise to reciprocate is written, the score is calculated as 0; if it is written, the score is calculated as full points.

The CPU 11 uses sentences that will receive high scores and sentences that will receive low scores in light of the above indicators (1) to (4) as training data, and repeatedly executes a process in the generative AI 300 to return a score in response to input of a sentence, so that the former will receive a high score and the latter will receive a low score.

The weighting is set as follows to calculate the emotional score: (1) 10%, (2) 40%, (3) 30%, and (4) 20%, but this is not limited to this.

Furthermore, the CPU 11 trains the generative AI 300 to generate comments that point out the factors that resulted in a low score in the indicators used to calculate the emotional score as areas for improvement, and to respond with these comments together with the emotional score. For example, if the indicator in (2) above is not specific enough, the generative AI 300 is trained to respond with more specific comments.

Depending on the learning status of the generative AI 300, a new learning process for generating the above advice information may be added, or the above learning process may be deleted, and the presence or absence of a learning process may be determined as appropriate depending on the situation.

After the generation AI 300 has performed the above-described learning, it generates advice information and transmits it to the intermediation server 100B (step 50). Then, when the intermediation server 100B receives the advice information, it transmits it to the user terminal 200 (step 51), and displays it in the advice display field 53 on the message creation screen of the user terminal 200, as shown in FIG. 5(A) (step 52).

The advice in Figure (A) has an emotional score of 74, and among the indicators above, the repeated use of words of gratitude and the conveying of praise for the other person are highly rated points, but the advice also includes a comment that the reason for the gratitude is not specific and is an area for improvement.

Then, when the user of the user terminal 200 improves the message based on the advice and presses a send button 54, for example in the lower right corner of the message creation screen, the user terminal 200 sends the message to the destination user terminal 200 (step 53).

[Message recipient extraction process]
Next, the process of extracting the recipients to whom the user should send the message when creating the thank-you/praise message by the intermediary servers 100A and 100B will be described. Fig. 6 is a sequence diagram showing the flow of the process of extracting recipients to whom the message creation support system should send the message. Figs. 7 to 9 are diagrams showing examples of user interfaces of the message recipient recommendation information that the intermediary server 100B displays on the user terminal 200.

As shown in FIG. 6, first, the user terminal 200 accepts a request to extract colleagues with low emotional evaluation values from among those who have a high business relationship value with the user, based on the user's behavioral history data stored in the intermediary server A (step 61).

The request for the extraction process is accepted, for example, when a pressing operation of a sending opportunity question button 71, which is provided on the user's My Page and is used to ask the generation system AI 300 about a sending opportunity (a sending recipient) for a message, is detected, as shown in FIG. 7.

When the extraction request is accepted, the user terminal 200 transmits the extraction request to the intermediation server 100A (step 62). When the intermediation server 100A receives the extraction request (step 63), it transmits an extraction processing request to the generation system AI 300 (step 64).

Upon receiving the extraction process request, the generative AI 300 extracts colleagues who have a strong business connection with the user (high business relationship value) based on the user's behavioral history data accumulated in the business tool (step 65).

In order for the generative AI 300 to execute this extraction process, the CPU of the intermediation server 100A trains the generative AI 300 to extract other users using, for example, the following data items and their corresponding judgment conditions. However, the data items and judgment conditions are not limited to these. In addition, any of the following data items may be judged individually, or multiple of the following data items may be judged in combination.

- Total number of sent and received chat messages: Colleagues with a total of 200 or more chat messages sent and received per month are determined to have a high professional relationship value.
- Total number of chat messages sent and received outside of working hours: Colleagues who send and receive more than 100 chat messages per month outside of working hours are determined to have a high professional relationship value.
- Total time spent in online meetings: 35 hours/month or more. Colleagues who attended the same online meetings were determined to have a high work relationship value.
- Total time spent in online meetings outside of working hours: 10 hours/month or more. Colleagues who attended the same online meetings outside of working hours are determined to have a high work relationship value.
- Total number of documents shared: Taking presentation documents as an example of documents, a colleague who shares more than 20 PowerPoint presentations per month or more than 400 slides per month is determined to have a high business relationship value. Documents can be of various types, such as documents and tables, and the judgment criteria can be set with reference to the numbers for presentation documents mentioned above.
- Total number of mentions: If the total number of mentions is 40 or more per month, colleagues who mention others will be determined to have a high working relationship value.
- Total number of emails sent and received: Colleagues with a total of 150 or more sent and received chats per month are determined to have a high professional relationship value.
- Total number of reactions (likes, emoji stamps) sent and received: Colleagues with a total of 200 or more reactions sent and received per month are determined to have a high working relationship value.

In addition, instead of making a judgment based on monthly behavioral history data as described above, the intermediation server 100A may train the generative AI 300 to make a judgment based on cumulative data, for example, as follows.
- Total number of sent and received chat messages: Extract colleagues with a cumulative total of 200 or more sent and received chat messages.
- Total time spent in online meetings: Extract colleagues who attended the same online meetings for a cumulative total of 35 hours or more.
- Total number of mentions: Extract colleagues who have made 40 or more mentions in total.

In addition, instead of selecting between high and other values, the business relationship value may be determined by a different method as described below.
- Business relationship value is judged in three stages (high, medium, low): For example, if the total number of sent and received chats is 200 or more, it is judged as high, if it is between 100 and 200, it is judged as medium, and if it is 100 or less, it is judged as low. Of course, it may be judged in more stages such as A, B, C, etc.
- Business relationship value is determined by score: It is normalized to a value of 1 per sent or received chat, so that if there are 200 chats, the value is 200, and the higher the number, the higher the business relationship value.

Depending on the learning status of the generative AI 300, a new learning process for extracting colleagues with high business relationship values may be added, or the above-mentioned learning process may be deleted, and the presence or absence of a learning process may be determined as appropriate depending on the situation.

After performing the above-described learning, the generation AI 300 generates information about colleagues with high business relationship values and transmits it to the intermediation server 100A (step 66).

When the CPU of the intermediation server 100A receives information about colleagues with high business relationship values from the generation AI 300 (step 67), it sends a request to the intermediation server 100B to extract colleagues with low emotional scores from among the colleagues with high business relationship values (step 68).

When the intermediation server 100B receives the extraction request (step 69), it sends an extraction processing request to the generation AI 300 (step 70).

The generation AI 300, which receives the extraction process request, extracts from among colleagues with high business relationship values those colleagues for whom the user has determined that the emotional scores of messages sent to the colleague are low (step 71).

In order for the generative AI 300 to execute the extraction process, the CPU 11 of the intermediation server 100B trains the generative AI 300 to extract colleagues using, for example, the following handling and judgment conditions. However, the conditions are not limited to these.

- Scope of emotional scores: The emotional scores are based on messages sent in the last three months, but can also go back further, such as six months or a year.
- Mathematical processing of emotional scores: One way to do this is to use statistical values such as the average, maximum, and minimum emotional scores for each recipient of messages sent so far, but other values obtained by mathematical processing can also be used.
- Emotional score determination (1): An average score of 70 or above is determined as high, between 40 and 70 as medium, and below 40 as low.
- Emotional score determination (2): A maximum value of 85 points or more is determined as high, 55 points or more but less than 85 points is determined as medium, and less than 55 points is determined as low.
- Emotional score determination (3): A minimum value of 55 points or more is determined as high, between 25 points and 55 points as medium, and less than 25 points as low.
- If no messages have been sent, the emotional score will be determined to be 0.
・The emotional score is adjusted taking into account the number of messages sent: A coefficient can be applied depending on the number of messages sent. For example, if the number of messages sent exceeds 10 per month or 100 in total, a coefficient of 1 is applied, and if it is less than that, a coefficient of 0.8 is applied.

In this embodiment, the intermediation server 100B trains the generative AI 300 to return the emotional scores of colleagues extracted using the above method, but instead, it may return the judgment results of the emotional scores (high, low, etc.).

As for the timing of calculating the emotional score, the system may automatically score the message when it is sent and store the score in the emotional score information database 33, so that calculations for all messages are completed in advance, or the system may calculate only the target message each time a request to extract the emotional score is received.

Depending on the learning status of the generative AI 300, a new learning process for extracting colleagues with low emotional scores may be added, or the above-mentioned learning process may be deleted. The presence or absence of a learning process may be determined as appropriate depending on the situation.

After performing the above-described learning, the generative AI 300 extracts colleagues with low emotional scores from among colleagues with high business relationship values, and transmits information on the extracted colleagues and their emotional scores to the intermediation server 100B as recommendation information for targets to whom messages should be sent (step 72).

The intermediary server 100B that receives the target person recommendation information transfers the target person recommendation information to the intermediary server 100A (step 73), and the intermediary server 100A transmits the target person recommendation information to the user terminal 200 (step 74), and displays it in the recipient list display field 72 (Figures 7 and 8) on the My Page of the user terminal 200 (step 75).

Figure 7 shows the state before the sending opportunity question button 71 is pressed (when there are no extraction results), but if the colleague extraction process has been performed in the past, target recommendation information as the previous extraction result may be displayed in the sending target list display field 72.

As shown in FIG. 8, the target recommendation information includes each target's name (handle name), department, business relationship value (high, medium, low), and emotional score (statistics), and a message creation button 73 for creating a message for that target is displayed, for example, on the right side of each target's information.

When the user notices that the emotional score of a particular subject among those displayed in the list is particularly low and presses the message creation button 73 for that subject, the above-mentioned thank you/praise message creation screen is superimposed on the My Page, as shown in FIG. 9. Thereafter, as described above, the user terminal 200 creates a thank you/praise message by accepting input from the user in the message input field 51 (step 76), and when the advice request button 52 is pressed, an advice request is sent and advice information is returned from the generative AI 300 via the intermediary server 100B.

In addition to the business relationship value in the candidate recommendation information above, information on what specific work support was provided (e.g., help with creating materials, etc.) may be displayed, and in addition to the emotional score, information on what the specific state of lack of gratitude and praise is (e.g., whether a small number of messages are sent, whether careless messages are sent, etc.) may be displayed.

[summary]
As described above, according to this embodiment, the message creation support system allows a user to evaluate and understand before sending a message whether the message created by the user properly expresses feelings of gratitude or praise to the recipient user, and allows the user to send a high-quality message. In addition, the system can induce the user to have a starting point for sending a message to other users who have a strong business relationship with the user but who are unable to properly convey feelings of gratitude or praise in their messages.

Second Embodiment
Next, a second embodiment of the present invention will be described. In this embodiment and the following embodiments, the same components as those in the first embodiment are denoted by the same reference numerals, and the description thereof will be omitted or simplified.

Figure 10 is a sequence diagram showing the flow of the process of extracting recipients of messages by the message creation support system according to this embodiment. Also, Figures 11 and 12 are diagrams showing examples of user interfaces of message recipient recommendation information that the intermediation servers 100A and 100B according to this embodiment display on the user terminal 200.

In the first embodiment described above, the user asks the generative AI 300 about opportunities to send a message via the sending opportunity question button 71 provided on the personal page. In this embodiment, the user asks the chatbot about opportunities to send a message on the chat screen, and the colleague extraction results are displayed in chat format. The user can also ask questions about the basis for selecting a colleague on the chat screen and receive a response from the generative AI 300.

As shown in FIG. 10, the processing of steps 61 to 71 is the same as that in FIG. 6 above. When the generative AI 300 extracts colleagues with low emotional scores from among colleagues with high business relationship values, in addition to the extracted recommendation information for recipients, if the user asks for the reason for determining that the business relationship value is high and the emotional score is low, the generative AI 300 also transmits information on the reason to the intermediary server 100B (step 72').

The intermediary server 100B receives the recipient information and the basis information and transfers it to the intermediary server 100A (step 73'), which then transmits it to the user terminal 200 (step 74') and displays it on the chat screen (step 75').

In order for the generative AI 300 to execute this process, when the CPU 11 of the intermediary server 100B is asked by the user about the basis for determining that a colleague has a high work relationship value and a low emotional score, the generative AI 300 is trained to respond via a chatbot with the information that serves as the basis (extraction logic).

As shown in FIG. 11, when a user inputs a question on the chat screen to ask the generative AI 300 which colleague to send a thank you/praise message to, the content of the question is displayed in the question display bubble 81a. Then, by the processing of steps 72' to 74' above, the message sending target recommendation information generated by the generative AI 300 is displayed in the reply display bubble 82a.

Furthermore, as shown in FIG. 12, when a user inputs a question asking the basis for determining that one of the colleagues included in the message sending target recommendation information has a high business relationship value and a low emotional score, the question is displayed in a question display bubble 81b, and the reply to the question is displayed in an answer display bubble 82b. Specifically, the logic behind the determination that the business relationship value is high is the reply that the number of non-business chats sent and received and the number of attendances at the same online conferences are high, and the logic behind the determination that the emotional score is low is that the message does not mention the specific actions of the recipient, and the recipient is not aware of the recipient's politeness.

In this way, according to this embodiment, the message creation support system can allow the user to intuitively understand, via the chatbot, the objective reasons why they should send a message of thanks or praise to another colleague.

Third Embodiment
Next, a third embodiment of the present invention will be described.

Figure 13 is a sequence diagram showing the flow of the process of extracting recipients of messages by the message creation support system according to this embodiment. Also, Figures 14 and 15 are diagrams showing examples of user interfaces of message recipient recommendation information that the intermediation servers 100A and 100B according to this embodiment display on the user terminal 200.

In each of the above-mentioned embodiments, colleagues with high business relationship values were extracted, and from among them, colleagues with low emotional scores were further extracted, and finally, message sending target recommendation information was presented to the user. However, in this embodiment, the user can select colleagues for which emotional scores are to be calculated from among colleagues with high business relationship values.

As shown in FIG. 13, steps 81 to 87 are similar to the processing in steps 61 to 67 in FIG. 6 of the first embodiment described above.

When the intermediation server 100A receives information about a colleague with a high business relationship value from the generation system AI 300 via the intermediation server 100B (step 87), it transmits the information to the user terminal 200 as transmission candidate information (step 88).

When the user terminal 200 receives the candidate information, it displays it in the recipient list display field 72 of the My Page (step 89). As shown in FIG. 14, in the recipient list display field 72, the emotional score field in the recipient recommendation information of the above embodiment is displayed blank, and an emotional score request button 74 is provided in addition to a message creation button 73.

When a user presses the emotional score request button 74 for any candidate (selecting the candidate for whom the emotional score is to be calculated), the user terminal 200 sends a request to extract the emotional score of the selected candidate to the intermediary server 100B (step 90).

When the CPU 11 of the intermediary server 100B receives a request to extract the emotional score of the selected candidate (step 91), it refers to the emotional score information database 33, extracts the statistical value of the emotional score of the selected candidate (step 92), and transmits the extracted emotional score information to the user terminal 200 (step 93).

When the user terminal 200 receives the emotional score information, it displays the emotional score in the corresponding candidate's column in the recipient list display column 72 on the My Page, as shown in FIG. 15 (step 94). In the example shown in the figure, "TAKE SUZUKI" is selected from the candidates, and his emotional score (average 32 points) is displayed.

When the user presses the message creation button 73 for the colleague from whom the emotional score was extracted, the above-mentioned thank you/praise message creation screen is superimposed on the My Page, as shown in the figure. Thereafter, as described above, the user terminal 200 creates a thank you/praise message by accepting input from the user in the message input field 51 (step 95), and when the advice request button 52 is pressed, an advice request is sent and advice information is returned from the generative AI 300 via the intermediary server 100B.

In this manner, in this embodiment, the message creation support system allows the user to select a candidate (in whom the user is interested) from among candidates to whom a message should be sent (colleagues with a high business relationship value), and allows the user to understand the emotional scores of messages sent to that candidate, thereby impressing a stronger impression on the user of the need to send a message to that candidate and encouraging the user to create a message.

Fourth Embodiment
Next, a fourth embodiment of the present invention will be described.

In each of the above-described embodiments, an example has been shown in which the intermediation server 100A and the intermediation server 100B exist as separate servers, and the intermediation server 100A executes the process of extracting colleagues with high business relationship values, and the intermediation server 100B executes the process of extracting colleagues with low emotional scores. However, the processes of both intermediation servers may be executed by a single server.

Figure 16 is a sequence diagram showing the flow of the process of extracting recipients of messages by the message creation support system according to this embodiment. The intermediary server 100C shown in the figure is a server that has a business tool management function like the intermediary server 100A, and also has an emotional communication tool management function like the intermediary server 100B.

As shown in the figure, in this embodiment, the processes of intermediary server 100A and intermediary server 100B shown in FIG. 6 of the first embodiment are handled by intermediary server 100C alone. In this way, even a single server can execute the extraction process in the first embodiment (and second embodiment) described above.

Fifth Embodiment
Next, a fifth embodiment of the present invention will be described.

Figure 17 is a sequence diagram showing the flow of the process of extracting recipients of messages by the message creation support system according to this embodiment. The intermediary server 100C shown in the figure is a server that has a management function for business communication tools like the intermediary server 100A, and also has a management function for emotional communication tools like the intermediary server 100B.

As shown in the figure, in this embodiment, the processes of intermediary server 100A and intermediary server 100B shown in FIG. 13 of the third embodiment are handled by intermediary server 100C alone. In this way, even a single server can execute the extraction process in the third embodiment described above.

<Modification>
The present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit and scope of the present invention.

In each of the above-described embodiments, an emotional score is calculated for expressions of gratitude and praise contained in a message, and advice information is generated, but emotions are not limited to gratitude and praise. For example, an emotional score may be calculated and advice information may be generated for other positive emotions toward other users, such as approval (mainly from a superior to a subordinate) and sympathy.

In each of the above-described embodiments, the recipient of the thank/praise message is identified by the process of extracting the recipient, and after the user creates the message, as described above in FIG. 4 to FIG. 6, an example is shown in which advice information is displayed on the user terminal 200 in response to a request for advice from the user in an unsent state, and the user sends the message in response to the advice. However, after creating a message, the user terminal 200 may send the message to the destination without going through the advice request. In this case, after the message is sent, the processes of steps 44 to 52 in FIG. 4 are executed in response to a user's evaluation request or automatically, and advice information (evaluation information) including the emotional score may be presented to the user after the fact.

In each of the above-described embodiments, after a message addressed to a recipient extracted by the process of extracting recipients of a thank-you/praise message is created, advice information on whether the expression of thanks/praise contained in the message is appropriate is generated and presented to the user. However, the generation and presentation of this advice information is not essential, and the process may be completed only by the process of extracting recipients of a thank-you/praise message (and presenting information on the extracted recipients to the user). That is, the process shown in FIG. 4 and FIG. 5 may be omitted, and only the process shown in FIG. 6 to FIG. 17 may be executed. In addition, after a colleague with a high business relationship value and a low emotional score is extracted and presented as a recipient, it is not essential for the user to actually create a message addressed to that recipient. Conversely, only the process of generating and presenting advice information shown in FIG. 4 and FIG. 5 may be executed, and the process of extracting and presenting message recipients shown in FIG. 6 to FIG. 17 may be omitted.

In each of the above-described embodiments, generative AI is used as the machine learning model, but the present invention can also be realized using other machine learning models.

Of the inventions described in the claims of this application, the invention described as an "information processing method" has each step performed automatically by at least one device such as a computer through software-based information processing, and is not performed by a human using a device such as a computer. In other words, the "information processing method" is an information processing method using computer software, and is not a method in which a human operates a computing tool such as a computer.

11...CPU
18: Memory unit 19: Communication unit 31: User information database 32: Advice request right information database 33: Emotional score information database 51: Message creation field 52: Advice request button 53: Advice display field 54: Send button 71: Sending opportunity question button 72: Sending target person list display field 73: Message creation button 74: Emotional score request button 81: Question display bubble 82: Reply display bubble 100A: Intermediate server 100B: Intermediate server 100C: Intermediate server 200: User terminal 300: Generative AI

Claims (4)

A message that has been created by a user using a business communication tool on a sender terminal and has not yet been sent to a recipient is input to a learning model that has been machine-learned to learn information regarding appropriate expressions of positive emotions toward a recipient that are included in the message, and
receiving, from the learning model, advisory information regarding the appropriate expression in the input message;
transmitting the advice information to the sender terminal and displaying it;
An information processing system including a control unit,
A storage unit is further provided for storing user information on a user of the business communication tool and advice request right information on a right of the user to request the advice information from the learning model in association with each other,
The control unit is
setting a timing for granting the advice request right information for each user so that a predetermined time interval or a message sending action is required between one advice request and the next advice request, so that the number of requests for the advice information from the user in a predetermined period is within a predetermined number in the entire organization to which the user belongs;
When an advice request requesting the advice information is received from a business communication tool used by the user, the information processing system determines whether or not the right exists based on the stored advice request right information, and if it determines that the right exists, requests the advice information from the learning model along with inputting the message. A message that has been created by a user using a business communication tool on a sender terminal and has not yet been sent to a recipient is input to a learning model that has been machine-learned to learn information regarding appropriate expressions of positive emotions toward a recipient that are included in the message, and
receiving, from the learning model, advisory information regarding the appropriate expression in the input message;
transmitting the advice information to the sender terminal and displaying it;
An information processing system including a control unit,
The control unit is
training the learning model to calculate an evaluation value regarding an expression of emotion in the input message based on a predetermined index regarding an expression of gratitude or praise included in the message;
receiving evaluation value information indicating the evaluation value calculated for the input message from the learning model, and transmitting the evaluation value information to the sender terminal for display;
calculating a statistical value by collecting evaluation values in the advice information created so far for each user to whom the message is to be sent, training the learning model so as to extract other users having low statistical values, and receiving information on the other users having low statistical values from the learning model;
and transmitting, to the sender terminal, recommendation information that recommends another user having a low statistical value as a new destination of the message. A message that has been created by a user using a business communication tool on a sender terminal and has not yet been sent to a recipient is input to a learning model that has been machine-learned to learn information regarding appropriate expressions of positive emotions toward a recipient that are included in the message, and
receiving, from the learning model, advisory information regarding the appropriate expression in the input message;
Transmitting the advice information to the sender terminal and displaying it;
storing user information on a user of the business communication tool and advice request right information on a right of the user to request the advice information from the learning model in association with each other;
setting a timing for granting the advice request right information for each user so that a predetermined time interval or a message sending action is required between one advice request and the next advice request, so that the number of requests for the advice information from the user in a predetermined period is within a predetermined number in the entire organization to which the user belongs;
An information processing method, when an advice request requesting the advice information is received from a business communication tool used by the user, determining whether or not the right exists based on the stored advice request right information, and if it is determined that the right exists, requesting the advice information from the learning model along with inputting the message. In the information processing device,
A step of inputting a message that has been created by a user using a business communication tool at a sender terminal and has not yet been sent to a recipient into a learning model that has been machine-learned to learn information regarding appropriate expressions of positive feelings toward a recipient contained in the message;
receiving, from the learning model, advisory information regarding the appropriate expression in the input message;
transmitting the advice information to the sender terminal and displaying it;
storing user information on a user of the business communication tool and advice request right information on a right of the user to request the advice information from the learning model in association with each other;
setting a timing for granting the advice request right information for each user so that a predetermined time interval or a message sending action is required between one advice request and the next advice request, so that the number of requests for advice information from the user in a predetermined period of time is within a predetermined number in the entire organization to which the user belongs;
When an advice request requesting the advice information is received from a business communication tool used by the user, the program executes a step of determining whether or not the right exists based on the stored advice request right information, and if it is determined that the right exists, requesting the advice information from the learning model along with inputting the message.

Images