Schedule adjustment device, schedule adjustment method, and schedule adjustment program

Abstract

To provide a schedule adjustment device, a schedule adjustment method, and a schedule adjustment program capable of autonomously collecting job knowledge and information necessary for core incidental business that is difficult to be standardized, and making and carrying out a plan for carrying out the business.SOLUTION: A schedule adjustment unit requests a large-scale language model management unit to estimate the possibility of schedule adjustment which is based on instruction information, adjustment manual information, and participant schedule data from a person requesting the schedule adjustment, and when the possibility of the schedule adjustment is unknown as a result of estimation made by the large-scale language model management unit, causes the large-scale language model management unit to prepare a confirmation text to be transmitted to a participant for whom the possibility of schedule adjustment is unknown and determine the possibility of schedule adjustment on the basis of an answer to the confirmation text from the participant, and the schedule adjustment unit determines schedule candidates.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a schedule adjustment device, a schedule adjustment method, and a schedule adjustment program that can support schedule adjustment for an event in which multiple people participate. [Background technology]

[0002] To improve business efficiency and productivity, the automation of business processes has been promoted through the introduction of advanced business systems and robotic process automation (RPA). One example of such automation is a schedule adjustment support device described in Patent Document 1, which supports schedule adjustment for events attended by multiple people. This device acquires schedule conditions for users attending the event from a client, acquires schedule files for users and equipment from a schedule management server, and then refers to the file specifications of the schedule management server to analyze the dates and times set in the schedule files. It then extracts time slots in which equipment is not reserved from the available times shared by all users as candidate dates and times, and transmits the candidate dates and times to the client. This allows the device to support schedule adjustment for events attended by multiple people, even if the participants use different schedule management servers. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2009-244953 [Patent Document 2] Japanese Patent Application Publication No. 2023-77370 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the schedule adjustment support device described in Patent Document 1 has a problem in that it cannot set a schedule if there is no free time common to all of the target users. For example, when setting a schedule for an urgent meeting, schedules must be adjusted for a limited period of time in the near future, and it may be difficult to find free time common to all of the target users. In such cases, without using the schedule adjustment support device, a schedule adjustment officer must make adjustments by inquiring with the relevant users, such as shifting the date of an already scheduled event, or determining whether attendance is mandatory and setting the target event for the scheduled time if it is not mandatory, which requires very cumbersome work.

[0005] In contrast to this, the schedule adjustment device described in Patent Document 2 makes it possible to include exclusion keywords such as "cannot attend" in existing schedules in advance, and by using such exclusion keywords, it becomes easy to adjust schedules.

[0006] However, because the above-mentioned exclusion keywords are set without considering the priority or content of the planned or scheduled event, the decision on whether or not to exclude a keyword in relation to the event may not be made appropriately. If schedule adjustments are made based on such exclusion keywords, appropriate adjustments may not be made, and there may be cases where multiple users are unable to reserve free time. In such cases, as described above, the schedule adjustment staff will have to perform cumbersome tasks such as inquiring about the relevant users.

[0007] Furthermore, it is necessary to set exclusion keywords, and it is not possible to judge words that are not set as exclusion keywords, which makes it difficult to accommodate the diverse schedule inputs made by users.

[0008] Therefore, the present invention aims to provide a schedule adjustment device, a schedule adjustment method, and a schedule adjustment program that can autonomously communicate with core-ancillary tasks (non-routine tasks) that require human judgment and are difficult to standardize, such as adjusting the schedules of all meeting participants taking into consideration agenda items, priorities, related tasks, etc., thereby collecting necessary business knowledge and information and formulating and executing plans for the execution of the tasks. Here, core-ancillary tasks refer to non-routine tasks that are not core, do not generate profits, and are closely linked to core tasks, making them difficult to separate from core tasks and outsource (outsource).

[0009] The present invention also aims to provide a schedule adjustment device, a schedule adjustment method, and a schedule adjustment program that support schedule adjustment for an event in which multiple people participate, and that can autonomously adjust the schedule of a participant whose date cannot be adjusted due to circumstances such as having other plans, regarding the setting of the schedule for the event.

[0010] Furthermore, the schedule adjustment method and schedule adjustment program of the present invention aim to estimate whether a schedule is adjustable by reading adjustment manual information that describes business knowledge and know-how related to adjustments.

[0011] Another object of the present invention is to provide a schedule adjustment method and a schedule adjustment program that can autonomously discover new know-how that is not recorded in the adjustment manual information during the process of adjusting a schedule and autonomously record this new know-how, thereby improving the accuracy of the adjustment function using the adjustment manual information. [Means for solving the problem]

[0012] In order to solve the above problems, the schedule adjustment device of the present invention is a schedule adjustment device that adjusts the schedules of multiple participants taking part in an event, and comprises a manual management unit, a large-scale language model management unit that manages and controls a large-scale language model, and a schedule adjustment unit, wherein the manual management unit manages adjustment manual information including know-how explanations that serve as a basis for determining schedule adjustment, and the schedule adjustment unit requests the large-scale language model management unit to estimate whether or not schedule adjustment is possible based on instruction information from a person requesting schedule adjustment, the adjustment manual information, and the schedule data of the participants, and when the result of the estimation by the large-scale language model management unit indicates that schedule adjustment is unclear, or when schedule adjustment is possible, the schedule adjustment unit causes the large-scale language model management unit to create a confirmation message to be sent to the participants whose schedule adjustment is unclear, and determines whether or not schedule adjustment is possible based on the participants' responses to the confirmation message, and the schedule adjustment unit determines schedule candidates based on the result of this determination.

[0013] The schedule adjustment device of the present invention further includes a conversation management unit that manages conversations with the client, and when the large-scale language model determines that a response from a participant contains adjustment know-how that is not included in the know-how explanation, the conversation management unit requests the large-scale language model management unit to create a new know-how explanation and causes the manual management unit to add the created new know-how explanation to the adjustment manual information.

[0014] The schedule adjustment method of the present invention is a schedule adjustment method using a schedule adjustment device that has a large-scale language model management unit that has a large-scale language model and a manual management unit that manages adjustment manual information, and is characterized by comprising the steps of: a step in which a person requesting schedule adjustment for an event inputs instruction information regarding the schedule adjustment; a step in which the large-scale language model management unit estimates whether or not the schedule can be adjusted based on the instruction information, the adjustment manual information, and the schedule data of the event participants; a step in which, when the result of the estimation by the large-scale language model management unit indicates that it is unclear whether or not the schedule can be adjusted, the large-scale language model management unit creates a confirmation message to be sent to the participants whose availability of schedule adjustment is unknown; a step in which the large-scale language model management unit determines whether or not the schedule can be adjusted based on the participants' responses to the confirmation message; and a step in which a schedule candidate is determined based on the determination result in the step of determining whether or not the schedule can be adjusted.

[0015] A schedule adjustment program of the present invention is a schedule adjustment program that operates in cooperation with a schedule adjustment device that includes a large-scale language model management unit that includes a large-scale language model, a manual management unit that manages adjustment manual information, and a conversation management unit, and includes the following steps: a step in which a requester of event schedule adjustment inputs instruction information related to the schedule adjustment to the conversation management unit; a step in which the large-scale language model management unit estimates whether or not the schedule can be adjusted based on the instruction information, the adjustment manual information, and schedule data of event participants stored in an external schedule management system; a step in which, when the result of the estimation by the large-scale language model management unit indicates that the schedule can be adjusted, the large-scale language model management unit creates a confirmation message to be sent to participants whose schedule adjustment is uncertain; a step in which the large-scale language model management unit determines whether or not the schedule can be adjusted based on responses from the participants to the confirmation message; and a step in which the conversation management unit determines schedule candidates based on the determination result in the step of determining whether or not the schedule can be adjusted. It is characterized by: [Effects of the Invention]

[0016] The schedule adjustment device, schedule adjustment method, and schedule adjustment program of the present invention are capable of autonomously collecting the necessary business knowledge and information, and formulating and executing plans for core ancillary tasks (non-routine tasks) that are difficult to standardize because they require communication and human judgment. [Brief explanation of the drawings]

[0017] [Figure 1] 1 is a block diagram showing a configuration of a schedule adjustment device according to an embodiment of the present invention. [Figure 2] 10 is a flowchart illustrating a procedure for estimating whether adjustment is possible in an embodiment of the present invention. [Figure 3A] 10 is a flowchart showing a procedure for recording adjustment know-how in an embodiment of the present invention. [Figure 3B] 10 is a flowchart showing a procedure for recording adjustment know-how in an embodiment of the present invention. [Figure 4] FIG. 2 is a sequence diagram showing the mutual processing relationships of the components of the schedule adjustment device according to the embodiment of the present invention. [Figure 5] FIG. 5 is a sequence diagram showing the mutual processing relationships among the components of the schedule adjustment device according to the embodiment of the present invention, and is a sequence diagram showing processing following that of FIG. 4. [Figure 6] FIG. 10 is a diagram showing a conversation history in an example of estimating whether adjustment is possible. [Figure 7] FIG. 10 is a diagram illustrating an example of an adjustment feasibility estimation template according to an embodiment of the present invention. [Figure 8] FIG. 8 is a diagram showing an example of input to the large-scale language model shown in FIG. 7. [Figure 9] FIG. 10 is a diagram illustrating an example of a template for checking the presence or absence of an adjustable basis according to an embodiment of the present invention. [Figure 10] FIG. 10 is a diagram showing an example of input to the large-scale language model shown in FIG. [Figure 11]FIG. 11 is a diagram showing an example of a conversational input for confirming the basis for the decision made in response to the input shown in FIG. 10. [Figure 12A] FIG. 10 is a diagram showing an example of a participating member schedule information table of a requester and participants in an example of estimating whether adjustment is possible. [Figure 12B] FIG. 10 is a diagram illustrating an example of a schedule condition table in an example of estimating whether adjustment is possible; [Figure 13A] FIG. 10 is a diagram showing an example of a schedule candidate table determined by schedule confirmation, showing a state before confirmation with participants is performed. [Figure 13B] FIG. 10 is a diagram showing an example of a schedule candidate table determined by schedule confirmation, showing the final state after confirmation. [Figure 14] FIG. 10 is a diagram illustrating an example of a template for checking the presence or absence of new know-how in the embodiment of the present invention. [Figure 15] FIG. 15 is a diagram showing an example of input to the large-scale language model shown in FIG. DETAILED DESCRIPTION OF THE INVENTION

[0018] A schedule adjustment device, a schedule adjustment method, and a schedule adjustment program according to an embodiment of the present invention will be described in detail below with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a schedule adjustment device 100 according to this embodiment. The schedule adjustment method is executed by the schedule adjustment device 100 shown in FIG. 1 in accordance with the procedures shown in FIGS. 2 to 5. The schedule adjustment program is also executed in accordance with the procedures shown in FIGS. 2 to 5 by the operation and control of each component included in the schedule adjustment device 100 shown in FIG. 1.

[0019] In this embodiment, a case will be described in which the event to be adjusted in the schedule is a meeting, but the event in the present invention is not limited to the meetings described below, and may be, for example, an event, a business trip, a trip, or an event in which multiple participants participate on the adjusted date and time. When adjusting the schedule for an event other than a meeting, adjustment manual information 161 (sometimes referred to as the "adjustment manual" in the figures, etc.) is used, which records know-how corresponding to the event.

[0020] As shown in FIG. 1, the schedule adjustment device 100 includes a conversation management unit 110, a large-scale language model management unit 120, a schedule adjustment unit 130, a proxy access unit 140, an adjustment communication access unit 150, and a manual management unit 160.

[0021] The schedule adjustment device 100 can be connected to an external schedule management system 200 via a proxy access unit 140. The schedule management system 200 stores schedule data of the participant M, and this schedule information can be read and written by the schedule adjustment unit 130 via the proxy access unit 140.

[0022] In this embodiment, participant M is assumed to be a person who plans to attend the conference and who belongs to the same company as requester C, and the estimation of whether adjustment is possible is performed on the assumption that participant M and requester C belong to the same company. On the other hand, in the present invention, the relationship between requester C and participant M can also be such that they belong to a group or organization different from the company. In this case, the assumption of the estimation of whether adjustment is possible can be changed to correspond to the group or the like to which they belong.

[0023] Here, the requester C is one of the participants M, and schedule adjustment is made for the requester C and at least one participant M other than the requester C. Alternatively, a person who is not included in the participants M may be selected as the requester C. In this case, there are multiple participants M, and schedule adjustment is made for these multiple participants M.

[0024] The conversation management unit 110 includes a request receiving unit 111 and a candidate presentation unit 112, and by cooperating with the large-scale language model management unit 120 via various interfaces, can, for example, act as a coordination agent to have a text conversation with a client C and accept a request for schedule setting. The request receiving unit 111 can receive a schedule adjustment request from the client C for adjusting the schedule for an event, and can have a conversation with the client C regarding the request. The request receiving unit 111 transmits an input sentence from the client C as request information to the large-scale language model management unit 120.

[0025] The conversation management unit 110 is configured to be able to exchange information with the client C via an interface. For example, the conversation management unit 110 can be connected to the client C's personal computer or mobile terminal by a dedicated application via a connector, by a browser via web input / output, or by a cloud system via Rest API, and information can be sent and received between them.

[0026] The candidate presentation unit 112 presents candidates adjusted by the schedule adjustment unit 130 to the client C as candidate presentation conversations. When the client C explains new adjustment know-how that is not included in the adjustment manual information 161, the candidate presentation unit 112 requests the large-scale language model 121 to create a know-how explanation (sometimes abbreviated as "know-how explanation" in drawings, etc.). Furthermore, when it is unclear whether adjustment is possible or not, or when adjustment is not possible, and new adjustment know-how is included in the response from the participant M to a confirmation message sent to the participant M, the candidate presentation unit 112 requests the large-scale language model 121 to create a know-how explanation. The large-scale language model 121 generates a know-how explanation from the candidate presentation conversation stored in the conversation history storage unit 123 and returns it to the candidate presentation unit 112. The candidate presentation unit 112 sends the know-how explanation to the manual management unit 160 and requests that the know-how explanation be added to the adjustment manual information 161.

[0027] The large-scale language model management unit 120 includes a large-scale language model 121, a conversation instruction information table 122, and a conversation history storage unit 123, and manages and controls these. By cooperating with the large-scale language model management unit 120, the conversation management unit 110 can have a text conversation with a client C, can receive a schedule setting request from the client C, and can present schedule adjustment candidates to the client C in natural language text.

[0028] The large-scale language model 121 is a computer language model composed of an artificial neural network with many parameters, such as a deep learning model pre-trained with a large corpus. With this model, when a question is given, an answer can be obtained immediately based on machine-learned data, and the accuracy of the answer can be improved by fine-tuning, etc.

[0029] The conversation instruction information table 122 is a table that stores conversation instruction information templates to be input to the large-scale language model 121, and in this embodiment, it stores only the adjustment feasibility estimation template exemplified in the right diagram of Fig. 7, the template for checking the presence of adjustment basis exemplified in the right diagram of Fig. 9, and the template for checking the presence of new know-how exemplified in the right diagram of Fig. 14. The conversation instruction information includes information such as the subject of the event, the content of the event, information about participant M of the event to be adjusted, and candidate dates and times for the schedule.

[0030] The adjustment feasibility estimation template illustrated in the right diagram of Fig. 7 is composed of instructions A11 for the large-scale language model 121, a know-how explanation A12 about an adjustable schedule read from the adjustment manual information 161, and an input and output template A13. The template for checking the presence of an adjustment feasibility basis illustrated in the right diagram of Fig. 9 is composed of instructions A21 for the large-scale language model 121, and an input and output template A22. The template for checking the presence of new know-how illustrated in the right diagram of Fig. 14 is composed of instructions A31 for the large-scale language model 121, a know-how explanation A32 about an adjustable schedule read from the adjustment manual information 161, and an input and output template A33.

[0031] The conversation history storage unit 123 stores the history of conversations with the requester C and the participant M via the conversation management unit 110. The conversation history is stored as conversations between the conversation management unit 110 as a coordination agent and the requester C and the participant M, as shown in FIG. 6, for example, and includes data entered by the requester C and the participant M, conversations related to schedule adjustment, and conversations presenting candidates.

[0032] The schedule adjustment unit 130 includes a schedule operation unit 131 , a participating member schedule information table 132 , an adjustment possibility estimation unit 133 , an adjustment possibility confirmation unit 134 , a schedule condition table 135 , and a schedule candidate table 136 .

[0033] The schedule operation unit 131 requests the schedule management system 200 for the participants' schedule information via the proxy access unit 140, records the schedule information in the participant member schedule information table 132, and determines candidate dates and times related to the judgment results of the confirmation message as schedule candidates.

[0034] 12A, the participant schedule information table 132 stores schedule information received from the schedule management system 200 via the proxy access unit 140. This schedule information includes the names of the organizer (requester C) and the participant M, the schedule date and time (for example, the start time and end time), and whether or not there is an appointment for the candidate date and time. If there is an appointment, the information also includes the start time, end time, meeting subject, information about the meeting participants, information about whether the meeting is an internal meeting or not and other meeting contents, and, if external participants are included, the names of the companies to which the participants belong.

[0035] The adjustment possibility estimation unit 133 requests adjustment manual information 161 from the manual management unit 160, requests the large-scale language model management unit 120 to estimate whether adjustment is possible based on the adjustment manual information 161, the conditions in the schedule condition table 135, and the schedule information of each participant stored in the participant schedule information table 132, and acquires the adjustment possibility information. In addition, in response to a request from the schedule operation unit 131, the adjustment possibility estimation unit 133 creates and records a list of date and time candidates for which scheduling is possible.

[0036] When the estimation result from the large-scale language model management unit 120 in response to the estimation request made by the adjustment feasibility estimation unit 133 indicates that a participant M is adjustable or that the adjustment feasibility is unknown, the adjustment feasibility confirmation unit 134 requests the large-scale language model management unit 120 to create a confirmation message regarding the adjustment feasibility to be sent to that participant M. This confirmation message is created as natural language text, and upon receiving it from the large-scale language model management unit 120, the adjustment feasibility confirmation unit 134 causes the confirmation message to be sent to the relevant participant M via the adjustment communication access unit 150, using the email service 151 or the chat service 152. The request to the large-scale language model management unit 120 to create the confirmation message is made autonomously or automatically by the adjustment feasibility confirmation unit 134, without any instructions from the requester C or the participant M.

[0037] The adjustment possibility confirmation unit 134 further receives a response from the participant M to the confirmation message, and requests the large-scale language model management unit 120 to determine whether or not adjustment is possible based on the content of the response. The adjustment possibility confirmation unit 134 also makes the request to determine the content of the response autonomously or automatically, without instructions from the requester C or the participant M.

[0038] As shown in FIG. 12B, the schedule condition table 135 records candidate date data and other schedule setting information, such as the date and time of the event, the names of the organizer and participants, the event subject (the meeting subject in the case of a meeting), and the content of the event (the meeting content in the case of a meeting).

[0039] As shown in FIG. 13A , the schedule candidate table 136 stores data on candidate dates for the schedule received from the schedule condition table 135 and the estimation results estimated by the adjustment feasibility estimation unit 133 for each candidate date, in association with each other. The schedule candidate table 136 records the adjustment order, the date and time of the event, the names of the organizer and participants, the candidate type, the people to be adjusted, and the confidence level. In this embodiment, the candidate type is one of “adjustable,” “unadjustable,” and “unclear whether it can be adjusted.” The adjustment order is determined by the adjustment feasibility estimation unit 133 based on the candidate type, the number of people to be adjusted, whether the people to be adjusted include anyone other than the organizer and participants, and so on. The adjustment feasibility for participant M is confirmed in descending order of the adjustment order, and the confirmation process ends when adjustment is possible. The confidence level is an index representing the accuracy of inference by a machine learning model such as a large-scale language model, and represents the probability that the inference result calculated based on the learned data is correct. The confidence level, also known as the confidence level, is expressed as a number between 0 and 1, with the closer to 1, the higher the probability of correctness.

[0040] The proxy access unit 140 is an interface that enables the schedule adjustment unit 130 and the schedule management system 200 to be connected to each other.

[0041] The coordination communication access unit 150 is an interface that enables information to be exchanged with the event participants M. This information can be exchanged in text form using an email service 151 and a chat service 152, or it can be exchanged in voice form by installing a voice conversion service (not shown).

[0042] The manual management unit 160 includes adjustment manual information 161 and is connected to the conversation management unit 110 and the schedule adjustment unit 130 so as to be able to exchange information with each other. The adjustment manual information 161 is a document that summarizes know-how in a conventional method of adjusting schedules between a requester C and a participant M through the exchange of natural language text by means of email, chat, or the like, or through oral conversation, and is read into the large-scale language model 121. The know-how explanations recorded in the adjustment manual information 161 include rules, customs, and practices related to scheduling meetings within the company to which the requester C and the participant M belong. It is assumed that the large-scale language model 121 stores common sense and etiquette for a working adult related to scheduling meetings.

[0043] In adjusting a schedule, if a client C explains new adjustment know-how that is not included in the adjustment manual information 161, the conversation management unit 110 requests the large-scale language model 121 to create an explanation of the know-how. The large-scale language model 121 generates a know-how explanation from a candidate presentation conversation stored in the conversation history storage unit 123 and returns it to the candidate presentation unit 112. When the know-how explanation is returned, the candidate presentation unit 112 sends the know-how explanation to the manual management unit 160 and requests that the sent know-how explanation be added to the adjustment manual information 161. The manual management unit 160 adds the know-how explanation to the adjustment manual information 161 based on the request.

[0044] Next, the procedure for schedule adjustment will be described with reference to Figs. 2 to 15. Fig. 2 is a flowchart showing the procedure for estimating whether adjustment is possible, and Figs. 3A and 3B are flowcharts showing the procedure from estimating adjustment know-how to recording, with "3B" in Fig. 3A being connected to "3B" in Fig. 3B. Fig. 4 is a sequence diagram showing the mutual processing relationships between the various parts of the schedule adjustment device 100, and Fig. 5 is a sequence diagram showing processing following Fig. 4. Figs. 6 to 13 are diagrams showing an application example of the estimation of whether adjustment is possible shown in Fig. 2, and Figs. 14 to 15 are diagrams showing examples of estimating and recording adjustment know-how shown in Figs. 3A and 3B.

[0045] In this embodiment, first, an adjustment possibility estimation process is performed, examples of which are shown in FIGS. 2 and 4 (mainly S106 to S115) and in FIGS. 6 to 13. If the result of this estimation is "adjustable" or "adjustability unknown," a confirmation message is sent to participant M, and adjustment possibility is confirmed based on the content of the response to this confirmation message (S116 to S125 in FIGS. 4 and 5). If, during this process, an explanation of the reason for adjustment possibility is received in a conversation with requester C, adjustment know-how estimation processes shown in FIGS. 3A, 3B, 5 (S126 to S130), 14, and 15 are performed, and if new know-how not included in adjustment manual information 161 is found, it is added to adjustment manual information 161 as a know-how explanation. Furthermore, a schedule is selected based on a conversation with requester C regarding the schedule candidates determined based on the adjustment possibility estimation result, and the reservation is executed (S131 to S145 in FIG. 5).

[0046] <Estimation of Adjustment Possibility> As shown in FIG. 4, a request for schedule adjustment is input from a requester C via the Web or chat (step S100 in FIG. 4). The input data is sent to the conversation management unit 110, and a conversation for the schedule adjustment request is held (steps S101 and S102 in FIG. 4). The conversation management unit 110 extracts meeting conditions from the conversation for the schedule adjustment request and stores them as schedule conditions in the schedule condition table 135. In this embodiment, it is assumed that the "scheduled start time," "meeting duration," "organizer," "participants," and "meeting subject" shown in FIG. 12B are stored in the schedule condition table 135. Data input by the requester C ("user input" in FIG. 4) is sent from the request receiving unit 111 to the large-scale language model management unit 120. Instruction information is created by embedding various information such as the schedule condition table into each template stored in the conversation instruction information table 122. The instruction information is input to the large-scale language model 121. The conversation with the requester C is stored in the conversation history storage unit 123 (step S103 in FIG. 4).

[0047] Next, the large-scale language model management unit 120, which has received the input from the requester C, inputs the input sentence of the requester C received from the request receiving unit 111 into the large-scale language model 121, and returns conversation instruction information, in which the adjustment possibility estimation template shown in FIG. 7 has become the state shown in FIG. 8, to the request receiving unit 111 (step S104 in FIG. 4). The request receiving unit 111 returns the returned conversation instruction information to the requester C via the interface (step S105 in FIG. 4). By repeating these steps S100 to S105, a conversation regarding the request reception is established. P1 to P4 in FIG. 6 are shown as an example of the conversation in steps S101 to S104 in FIG. 4.

[0048] Next, the schedule operation unit 131 requests the schedule information of the participants from the proxy access unit 140 based on the conditions registered in the schedule condition table 135 (step S106 in FIG. 4, step S1 in FIG. 2). The proxy access unit 140 acquires the schedule information of the participants from the schedule management system 200 and returns it to the schedule operation unit 131. The schedule operation unit 131 records the returned schedule information in the participant member schedule information table 132 (steps S107 to S109 in FIG. 4). In this embodiment, the schedule operation unit 131 requests the schedules of requester C's "Ichiro Sato" and participant M's "Jiro Tanaka" using the scheduled date "October 18th" as a key, and acquires all of the schedules of requester C and participant M scheduled for that day, "October 18th," from the schedule management system 200.

[0049] Next, the adjustment possibility estimation unit 133 creates a list of candidate dates and times for which scheduling is possible from the schedule information recorded in the participant schedule information table 132. The created list is recorded in the schedule candidate table 136. The participants' common free time can be considered as time for which scheduling is possible. If there are no candidates for which scheduling is possible, it estimates whether or not a date and time for which some event is scheduled in participant M's schedule is adjustable. The adjustment possibility estimation unit 133 works in cooperation with the large-scale language model management unit 120 and the manual management unit 160 to estimate whether adjustment is possible (steps S110 to S115 in FIG. 4, steps S2 to S5 in FIG. 2).

[0050] The adjustment feasibility estimation unit 133 makes the determination by acquiring adjustment manual information 161 managed by the manual management unit 160, and transmitting it to the large-scale language model management unit 120 together with schedule information of the participating members for whom the adjustment feasibility determination is to be made (participating member schedule information table in FIG. 12A) and instruction information, such as schedule conditions, obtained from a conversation with the client C (steps S111 to S113 in FIG. 4, steps S2 to S4 in FIG. 2). Here, the conversation instruction information is schedule conditions created from input data from the client C. The large-scale language model management unit 120 obtains an adjustment possibility estimation template for confirming whether adjustment is possible from the conversation instruction information table 122, and inputs the adjustment possibility estimation template with the adjustment manual information 161 embedded therein, the schedule conditions, and the schedule information for which an adjustment possibility determination is to be made into the large-scale language model 121, thereby obtaining a response of either "adjustable," "unadjustable," or "adjustability unknown" as the adjustment possibility for each input schedule (step S114 in FIG. 4, steps S6 to S8 in FIG. 2).

[0051] In this embodiment, the instruction A11 and input and output templates A13 shown in FIG. 7 are stored in the conversation instruction information table 122 as an adjustment possibility estimation template, and this adjustment possibility estimation template includes adjustment manual information 161 between the instructions A11. Depending on the embodiment, the input and output templates A13 may be stored in a separate storage unit, or multiple templates may be prepared. By providing the adjustment manual information 161, in which know-how explanations are recorded, to the large-scale language model 121 each time estimation is performed, it is possible to obtain individually optimized responses. Furthermore, the input and output templates A13 can improve the accuracy of output values by providing a demo within the prompt to learn the context.

[0052] 8 shows an attempt to obtain an adjustment feasibility estimation result by providing the conference schedule, conference subject, etc. of participant M, whose adjustment feasibility is to be estimated, as input values to the adjustment feasibility estimation template of FIG. 7, to the large-scale language model 121. The large-scale language model 121 outputs, as output values (not shown), whether adjustment is possible and the confidence level of the estimation. Then, the adjustment possibility estimation unit 133 records a list of date and time candidates for which a schedule can be set in the schedule candidate table 136 based on the obtained adjustment possibility information.

[0053] <Check whether adjustment is possible> As a result of the estimation of adjustment feasibility, if there are candidates for which the estimation result is adjustable and the adjustment feasibility is unclear and which require confirmation, the adjustment feasibility confirmation unit 134 performs confirmation (step S116 in FIG. 4). The adjustment feasibility confirmation unit 134 first sends the schedule candidates for which confirmation is required to the large-scale language model 121 and requests the creation of a confirmation draft (step S117 in FIG. 4). The large-scale language model 121 creates a confirmation message such as "Mr. Tanaka, Mr. Ichiro Sato would like to have a one-hour meeting between 12:00 and 14:00 on October 18th. Mr. Tanaka already has plans, is this possible?" (see P5 in FIG. 6) and returns the created draft to the adjustment feasibility confirmation unit 134 (step S118 in FIG. 4). The adjustment possibility confirmation unit 134 sends the received confirmation draft as a confirmation message to the participant M of the schedule that needs to be confirmed via the adjustment communication access unit 150 and the email service 151 or chat service 152 (steps S119 and S120 in FIG. 4).

[0054] The adjustment possibility confirmation unit 134 receives a reply from the schedule participant M who has received the confirmation draft (steps S121 to S123 in FIG. 4, P6 in FIG. 6) via the adjustment communication access unit 150. In order to determine whether the reply content is adjustment possible or not, the adjustment possibility confirmation unit 134 adds the reply content "Yes, the meeting starting at 1:00 pm can be adjusted" (see P6 in FIG. 6) to the input field for the conversation instruction information as shown in FIG. 10 and sends it to the large-scale language model 121 to request a determination on whether the adjustment is possible (step S124 in FIG. 5). The large-scale language model 121 returns the determination result indicating whether the adjustment is possible or not to the adjustment possibility confirmation unit 134 (step S125 in FIG. 5).

[0055] If the determination result indicates that the basis for adjustment feasibility requires confirmation, the adjustment feasibility confirmation unit 134 again performs confirmation (step S116 in FIG. 4). The adjustment feasibility confirmation unit 134 requests the large-scale language model 121 to create a confirmation draft (step S117 in FIG. 4). The large-scale language model 121 creates conversation instruction information by adding a confirmation statement regarding the basis for confirmation, "Why was it determined that the meeting starting at 1:00 p.m. can be adjusted?" (see P7 in FIG. 6), to the template for checking the presence or absence of adjustment feasibility shown in FIG. 9, as shown in the input field in FIG. 11, and returns the created draft to the adjustment feasibility confirmation unit 134 (step S118 in FIG. 4). The adjustment feasibility confirmation unit 134 sends the received draft confirmation to the participants of the schedule requiring confirmation via the email service 151 or chat service 152 via the adjustment communication access unit 150 (steps S119 and S120 in FIG. 4).

[0056] The adjustment possibility confirmation unit 134 receives a reply from schedule participant M who has received the confirmation draft (steps S121 to S123 in FIG. 4, P6 in FIG. 6) via the adjustment communication access unit 150, and in order to determine whether the content of the reply is a basis for adjustment, adds the reply content "The review is a meeting for internal members only, so adjustment is possible." (see P8 in FIG. 6) to the output field of the conversation instruction information as shown in FIG. 11, sends it to the large-scale language model 121, and requests a determination on adjustment possibility (step S124 in FIG. 5). The large-scale language model 121 returns the determination result, ie, whether adjustment is possible or not, to the adjustment possibility confirmation unit 134 (step S125 in FIG. 5).

[0057] <Check for new know-how> In the confirmation of whether adjustment is possible (step S116 in FIG. 4 to step S125 in FIG. 5), if participant M explains new adjustment know-how that is not included in the adjustment manual information 161, the candidate presentation unit 112 requests the large-scale language model 121 to create an explanation of the know-how (steps S126 and S127 in FIG. 5, step S11 in FIG. 3).

[0058] Upon receiving the request, the large-scale language model 121 acquires the candidate presentation conversation, for example, P8 in FIG. 6, stored in the conversation history storage unit 123 (step S12 in FIG. 3A), and inserts the adjustment manual information 161 into the template for checking the presence of new know-how shown in FIG. 14, and reads the schedule candidate table. The candidate type information in the read schedule candidate table is used when checking the possibility of the existence of new know-how (step S16 in FIG. 3B).

[0059] After reading the schedule candidate table, the large-scale language model 121 adds the know-how explanation "The review is an internal meeting, so it can be adjusted" corresponding to the acquired conversation "The review is an internal meeting, so it can be adjusted" to the input field, as shown in FIG. 15, and replies to the candidate presentation unit 112 as conversation instruction information including a know-how draft (step S128 in FIG. 5, steps S12 to S15 in FIG. 3A).

[0060] The candidate presentation unit 112 checks whether there is a possibility of new know-how based on the received conversation instruction information (FIG. 15) (step S16 in FIG. 3B).

[0061] If a meeting that was estimated to be adjustable is confirmed to be either adjustable or not adjustable, it is determined that there is no possibility of new know-how existing ("None" in step S16 of FIG. 3B), and the process of checking for the existence of new know-how is terminated without adding a know-how explanation to the adjustment manual information 161 (step S17 of FIG. 3B).

[0062] When the determination in step S16 in FIG. 3B is "yes," for example, when a meeting that was estimated to be unreliable is later determined to be adjustable, it is determined that there is a possibility that new know-how exists ("yes" in step S16 in FIG. 3B), and the candidate presentation unit 112 determines whether new know-how exists in the conversation history data (step S18 in FIG. 3B). As a result, when it is determined that new know-how exists, the candidate presentation unit 112 requests the large-scale language model 121 to create a know-how explanation (steps S126 and S127 in FIG. 5), sends the created know-how explanation to the manual management unit 160, and requests that the know-how explanation be added to the adjustment manual information 161. Based on the request, the manual management unit 160 adds the know-how explanation to the adjustment manual information 161, thereby acquiring new adjustment know-how (step S130 in FIG. 5, step S19 in FIG. 3B).

[0063] If it is determined that no new know-how exists as a result of checking the existence of new know-how in the conversation history data (step S18 in FIG. 3B), the know-how explanation A32 in the template for checking the existence of new know-how shown in FIG. 14 is referenced to check whether known know-how exists (step S20 in FIG. 3B). If known know-how exists ("Yes" in step S20 in FIG. 3B), the know-how explanation is not added to the adjustment manual information 161, and the process of checking the existence of new know-how ends (step S17 in FIG. 3B).

[0064] If there is no known know-how ("None" in step S20 of FIG. 3B), a confirmation message is sent to participant M to ask why adjustment is possible (step S21 of FIG. 3B, step S120 of FIG. 4). Then, it is determined whether new know-how exists in the answer (answer in the conversation history) (step S123 of FIG. 4) (step S22 of FIG. 3B, step S124 of FIG. 5).

[0065] In this judgment (step S22 in FIG. 3B), if it is determined that no new know-how exists, the process of checking whether or not new know-how exists is terminated (step S17 in FIG. 3B) without adding a know-how explanation to the adjustment manual information 161; if it is determined that new know-how exists, the large-scale language model 121 is requested to create a know-how explanation (steps S126 and S127 in FIG. 5), and the manual management unit 160 is requested to add the created know-how explanation to the adjustment manual information 161; based on the request, the manual management unit 160 adds the know-how explanation to the adjustment manual information 161 and acquires new adjustment know-how (step S130 in FIG. 5, step S19 in FIG. 3B).

[0066] <Select schedule and make reservation> Upon receiving the result of the determination as to whether or not adjustment is possible from the large-scale language model 121 (step S125 in FIG. 5), the adjustment possibility confirmation unit 134 updates the list of date and time candidates for which scheduling is possible on the schedule candidate table 136 (step S131 in FIG. 5).

[0067] Next, the candidate presentation unit 112 in the conversation management unit 110 converses with the requester C based on the participating member schedule information table 132 determined by the schedule adjustment unit 130, and presents schedule candidates (steps S132 to S134 in FIG. 5, P7 in FIG. 6). The conversation in the candidate presentation unit 112 is carried out in cooperation with the large-scale language model management unit 120, similar to the conversation of request acceptance by the request acceptance unit 111 (steps S135 to S140 in FIG. 5).

[0068] When client C selects a desired schedule candidate during the conversation with client C (steps S135 to S137 in FIG. 5, P8 in FIG. 6), the selection is sent to the large-scale language model management unit 120 via the conversation management unit 110 (step S138 in FIG. 5), and the large-scale language model management unit 120 returns a response to the effect that the request has been accepted to client C's web chat or the like via the conversation management unit 110 (steps S139 and S140 in FIG. 5), and requests the schedule adjustment unit 130 to register the specified schedule (steps S141 and S142 in FIG. 5).

[0069] Upon receiving the request, the schedule adjustment unit 130 instructs the proxy access unit 140 to register the schedule via the schedule operation unit 131 (steps S143 and S144 in FIG. 5). Upon receiving the instruction, the proxy access unit 140 registers the schedule in the schedule management system 200 (step S145 in FIG. 5). After completing the registration in the schedule management system 200, the schedule operation unit 131 notifies the large-scale language model 121 that the schedule has been registered, and the large-scale language model 121 notifies the requester C via the conversation management unit 110 that the schedule has been registered.

[0070] Next, a specific example of estimation of whether adjustment is possible will be described with reference to FIGS. In this example, the conversation between Sato Ichiro as the client C and the conversation management unit 110 as the adjustment agent, as shown in Fig. 6, is recorded in the conversation history storage unit 123 as a conversation history for each conversation phase. In this conversation, the schedule adjustment device 100 operates based on the conversation between Sato Ichiro and the conversation management unit 110, thereby performing the processes shown in Figs. 7 to 13, and the meeting desired by Sato Ichiro with Tanaka Jiro (participant M) from 13:00 to 14:00 on October 18th is arranged. The specific process will be explained below.

[0071] As shown in Fig. 7, the conversation instruction information table 122 stores templates for recording the adjustment feasibility judgment and the acquisition and recording of know-how for each conversation phase. The schedule adjustment unit 130 requests the large-scale language model 121 to estimate the adjustment feasibility by providing the adjustment feasibility estimation template shown in Fig. 7 (see step S114 in Fig. 4). Here, the adjustment feasibility estimation template (Fig. 7) is a template for the large-scale language model 121 that is prepared in advance to apply information indicated in the candidate presentation conversation with the requester C.

[0072] In the "adjustment know-how insertion space" (A12) in FIG. 7, an explanation of know-how that can be adjusted is input from the adjustment manual information 161.

[0073] The adjustment possibility estimation template shows an input and output template A13. As shown in this template, the large-scale language model 121 presents a self-diagnosed confidence level for the accuracy of the estimation based on the results of structural analysis of the sentence indicating participant M's schedule and the results of comparative analysis with the know-how explanation text. The confidence level is expressed as a number from 0 or above, indicating a low confidence level, to 1 or below, indicating the highest confidence level. In the example shown in FIG. 7, the words "in-house" and "(tentative)", which are highly relevant to the know-how explanation text, are included, so the numbers are "0.7" and "0.9", respectively.

[0074] 12A shows the participating member schedule information table 132 in which the schedule for 10 / 18 (October 18th) is recorded, with Sato Ichiro and Tanaka Jiro as organizers or participants. This table is data returned from the schedule management system 200 in response to a request from the schedule adjustment unit 130.

[0075] FIG. 12B shows the schedule conditions (step S105 in FIG. 4) set for the request reception phase, such as "Please arrange a meeting with Tanaka Jiro on October 18th" (P1 in FIG. 6) and "The agenda will be a weekly report from Company C. Please arrange a one-hour meeting" (P3 in FIG. 6), spoken by Sato Ichiro, and recorded in the schedule condition table 135.

[0076] First, the schedule adjustment unit 130 refers to the participating member schedule information table 132 (FIG. 12A) and determines that Sato Ichiro's free time is from 12:00 to 14:00.

[0077] Next, the schedule adjustment unit 130 sends a request to the large-scale language model management unit 120 to estimate whether adjustment is possible for the four meetings (meeting IDs #100, #101, #102, and #103) that Ichiro Sato is hosting or attending and the meetings (meeting IDs #104 and #105) that correspond to Ichiro Sato's free time in the participating member schedule information table 132 (FIG. 12A) (step S114 in FIG. 4). The large-scale language model management unit 120 estimates whether adjustment is possible based on the know-how explanations "undecided, tentative schedule" and "schedule with no external participants" (know-how explanation A12 in FIG. 7). Of the six meetings, for meeting ID #104, the schedule adjustment unit 130 replies "adjustable" because it is "not decided, tentative," for meeting ID #105, the schedule adjustment unit 130 replies "unclear whether it can be adjusted" because it may be "a schedule without external participants," and for the other four meeting IDs #100, 101, 102, and 103, the schedule adjustment unit 130 replies "not adjustable" because there may be external participants (step S115 in Figure 4).

[0078] Based on the above schedule confirmation (P4 in FIG. 6), of the two candidates shown in schedule candidate table 136 in FIG. 13A, confirmation is made with Tanaka Jiro about the candidates (meeting IDs #104 and #105) for which schedule adjustment is possible, i.e., the candidate for which it was unclear whether the schedule could be adjusted and the candidate for which the schedule adjustment was possible (steps P5 to P8 in FIG. 6). As a result, it was confirmed that meeting ID #105 was adjustable, and so it is determined to be a candidate, as shown in FIG. 13B. This schedule is determined to be "reservable" and has a confidence level of 1.0.

[0079] Next, for ID#105, which has the highest adjustment ranking, the schedule adjustment unit 130 sends a confirmation message to Tanaka Jiro regarding whether or not adjustment is possible via the email service 151 or chat service 152 via the adjustment communication access unit 150 (step S119 in FIG. 4, P5 in FIG. 6). This confirmation message is created at the request of the schedule adjustment unit 130 to the large-scale language model management unit 120 (steps S117 and S118 in FIG. 4).

[0080] Upon receiving the confirmation message, Tanaka Jiro replies that the appointment from 13:00 to 14:00 (meeting ID #105) is an "internal regular meeting" and therefore can be adjusted (P6 to P8 in FIG. 6, step S123 in FIG. 4). The adjustment feasibility confirmation unit 134 adds the content of the conversation (steps S5 and S6 in FIG. 6) to the input field for conversation instruction information as shown in FIG. 10 for the template for checking the presence of adjustment feasibility basis shown in FIG. 9, and sends it to the large-scale language model 121 to request a determination on whether adjustment is possible (step S124 in FIG. 5). The large-scale language model 121 returns a determination result indicating whether adjustment is possible or not to the adjustment feasibility confirmation unit 134 (step S125 in FIG. 5).

[0081] The adjustment feasibility estimation unit 133 determines that the basis for adjustment feasibility in the reply from Tanaka Jiro needs to be confirmed, and requests the large-scale language model 121 to create a draft confirmation statement (steps S116 and S117 in FIG. 4). The large-scale language model 121 creates conversation instruction information by adding a confirmation statement regarding the basis for confirmation, "Why was it determined that the meeting starting at 1:00 p.m. can be adjusted?", to the template for confirming the presence or absence of adjustment feasibility shown in FIG. 9, as shown in the input field in FIG. 11, and returns the created draft confirmation statement to the adjustment feasibility confirmation unit 134 (step S118 in FIG. 4). The adjustment feasibility confirmation unit 134 sends the received draft confirmation statement to Tanaka Jiro, the participant who needs to be confirmed, via the adjustment communication access unit 150, by email service 151 or chat service 152 (steps S119 and S120 in FIG. 4, P7 in FIG. 6).

[0082] The adjustment possibility confirmation unit 134 receives a reply from Tanaka Jiro after receiving the draft confirmation (steps S121 to S123 in FIG. 4, P8 in FIG. 6) via the adjustment communication access unit 150. In order to determine whether the content of the reply is a basis for whether adjustment is possible, the adjustment possibility confirmation unit 134 adds the reply content "The review is a meeting only for internal members, so adjustment is possible" to the output field of the conversation instruction information as shown in FIG. 11 and sends it to the large-scale language model 121 to request a determination on whether adjustment is possible (step S124 in FIG. 5). The large-scale language model 121 returns the determination result that adjustment is possible to the adjustment possibility confirmation unit 134 (step S125 in FIG. 5).

[0083] The schedule adjustment unit 130, having received the determination result from the large-scale language model 121, presents the schedule candidate to the conversation management unit 110, assuming that the schedule candidate has been determined (step S132 in FIG. 5), and the conversation management unit 110 then transmits the schedule candidate to Sato Ichiro (step S134 in FIG. 5, P9 in FIG. 6).

[0084] Having received the proposed schedule, Ichiro Sato replies to the conversation management unit 110 that he accepts the proposed schedule, "October 18th, 13:00-14:00" (steps S135 and S136 in FIG. 5, P10 in FIG. 6). Upon receiving this reply, the conversation management unit 110 transmits the schedule selected by Ichiro Sato, "October 18th, 13:00-14:00," to the schedule adjustment unit 130, which then requests the schedule management system 200 to reserve the schedule via the proxy access unit 140, and the reservation is made in the schedule management system 200 (steps S142-S145 in FIG. 5).

[0085] Next, a specific example of estimation and acquisition of adjustment know-how will be described with reference to FIGS. 14, a draft know-how explanation created in the large-scale language model management unit 120 is added based on the explanation received from Tanaka Jiro as participant M during adjustment (step S123 in FIG. 4, P8 in FIG. 6) (step S129 in FIG. 5). The conversation management unit 110 transmits the conversation instruction information shown in FIG. 11 to the manual management unit 160, and the written draft know-how explanation is added to the adjustment manual information 161.

[0086] As configured as above, the above embodiment provides the following effects. For core ancillary tasks (non-routine tasks) that require communication, human judgment, and are difficult to standardize, in cases where human judgment is required, communication can be carried out autonomously without seeking instructions from the client, making it possible to automatically collect the business knowledge and information necessary for judgment, and enabling the required work execution plans to be drawn up and executed quickly while ensuring a certain level of reliability.

[0087] In the above embodiments, a schedule adjustment device, a schedule adjustment method, and a schedule adjustment program that adjust schedules for meetings, events, business trips, travel, and other events have been described. However, as another aspect in which a specified process is performed using a manual management unit and a large-scale language model management unit, this can also be embodied as a core-ancillary business processing device, a core-ancillary business processing method, and a core-ancillary business processing program that enable automation or labor-saving of core-ancillary tasks.

[0088] More specifically, the core-ancillary task processing device comprises a manual management unit that manages adjustment manual information including know-how explanations that serve as the basis for processing core-ancillary tasks, a large-scale language model management unit that manages and controls large-scale language models, and a core-ancillary task management unit that manages and executes core-ancillary tasks, and the core-ancillary task management unit requests the large-scale language model management unit to create processing procedures to be executed by the core-ancillary task management unit based on instruction information from the client and the adjustment manual information, and the core-ancillary task management unit executes the core-ancillary tasks in accordance with the created processing procedures. Therefore, since the core-ancillary tasks are executed automatically according to the client's instructions, labor savings can be achieved by reducing the number of personnel and costs involved in core-ancillary task processing.

[0089] Furthermore, if the large-scale language model management unit is unable to determine the processing procedure for the core-ancillary task to be created, the core-ancillary task management unit will have the large-scale language model management unit create a confirmation message to send to the client, send this confirmation message to the client, and have the large-scale language model management unit determine whether the processing procedure for the core-ancillary task can be determined based on the client's response to the confirmation message, and based on the result of this determination, the core-ancillary task management unit will determine the processing procedure and execute the processing. Here, instructions from the client and transmission to the client can be made through the conversation management unit, as in the above embodiment.

[0090] The know-how explanation and adjustment manual information have the same configuration as the know-how explanation and adjustment manual information of the above embodiment, and the adjustment manual information records know-how explanations related to the processing of core ancillary business. When the large-scale language model determines that there is new know-how in the client's response to the confirmation statement, as in the above embodiment, it requests the large-scale language model management unit to create a new know-how explanation and has the manual management unit add the new know-how explanation to the adjustment manual information.

[0091] For example, in the case of "overseas business trip procedures" as a core ancillary task, when the requester explains the planned business trip date, destination, and purpose of the visit, the system retrieves past business trip approvals stored in the coordination manual information and uses the Internet to search for candidate accommodations and presents them to the requester. Once the requester selects a lodging option from the candidates, the business trip approval is created. When the core ancillary task management department receives an explanation from the requester, it checks the large-scale language model based on the coordination manual information to see if there are any problems with the purpose of the visit, etc.

[0092] Furthermore, if the core ancillary task is "travel expense settlement," when the requester explains the purpose and route of the business trip and submits a photo of the receipt as evidence, the core ancillary task management department uses the coordination manual information to check against the large-scale language model whether there are any problems with the purpose of the business trip or the way the photo is taken, and if there are any problems, contacts the requester and requests that they resubmit the correct purpose of the business trip and the photo. If there are no problems, the settlement process is carried out using the travel expense settlement system.

[0093] Here, the core-attached business processing method is a method of executing core-attached business using the core-attached business processing device, and the core-attached business processing program is a program that executes core-attached business in cooperation with the core-attached business processing device.

[0094] Modifications will be described below. In the above embodiment, communication with participant M was carried out via the adjustment communication access unit 150 provided separately from the conversation management unit 110, but it is also possible to exchange information with participant M via the conversation management unit 110 without providing the adjustment communication access unit 150. This simplifies the configuration of the schedule adjustment device and reduces the cost of configuring the device. Although the present invention has been described with reference to the above-mentioned embodiment, the present invention is not limited to the above-mentioned embodiment, and improvements or modifications can be made within the scope of the invention or the spirit of the invention. [Explanation of symbols]

[0095] 100 Schedule Adjustment Device 110 Conversation Management Department 111 Request Reception Department 112 Candidate presentation section 120 Large-scale Language Model Management Unit 121 Large-scale language models 122 Conversational Instruction Information Table 123 Conversation history memory unit 130 Schedule Coordination Department 131 Schedule operation section 132 Participant Schedule Information Table 133 Adjustability estimation unit 134 Adjustability confirmation section 135 Schedule Condition Table 136 Schedule Candidate Table 140 Proxy Access Department 150 Adjustment communication access section 151 Email Service 152 Chat Service 160 Manual Management Department 161 Adjustment Manual Information 200 Schedule Management System C. Client M participants

Claims

1. A schedule adjustment device that adjusts schedules for multiple participants in an event, comprising: a manual management unit; a large-scale language model management unit that manages and controls a large-scale language model; and a schedule adjustment unit; the manual management unit manages adjustment manual information including know-how explanations that serve as a basis for determining the schedule adjustment; the schedule adjustment unit requests the large-scale language model management unit to estimate whether the schedule adjustment is possible based on instruction information from a requester of the schedule adjustment, the adjustment manual information, and schedule data of the participants; When the result of estimation by the large-scale language model management unit is that it is unclear whether the schedule can be adjusted, and when it is possible to adjust the schedule, the schedule adjustment unit causes the large-scale language model management unit to create a confirmation message to be sent to the participant for whom it is unclear whether the schedule can be adjusted, and determines whether the schedule can be adjusted based on the participant's response to the confirmation message, and the schedule adjustment unit determines a schedule candidate based on the determination result. A schedule adjustment device comprising:

2. a conversation management unit that manages a conversation with the client, When the large-scale language model determines that the response from the participant includes the adjustment know-how that is not included in the know-how explanation, the conversation management unit requests the large-scale language model management unit to create a new know-how explanation, and causes the manual management unit to add the created new know-how explanation to the adjustment manual information.

2. The schedule adjustment device according to claim 1.

3. A schedule adjustment method using a schedule adjustment device including a large-scale language model management unit that includes a large-scale language model and a manual management unit that manages adjustment manual information, a step in which a requester of an event schedule adjustment inputs instruction information regarding the schedule adjustment; the large-scale language model management unit estimating whether the schedule adjustment is possible or not based on the instruction information, the adjustment manual information, and schedule data of participants of the event; when it is unclear whether the schedule can be adjusted as a result of the estimation by the large-scale language model management unit, the large-scale language model management unit creates a confirmation message to be sent to the participant whose schedule can be adjusted is unclear; a step in which the large-scale language model management unit determines whether or not the schedule adjustment is possible based on the responses from the participants to the confirmation message; and determining a schedule candidate based on a result of the determination in the step of determining whether or not the schedule can be adjusted. A schedule adjustment method comprising:

4. A schedule adjustment program that operates in cooperation with a schedule adjustment device that includes a large-scale language model management unit that includes a large-scale language model, a manual management unit that manages adjustment manual information, and a conversation management unit, a step in which a requester of an event schedule adjustment inputs instruction information regarding the schedule adjustment to the conversation management unit; a step in which the large-scale language model management unit estimates whether the schedule adjustment is possible based on the instruction information, the adjustment manual information, and schedule data of participants of the event stored in an external schedule management system; when it is unclear whether the schedule can be adjusted as a result of the estimation by the large-scale language model management unit, the large-scale language model management unit creates a confirmation message to be sent to the participant whose schedule can be adjusted is unclear; a step in which the large-scale language model management unit determines whether or not the schedule adjustment is possible based on the responses from the participants to the confirmation message; and a step in which the conversation management unit determines a schedule candidate based on a result of the determination in the step of determining whether or not the schedule can be adjusted. A schedule adjustment program.

Images