Retraining system, retraining method and program

Abstract

To provide a retraining system for realizing a large language model useful for a business support system.SOLUTION: In a retraining system (1): a large language model acquisition unit (101) acquires a pre-trained large language model; a training data acquisition unit (102) acquires training data for retraining the large language model, the training data having been created on the basis of activity data representing activity of a user performed in a business support system that supports business of the user; a retraining unit (103) creates a business support model specific to the business support system, by executing retraining of the large language model on the basis of the training data.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present disclosure relates to a relearning system, a relearning method, and a program.

Background Art

[0002] Conventionally, a business support system for supporting a user's work is known. For example, Patent Document 1 describes groupware that supports communication between users belonging to an organization as an example of a business support system. The business support system of Patent Document 1 inputs data indicating the usage pattern of groupware by the user to a learned model that estimates whether or not a period of no input during which the user does not perform an operation occurs. The learned model estimates whether or not a period of no input occurs based on the data.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, large language models used in the field of natural language processing are different from learned models that perform simple classification such as Patent Document 1 and need to analyze a very large number of languages. For this reason, conventional technologies including Patent Document 1 have not been able to realize a large language model useful for a business support system.

[0005] One of the objects of the present disclosure is to realize a large language model useful for a business support system.

Means for Solving the Problems

[0006] The relearning system according to the present disclosure includes a large language model acquisition unit that acquires a pre-trained large language model, a training data acquisition unit that acquires training data for relearning the large language model, the training data being created based on activity data indicating the activities of the user performed in a business support system that supports the user's business, and a relearning unit that creates a business support model specific to the business support system by executing relearning of the large language model based on the training data.

Advantages of the Invention

[0007] The present disclosure can realize a large language model useful for a business support system.

Brief Description of the Drawings

[0008]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Embodiments for Carrying Out the Invention

[0009] [1. Hardware Configuration of the Relearning System] An example of an embodiment of a relearning system, a relearning method, and a program according to the present disclosure will be described. FIG. 1 is a diagram showing an example of the hardware configuration of the relearning system. For example, the relearning system 1 includes a relearning terminal 10 and a business support system 2. The business support system 2 includes a business support server 20 and a user terminal 30. Each of the relearning terminal 10, the business support server 20, and the user terminal 30 is connected to a network N such as the Internet or a LAN.

[0010] The relearning terminal 10 is a computer that executes the relearning described below. For example, the relearning terminal 10 is a personal computer, a server computer, or a tablet. For example, the relearning terminal 10 includes a control unit 11, a storage unit 12, a communication unit 13, an operation unit 14, and a display unit 15. For example, the control unit 11 includes at least one processor. The storage unit 12 includes at least one of a volatile memory such as a RAM and a non-volatile memory such as a flash memory. The communication unit 13 includes at least one of a communication interface for wired communication and a communication interface for wireless communication. The operation unit 14 is an input device such as a mouse, a touch panel, or a keyboard. The display unit 15 is a liquid crystal or organic EL display.

[0011] The business support server 20 is a server computer. For example, the business support server 20 includes a control unit 21, a storage unit 22, and a communication unit 23. The hardware configuration of each of the control unit 21, the storage unit 22, and the communication unit 23 may be the same as that of the control unit 11, the storage unit 12, and the communication unit 13, respectively.

[0012] The user terminal 30 is a user's computer. For example, the user terminal 30 is a personal computer, a tablet, a smartphone, or a wearable terminal. For example, the user terminal 30 includes a control unit 31, a storage unit 32, a communication unit 33, an operation unit 34, and a display unit 35. The hardware configuration of each of the control unit 31, the storage unit 32, the communication unit 33, the operation unit 34, and the display unit 35 may be the same as that of the control unit 11, the storage unit 12, the communication unit 13, the operation unit 14, and the display unit 15, respectively.

[0013] Note that the program stored in at least one of the storage units 12, 22, and 32 may be supplied to at least one of the relearning terminal 10, the business support server 20, and the user terminal 30 via the network N. Also, the hardware configuration of at least one of the relearning terminal 10, the business support server 20, and the user terminal 30 is not limited to the example in FIG. 1. For example, at least one of the relearning terminal 10, the business support server 20, and the user terminal 30 may include at least one of a reading unit (e.g., a memory card slot) that reads an information storage medium and an input / output unit (e.g., a USB terminal) that connects to the information storage medium. In this case, the program stored in the information storage medium may be supplied to at least one of the relearning terminal 10, the business support server 20, and the user terminal 30 via at least one of the reading unit and the input / output unit.

[0014] Also, the computer included in the relearning system 1 is not limited to the example in FIG. 1. The relearning system 1 may include at least one computer. For example, the relearning system 1 may include only the relearning terminal 10. In this case, the business support system 2 exists outside the relearning system 1. For example, the relearning system 1 may include a computer not shown in FIG. 1.

[0015] [2. Overview of the Relearning System] In this embodiment, the case where the business support system 2 is groupware is taken as an example. For example, the business support system 2 supports the business of users belonging to an organization such as a company or an administrative agency. The user uses the business support function that the business support system 2 has. The business support function is a function that a program developed for business support (e.g., a program stored in the business support server 20 or a logram stored in the user terminal 30) has.

[0016] The business support function may be various known functions. For example, the business support function may be a thread function for users to communicate in a thread, a schedule function for users to manage schedules, a database function for users to manage databases, a file management function for users to manage files, or a mail management function for users to manage e-mails. The business support function may also be other functions implemented in known groupware.

[0017] Note that the business support system 2 may be a system not classified as groupware. For example, the business support system 2 may be a system that supports operations not related to an organization, a system that does not have functions for interlocking with other products, or other systems. The business support system 2 may be any system that supports any operation of a user.

[0018] FIG. 2 is a diagram showing an example of a screen displayed on the user terminal 30. For example, when a user operates the user terminal 30 to log in to the business support system 2, the user terminal 30 causes the display unit 35 to display a business support screen SC for supporting the user's operations. The user can use any business support function from the business support screen SC. In the example of FIG. 2, the business support screen SC when the user uses the thread function is shown. The user can post any message to the thread displayed on the business support screen SC.

[0019] In addition to threads such as those in FIG. 2, various texts are registered in the business support system 2. If the business support system 2 can perform language analysis on the texts registered in itself, it is considered that the user's operations can be effectively supported. For example, if the business support system 2 can analyze the texts registered in a thread to create a summary of the thread, the user can easily understand the topic in the thread. Similarly, for other business support functions other than the thread function, language analysis of texts is considered useful. However, it is very time-consuming for the operating company of the business support system 2 to create a large language model for language analysis from scratch.

[0020] Therefore, the relearning system 1 creates a business support model specific to the business support system 2 by performing relearning of a pre-trained large language model. The pre-trained large language model may be a known model used in the field of natural language processing. For example, the large language model may be a recurrent neural network type model, a long short-term memory network type model, or a Transformer type model.

[0021] In this embodiment, the case where the large language model is a Transformer type model is taken as an example. For example, the large language model may be a GPT (Generative Pre-trained Transformer) or BERT (Bidirectional Encoder Representations from Transformers) model. PaLM (Pathways Language Model) or LLaMA (Large Language Model Meta AI), which may be classified as a Transformer type model, may be used as the large language model. The large language model may be other models using machine learning techniques.

[0022] For example, when text is input to the large language model, the large language model divides the text into a plurality of tokens. Tokens are elements for the large language model to understand the text. For example, tokens may be words, morphemes, phrases, or characters. The large language model converts each of the plurality of tokens into an embedding representation. The large language model understands the context of the text input to itself based on the order of the embedding representations of each of the plurality of tokens. The large language model performs an output according to its task based on the context of the text. The large language model executes these series of processes based on the parameters of each layer such as the intermediate layer and the output layer. The large language model may have hundreds of millions to hundreds of billions of parameters.

[0023] For example, the pre-learning of the large-scale language model is performed by another computer that is not included in the re-learning system 1. The operating company of the business support system 2 acquires the large-scale language model that has been pre-learned by another computer, either for a fee or free of charge. The large-scale language model can support the user's business to some extent through pre-learning. However, since the pre-learning of the large-scale language model is performed based on general-purpose data that is unrelated to business support, it is not possible to perform language analysis specific to business support. Therefore, the re-learning system 1 re-learns the large-scale language model based on data specific to business support.

[0024] Re-learning is a process for changing at least a part of a large-scale language model. For example, re-learning is adjustment of parameters of a large-scale language model. Re-learning may be replacement of some layers included in a large-scale language model. In this embodiment, a case where fine-tuning corresponds to re-learning is taken as an example. For example, fine-tuning is adjustment of parameters of a large-scale language model so that the parameters are adapted to a specific task. The meaning of fine-tuning may be a well-known meaning with various theories. Fine-tuning is not limited to the example of this embodiment.

[0025] Note that the re-learning is not limited to fine tuning. Re-learning may be any process for changing at least a part of the large-scale language model. For example, the re-learning may be transfer learning or distillation. There are various theories on the meaning of transfer learning or distillation, but transfer learning or distillation may have a known meaning among various theories. The re-learning may be a process other than fine tuning, transfer learning, and distillation. Hereinafter, the large-scale language model after re-learning is referred to as a business support model. The part simply describing the large-scale language model means a pre-trained large-scale language model (a large-scale language model before re-learning).

[0026] FIG. 3 is a diagram showing an example of the relationship between a large language model and a business support model. For example, pre-training of the large language model M1 is performed so that the large language model M1 can process a specific task. The task of this embodiment is a task for business support, different from general tasks. Hereinafter, the task of this embodiment will be referred to as a business support task. In the example of FIG. 3, creating a summary of a thread corresponds to the business support task.

[0027] Note that the business support task is not limited to creating a summary of a thread. For example, the business support task may be creating a summary of text other than the messages of a thread (for example, the body of an email, a comment posted by the comment function of a database, or the text of a website such as a help page), automatically creating some text, machine translation, answering a user's question, extracting data registered in the business support system 2, classifying data registered in the business support system 2, analyzing a user's behavior, predicting a user's behavior, or other tasks.

[0028] For example, assume that thread data indicating a thread posted to the business support system 2 is input to the large language model M1. The large language model M1 analyzes the text of the thread indicated by the thread data based on the parameters adjusted by pre-training. The large language model M1 outputs summary data indicating a summary of the thread based on the analysis result of the text. Since the pre-training of the large language model M1 is performed based on general-purpose data, the large language model M1 can create a summary with a certain degree of accuracy, but may not be able to create a summary specific to business support.

[0029] For example, assume that in a general summary, the beginning and the end of the text to be summarized are important. In this case, pre-training of the large language model M1 is performed so that the large language model M1 creates a summary based on the beginning and the end of the text. The large language model M1 creates a summary of the thread based on the beginning and the end of the thread indicated by the thread data input to itself.

[0030] On the other hand, in the summary for business support, assume that the part including the meeting date and the text entered by the senior user is more important than the beginning and the end of the text. In this case, if these contents are not included at the beginning and the end of the text, the summary created by the large language model M1 may not be very useful. For example, the summary created by the large language model M1 may not include the meeting date and may not include the text entered by the senior user.

[0031] Therefore, the relearning system 1 of the present embodiment executes relearning of the large language model M1 based on the activity data indicating the user's activity in the business support system 2. The activity is the user's input to the business support system 2. Since the activity data contains information useful for business support, a business support model M2 useful for business support is created by the relearning. In the example of FIG. 3, the thread data corresponds to the activity data. Examples of activity data other than the thread data will be described later.

[0032] For example, in the business support model M2, training data indicating the relationship between the thread prepared for relearning and the correct summary created from the thread is learned. In the present embodiment, the case where the operating company of the business support system 2 creates the correct summary is taken as an example. For example, training data is created so that the summary includes the messages including the meeting date and the messages of the senior user among the messages posted to the thread. By executing relearning based on such training data, the business support model M2 can create a summary specific to business support.

[0033] As described above, the relearning system 1 creates a business support model M2 specific to business support by performing relearning of a large language model M1 on which prelearning based on general-purpose data has been executed. For example, it is not realistic for the operating company of the business support system 2 to perform learning of the business support model M2 from scratch, but the relearning system 1 can realize a business support model M2 useful for the business support system 2 by relearning the large language model M1. Hereinafter, the details of the relearning system 1 will be described.

[0034] [Functions Realized by the Relearning System] FIG. 4 is a diagram showing an example of functions realized by the relearning system 1.

[0035] [3-1. Functions Realized by the Relearning Terminal] For example, the relearning terminal 10 includes a data storage unit 100, a large language model acquisition unit 101, a training data acquisition unit 102, and a relearning unit 103. The data storage unit 100 is realized by the storage unit 12. Each of the large language model acquisition unit 101, the training data acquisition unit 102, and the relearning unit 103 is realized by the control unit 11.

[0036] [Data Storage Unit] The data storage unit 100 stores data necessary for relearning. For example, the data storage unit 100 stores the large language model M1. The large language model M1 includes parameters (parameters before relearning) adjusted by prelearning and a program for processing such as calculation of embedding representations.

[0037] The parameters of the large language model M1 are referred to by the program of the large language model M1. For example, the parameters are weight coefficients and biases. The parameters may be various known parameters. The parameters are not limited to weight coefficients and biases. For example, the parameters may be a matrix referred to during calculation of embedding representations, a positional encoding referred to in encoding of token positions, or other parameters.

[0038] The program of the large language model M1 shows the internal processing of the large language model M1. For example, the program of the large language model M1 includes an encoder that calculates embedded representations, a decoder that creates outputs specific to the business support task from the embedded representations, an output layer that performs the final output, and the processing of other layers. In these processes, parameters are referenced. There may be parameters specific to a certain business support task. The program of the large language model M1 may be a known program.

[0039] For example, when the relearning by the relearning unit 103 is completed, the data storage unit 100 stores the relearned business support model M2. In this embodiment, an example is given where the large language model M1 and the relearned business support model M2 are separate data. These may not be separate data, and the large language model M1 may be overwritten by the business support model M2. The data storage unit 100 stores not only the large language model M1 and the business support model M2 but also the training database DB1.

[0040] FIG. 5 is a diagram showing an example of the training database DB1. The training database DB1 is a database in which the training data necessary for the relearning of the large language model M1 is stored. For example, the training data includes an input part that is input to the large language model M1 during relearning and an output part that is the correct answer during relearning. The combination of the input part and the output part may be any combination according to the business support task. The training data may be prepared manually or by a known tool.

[0041] The input part of the training data may be the activity data itself or other data created based on the activity data. Since the large language model M1 performs language analysis, the activity data represents the text input by the user. The activity data may represent other activities other than text. For example, the activity may represent the user's reaction (e.g., like or emoji reaction), view count, reply count, text posting date and time, or other content.

[0042] The output part of the training data indicates the content corresponding to the business support task. For example, if the business support task is the output of some text such as a summary, the output part of the training data is the correct text. If the business support task is the output of some classification (label), the output part of the training data is the correct classification. If the business support task is the extraction of some data, the output part of the training data is the correct data (the data to be extracted from the input part of the training data). If the business support task is the sorting of some data, the output part of the training data is the correct order.

[0043] In this embodiment, the case where the creation of the thread summary corresponds to the business support task is taken as an example. Therefore, in the example of FIG. 5, the input part of the training data is the thread data for re - learning. The thread data used as the input part of the training data may be the thread data actually existing in the business support system 2 or the thread data indicating the content of a virtual thread prepared by the operating company of the business support system 2. In the example of FIG. 5, the output part of the training data is summary data indicating the correct summary created from the thread data that is the input part of the training data.

[0044] Note that the training data is not limited to the examples of this embodiment. The training data only needs to show content corresponding to the business support task. The input part of the training data may be learning activity data or data created based on the learning activity data. The output part of the training data only needs to show the correct answer as the processing result for the input part of the training data. The input part and the output part of the training data only need to show content corresponding to the business support task. Hereinafter, an example of the training data corresponding to the business support task will be described, but the training data is not limited to the following example.

[0045] For example, when the business support task is to create a summary of text other than the messages of the thread (for example, the body of an email, the comments posted by the comment function of a database, or the text of a website such as a help page), the input part of the training data shows the text to be summarized. In this case, the output part of the training data shows the correct summary created from the text.

[0046] For example, when the business support task is to automatically create some text (for example, the message posted to the thread, the body of the email, the comment posted by the comment function of the database, or the text of a website such as a help page), the input part of the training data shows the conditions of the text to be created. In this case, the output part of the training data shows the correct text corresponding to the conditions.

[0047] For example, when the business support task is machine translation, the input part of the training data is the text before translation (the text in the source language). In this case, the output part of the training data is the correct text after translation. For example, when the business support task is to answer a user's question, the input part of the training data shows the question prepared for re-learning. In this case, the output part of the training data shows the correct answer to the question.

[0048] For example, when the business support task is the extraction of data registered in the business support system 2 (for example, extraction of the sender's address of an email, extraction of the recipient's address of an email, extraction of the organization and name described in the signature of an email, extraction of the organization and name from the image data of a business card), the input part of the training data indicates the data to be the extraction source. In this case, the output part of the training data indicates the correct content extracted from the data.

[0049] For example, when the business support task is the classification of data registered in the business support system 2 (for example, classification of topics on a thread, classification of the content described in an email, or classification of files registered in the business support system 2), the input part of the training data indicates the data for re-learning. In this case, the output part of the training data indicates the classification of the data. For example, when the business support task is the analysis or prediction of a user's behavior in the business support system 2, the input part of the training data indicates the user's behavior. In this case, the output part of the training data indicates the analysis result or prediction result of the behavior.

[0050] Note that the data stored in the data storage unit 100 is not limited to the above examples. The data storage unit 100 may store any data. For example, the data storage unit 100 may store a re-learning program indicating a series of processes in the re-learning of the large language model M1. When the business support model M2 is created for each user organization, the data storage unit 100 may store the business support model M2 for each organization.

[0051] [Large Language Model Acquisition Unit] The large language model acquisition unit 101 acquires a pre-trained large language model M1. In this embodiment, since the pre-trained large language model M1 is stored in the data storage unit 100, the large language model acquisition unit 101 acquires the pre-trained large language model M1 from the data storage unit 100. When the pre-trained large language model M1 is stored in a computer or information storage medium other than the re-training terminal 10, the large language model acquisition unit 101 acquires the pre-trained large language model M1 from the other computer or information storage medium.

[0052] [Training data acquisition unit] The training data acquisition unit 102 acquires training data for re-training the large language model M1, which is created based on activity data indicating the user's activities performed in the business support system 2 that supports the user's business. In this embodiment, since the training data is stored in the training database DB1, the large language model acquisition unit 101 acquires the training data from the training database DB1. When the training data is stored in a computer or information storage medium other than the re-training terminal 10, the training data acquisition unit 102 acquires the training data from the other computer or information storage medium.

[0053] [Re-training unit] The re-training unit 103 creates a business support model M2 specific to the business support system 2 by performing re-training of the large language model M1 based on the training data. In this embodiment, an example is given where the re-training unit 103 performs re-training of the large language model M1 based on the algorithm of supervised learning. The re-training algorithm may be a known algorithm. For example, the re-training unit 103 may perform re-training of the large language model M1 based on the algorithm of semi-supervised learning or unsupervised learning.

[0054] For example, the relearning unit 103 executes the relearning of the large language model M1 by adjusting the parameters of the large language model M1 such that when the input portion of the training data is input to the large language model M1, the output portion of the training data is output from the large language model M1, based on an algorithm for supervised learning. The relearning unit 103 calculates a loss indicating the error between the output from the large language model M1 when the input portion of the training data is input to the large language model M1 and the output portion of the training data, based on a known loss function. The relearning unit 103 executes the relearning of the large language model M1 by adjusting the parameters of the large language model M1 such that the calculated loss becomes smaller. The relearning unit 103 may repeatedly execute the relearning of the large language model M1 until the loss becomes less than a threshold value.

[0055] For example, the large language model M1 during relearning divides the text indicated by the input portion of the training data into a plurality of tokens. The large language model M1 encodes each of the plurality of tokens and converts them into a sequence of embedding representations. The large language model M1 predicts its continuation as necessary based on the sequence of embedding representations and calculates an output corresponding to the business support task. The greater the difference between the output of the large language model M1 and the output portion of the training data, the greater the loss. The smaller the difference between the output of the large language model M1 and the output portion of the training data, the smaller the loss. These series of processes are executed based on the program of the large language model M1 (for example, the encoder, decoder, output layer, and other layers) and the current parameters of the large language model M1.

[0056] For example, assume that the input part of the training data shows text corresponding to the activity data. Further, assume that the output part of the training data is the correct text corresponding to the said text. In this case, the business support task becomes a task of creating other text based on some text. When the relearning unit 103 inputs the input part of the training data to the large language model M1, the large language model M1 outputs other text according to the arrangement of the embedding representations of the text shown by the input part. The relearning unit 103 executes the relearning of the large language model M1 based on the difference between the said other text and the output part of the training data.

[0057] For example, assume that the input part of the training data shows text based on the activity data. Further, assume that the output part of the training data is the classification that is the correct answer for the said text. In this case, the business support task becomes a task of classifying some text. When the relearning unit 103 inputs the input part of the training data to the large language model M1, the large language model M1 outputs an estimated classification result according to the arrangement of the embedding representations of the text shown by the input part. The relearning unit 103 executes the relearning of the large language model M1 based on the difference between the estimated classification result of the said classification and the output part of the training data.

[0058] In the present embodiment, taking the case where the summary creation of the thread corresponds to the business support task as an example, the relearning unit 103 inputs the thread data, which is the input part of the training data, to the large language model M1. The large language model M1 outputs summary data showing a summary according to the arrangement of the embedding representations of the thread shown by the said thread data. The relearning unit 103 executes the relearning of the large language model M1 based on the difference between the output summary data and the output part of the training data. Similarly for the various examples of training data described above, the relearning unit 103 may input the input part of the training data to the large language model M1 and execute the relearning of the large language model M1 based on the difference between the output from the large language model M1 and the output part of the training data.

[0059] Note that the algorithm used for re - learning may be a known algorithm. For example, the re - learning unit 103 performs re - learning of the large - scale language model M1 based on the error backpropagation method. The re - learning unit 103 may perform re - learning of the large - scale language model M1 based on other algorithms other than the error backpropagation method. For example, the re - learning unit 103 may perform re - learning of the large - scale language model M1 based on the gradient descent method, the momentum method, the quasi - Newton method, the conjugate gradient method, the local search method, or other algorithms.

[0060] In this embodiment, the re - learning unit 103 executes a series of re - learning processes by executing the re - learning program stored in the data storage unit 100. The re - learning unit 103 records the large - scale language model M1 for which re - learning has been completed in the data storage unit 100 as the business support model M2. The re - learning unit 103 transmits the business support model M2 to the business support server 20. The business support model M2 transmitted to the business support server 20 is provided for user use.

[0061] [3 - 2. Functions Realized by the Business Support Server] For example, the business support server 20 includes a data storage unit 200 and a business support unit 201. The data storage unit 200 is realized by a storage unit 22. The business support unit 201 is realized by a control unit 21.

[0062] [Data Storage Unit] The data storage unit 200 stores data necessary for user business support. For example, the data storage unit 200 stores a user database DB2.

[0063] FIG. 6 is a diagram showing an example of the user database DB2. The user database DB2 is a database in which various data of users who use the business support system 2 are stored. For example, the user database DB2 stores the organization ID of the organization to which the user belongs, the user ID of the user, and activity data. Any data may be stored in the user database DB2. For example, other data such as the password for user login may be stored in the user database DB2.

[0064] The organization ID is the ID of the organization that has contracted with the business support system 2. The user ID is the ID of the user belonging to the organization. For example, the user ID is used as a login account for the user to log in to the business support system 2. At least one user ID is associated with one organization ID. The activity data is data indicating the activity of the user. For example, the activity data indicates the date and time when the activity was performed and the specific content of the activity.

[0065] In the present embodiment, since the posting of a message to a thread corresponds to an activity, the activity data indicates the text that is the specific content of the message. The activity data may include the date and time when the user posted a message to the thread, a reaction to the message, or other information. Each time a user of a certain organization posts a message to a thread, the activity data is created by the business support unit 201 and associated with the organization ID of the organization and the user ID of the user, and the activity data is stored in the user database DB2.

[0066] Note that the data stored in the data storage unit 200 is not limited to the user database DB2. The data storage unit 200 may store any data. For example, the data storage unit 200 may store the relearned business support model M2. The data storage unit 200 may store various data such as the content of threads registered in the business support system 2 separately from the activity data.

[0067] [Business Support Department] The business support department 201 executes various processes for supporting the user's business. For example, the business support department 201 creates activity data based on the operation data indicating the operation content of the user acquired from the user terminal 30. The business support department 201 updates the user database DB2 so that the user ID of the user and the created activity data are associated with each other. The business support department 201 transmits the activity data stored in the user database DB2 to the relearning terminal 10. The training data is created based on the activity data.

[0068] For example, the business support department 201 supports the user's business based on the relearned business support model M2. In this embodiment, since the case where creating a summary of a thread corresponds to a business support task is taken as an example, the business support department 201 inputs the thread data of the thread that the user wishes to create a summary for to the business support model M2. The relearned business support model M2 calculates the embedded representation of the thread indicated by the thread data. The relearned business support model M2 creates summary data indicating a summary corresponding to the calculated embedded representation. The business support department 201 transmits the summary data created by the relearned business support model M2 to the user terminal 30. The business support department 201 similarly supports the user's business for other business support tasks by inputting data indicating some text to the business support model M2 and based on the data output from the business support model M2.

[0069] [3-3. Functions Implemented on the User Terminal] For example, the user terminal 30 includes a data storage unit 300, a display control unit 301, and an operation reception unit 302. The data storage unit 300 is realized by the storage unit 32. The display control unit 301 and the operation reception unit 302 are realized by the control unit 31.

[0070] [Data storage unit] The data storage unit 300 stores data necessary for the user to use the business support system 2. For example, the data storage unit 300 stores a browser. For example, the data storage unit 300 stores a program dedicated to the business support system 2.

[0071] [Display control unit] The display control unit 301 causes the display unit 35 to display various screens for business support. For example, the display control unit 301 causes a screen in the business support system 2 to be displayed on the browser. The display control unit 301 causes a screen in the business support system 2 to be displayed on a program dedicated to the business support system 2. After the business support model M2 is created, the user can use the business support model M2 from the business support screen SC. In the example of FIG. 2, when the user selects "Create summary", summary data of the displayed thread is created by the business support model M2.

[0072] [Operation reception unit] The operation reception unit 302 receives operations on each screen displayed by the display control unit 301 on the display unit 35. The operation reception unit 302 transmits operation data indicating the operation content performed by the user to the business support server 20.

[0073] [4. Processes Executed in the Relearning System] FIG. 7 is a diagram showing an example of a process executed in the relearning system 1. The control units 11, 21, and 31 execute the programs stored in the storage units 12, 22, and 32, respectively, whereby the process of FIG. 7 is executed.

[0074] As shown in FIG. 7, the business support server 20 creates activity data based on the operation data received from the user terminal 30 and stores it in the user database DB2 (S1). The business support server 20 transmits the activity data stored in the user database DB2 to the relearning terminal 10 (S2). The relearning terminal 10 receives the activity data from the business support server 20 (S3). The relearning terminal 10 stores the training data created based on the activity data in the training database DB1 (S4).

[0075] The relearning terminal 10 acquires the pre-trained large language model M1 stored in the storage unit 12 (S5). The relearning terminal 10 acquires an arbitrary number of training data from the training database DB1 (S6). The relearning terminal 10 creates a business support model M2 by relearning the pre-trained large language model M1 based on the training data acquired in S6 (S7). The relearning terminal 10 transmits the business support model M2 created in S7 to the business support server 20 (S8). The business support server 20 receives the business support model M2 from the relearning terminal 10 (S9). The business support server 20 executes a process for allowing the user to use the business support model M2 with the user terminal 30 (S10), and this process ends.

[0076] [Summary of Embodiment] The relearning system 1 of this embodiment acquires a pre-trained large language model M1. The relearning system 1 acquires training data for relearning the large language model M1, which is created based on activity data. The relearning system 1 executes relearning of the large language model M1 based on the training data. For example, although it is not realistic for the operating company of the business support system 2 to create a business support model M2 from scratch, the relearning system 1 can realize a business support model M2 useful for the business support system 2 by relearning the pre-trained large language model M1. For example, compared with the general large language model M1, the relearning system 1 can effectively support the user's business based on the business support model M2 in which the user's activities in the business support system 2 are learned. Since the large language model M1 has basic language analysis capabilities, the relearning system 1 can create a highly accurate business support model M2 even with a small amount of training data.

[0077] [6. Modification Example] Note that the present disclosure is not limited to the examples of the embodiments. The present disclosure can be modified without departing from the spirit of the present disclosure.

[0078] [6-1. Modification Example 1] In the embodiment, the training data including the thread data for relearning and the summary data that is the correct answer during relearning is described as an example. The relearning system 1 can provide more effective business support by creating a plurality of business support models M2 based on the training data created from various perspectives, rather than executing relearning with the training data created from such a single perspective. Therefore, the relearning system 1 of Modification Example 1 executes multi-stage relearning based on the training data created from various perspectives.

[0079] The training data acquisition unit 102 of Modification Example 1 acquires a plurality of training data created from different viewpoints. A viewpoint is a criterion for creating training data. In other words, a viewpoint is the type of activity data used in creating training data. In Modification Example 1, as an example of a plurality of training data created from different viewpoints, training data created for each business support task and training data created for each organization will be described. Based on the training data created from these viewpoints, multi-stage re-learning is executed.

[0080] FIG. 8 is a diagram showing an example of multi-stage re-learning. The re-learning unit 103 of Modification Example 1 creates a plurality of business support models M2 by executing multi-stage re-learning of the large language model M1 based on the plurality of training data acquired by the training data acquisition unit 102 of Modification Example 1. For example, the re-learning unit 103 creates common business support models M2A1 and M2A2, which are the first-stage business support models M2, by executing first-stage re-learning based on the training data created from a certain viewpoint. Hereinafter, when the common business support models M2A1 and M2A2 are not distinguished, they are simply referred to as the common business support model M2A. The common business support model M2A may be only one or three or more.

[0081] For example, the re-learning unit 103 creates specific business support models M2B1 and M2B2, which are the second-stage business support models M2, by executing second-stage re-learning based on the training data created from other viewpoints. Hereinafter, when the specific business support models M2B1 and M2B2 are not distinguished, they are simply referred to as the common business support model M2B. The common business support model M2B may be only one or three or more. The re-learning unit 103 may execute re-learning of three or more stages.

[0082] In Modification 1, as in the embodiment, it is assumed that users belonging to each of a plurality of organizations use the business support system 2. Further, it is assumed that the user database DB2 stores the activity data of users of each of the plurality of organizations. For example, the training data acquisition unit 102 acquires common training data, which is common to a plurality of organizations and is created based on the activity data of each of the plurality of organizations, for each business support task performed by the business support model M2.

[0083] For example, it is assumed that there are two business support tasks: summary creation of threads and machine translation. In this case, the training data acquisition unit 102 acquires common training data created based on the thread data (an example of activity data) of each of the plurality of organizations for the creation of the common business support model M2A1 that performs summary creation of threads. For example, the training data acquisition unit 102 acquires common training data created based on the activity data indicating some text input by users belonging to each of the plurality of organizations for the creation of the common business support model M2A2 that performs machine translation.

[0084] For example, the relearning unit 103 creates a common business support model M2A, which is a common business support model common to a plurality of organizations, by performing relearning of the large language model M1 based on the common training data for each business support task. The common business support model M2A of Modification 1 is a model specific to a certain business support task but not a model specific to a particular organization. The common business support model M2A can generally support the operations of each of the plurality of organizations. The common business support model M2A is an intermediate model for creating the specific business support model M2B described later. The method of relearning based on the common training data is the same as the relearning based on the training data described in the embodiment.

[0085] For example, assume that there are two business support tasks: creating a summary of a thread and machine translation. In this case, the relearning unit 103 creates a common business support model M2A1 for creating a summary of a thread by executing relearning of the large language model M1 based on common training data created from thread data (an example of activity data) of each of a plurality of organizations. The common business support model M2A1 for creating a summary of a thread has higher accuracy in creating a summary than the large language model M1, but cannot perform summary creation specific to a particular organization.

[0086] For example, the relearning unit 103 creates a common business support model M2A2 for machine translation by executing relearning of the large language model M1 based on common training data created from activity data indicating some text input by users belonging to each of a plurality of organizations. The common business support model M2A2 for machine translation has higher accuracy in machine translation than the large language model M1, but cannot perform machine translation specific to a particular organization.

[0087] For example, the training data acquisition unit 102 acquires, for each organization, specific training data that is training data specific to the organization and is created based on the activity data of the organization. In Modification Example 1, a case where the activity data used for creating the specific training data and the activity data used for creating the common training data are the same is described, but these may be different.

[0088] For example, assume that there are two business support tasks: creating a summary of a thread and machine translation. In this case, the training data acquisition unit 102 acquires specific training data created based on the thread data of a first organization for creating a specific business support model M2B1 that performs summary creation specific to the first organization. The training data acquisition unit 102 acquires specific training data created based on the thread data of a second organization for creating a specific business support model M2B2 that performs summary creation specific to the second organization.

[0089] For example, for creating a specific business support model M2B3 that performs machine translation specific to a certain first organization, the training data acquisition unit 102 acquires specific training data created based on activity data indicating some text input by a user belonging to the first organization. The training data acquisition unit 102 acquires specific training data created based on activity data indicating some text input by a user belonging to the second organization for creating a specific business support model M2B4 that performs machine translation specific to another second organization.

[0090] For example, for each organization, the relearning unit 103 creates a specific business support model M2B, which is a specific business support model M2 specific to the organization, by performing relearning of the common business support model M2A based on the specific training data of the organization. The specific business support model M2B of Modification Example 1 is a model specialized for a specific business support task of a specific organization. The specific business support model M2B can support a specific business of a specific organization. The specific business support model M2B is the final model provided to users belonging to a specific organization. The method of relearning based on specific training data is the same as the relearning based on training data described in the embodiment.

[0091] For example, assume that there are two business support tasks: summary creation of threads and machine translation. In this case, the relearning unit 103 creates a specific business support model M2B1 for summary creation specific to the first organization by performing relearning of the common business support model M2A1 for summary creation based on specific training data created from the thread data of the first organization. The relearning unit 103 creates a specific business support model M2B2 for summary creation specific to the second organization by performing relearning of the common business support model M2A1 for summary creation based on specific training data created from the thread data of the second organization.

[0092] For example, the relearning unit 103 creates a specific business support model M2B3 for machine translation specific to the first organization by performing relearning of the common business support model M2A2 for machine translation based on specific training data created from activity data indicating some text input by a user belonging to the first organization. The relearning unit 103 creates a specific business support model M2B4 for machine translation specific to the second organization by performing relearning of the common business support model M2A2 for machine translation based on specific training data created from activity data indicating some text input by a user belonging to the second organization.

[0093] In the above description, the case where there are two business support tasks and two organizations has been described as an example, but the number of business support tasks and organizations may be arbitrary. For example, when there are three or more business support tasks, three or more common business support models M2A may be created. When there are three or more organizations, three or more specific business support models M2B may be created for each business support task.

[0094] For example, the relearning unit 103 records the specific business support model M2B of each of the plurality of organizations in the data storage unit 100. The relearning unit 103 transmits the specific business support model M2B of each of the plurality of organizations to the business support server 20. The business support server 20 records in the data storage unit 200 by associating the organization ID of each of the plurality of organizations with the specific business support model M2B of the organization. When a user belonging to a certain organization logs in to the business support system 2, the business support unit 201 supports the business of the user based on the specific business support model M2B associated with the organization ID of the organization.

[0095] In addition, when a user belonging to an organization that has not created the specific business support model M2B logs in to the business support system 2, the business support department 201 may support the user's business based on the common business support model M2A. In this case, the business support department 201 may support the user's business based on the specific business support model M2B associated with the organization ID of another organization (for example, an organization with the same industry type or employee scale), instead of the common business support model M2A.

[0096] The relearning system 1 of Modification Example 1 acquires a plurality of training data created from different perspectives. The relearning system 1 creates a plurality of business support models M2 by performing multi-stage relearning of the large language model M1 based on the plurality of training data. Thereby, the relearning system 1 can realize a more useful business support model M2. For example, the relearning system 1 can realize more accurate business support through multi-stage relearning.

[0097] In addition, the relearning system 1 creates a common business support model M2A based on common training data for each business support task. The relearning system 1 creates a specific business support model M2B for each organization based on the specific training data of that organization. Thereby, after the relearning system 1 creates the common business support model M2A based on more common training data, it can create the specific business support model M2B specialized in the tendencies of individual organizations, so that it can efficiently create the business support model M2 specific to each organization. For example, the relearning system 1 can realize highly accurate business support according to a specific business support task and a specific organization.

[0098] [6-2. Modification Example 2] For example, the multi-stage relearning method described in Modification 1 is not limited to the multi-stage relearning method according to the organization described in Modification 1. In Modification 2, as an example of another method of multi-stage relearning, multi-stage relearning according to business support tasks and business support functions will be described. When there are a plurality of business support systems 2, the business support function may not be a function of a single business support system 2, but may be the business support system 2 itself. For example, when there are a plurality of groupwares, each individual groupware may correspond to a business support function.

[0099] For example, the training data acquisition unit 102 acquires common training data, which is common to a plurality of business support functions and is created based on the activity data of each of the plurality of business support functions of the business support system 2, for each business support task performed by the business support model M2.

[0100] For example, assume that there are two business support tasks: summary creation and machine translation. The summary creation in Modification 2 is not limited to threads, but is assumed to be summary creation for the entire business support system 2. In this case, the training data acquisition unit 102 acquires common training data for summary creation, which is created based on the activity data of each of the plurality of business support functions, for creating a common business support model M2A1 that performs summary creation. For example, the training data acquisition unit 102 acquires common training data for machine translation, which is created based on the activity data of each of the plurality of business support functions, for creating a common business support model M2A2 that performs machine translation.

[0101] For example, for each business support task, the relearning unit 103 creates a common business support model M2A, which is a common business support model M2 for multiple business support functions, by performing relearning of the large language model M1 based on common training data. The common business support model M2A in Variation 2 is a model specific to a certain business support task, but not a model specific to a particular business support function. The common business support model M2A can generally support the work of users who utilize each of the multiple business support functions. The common business support model M2A is an intermediate model for creating a specific business support model M2B described later. The method of relearning based on common training data is the same as the relearning based on the training data described in the embodiment.

[0102] For example, assume that there are two business support tasks: summary creation and machine translation. In this case, the relearning unit 103 creates a common business support model M2A1 for summary creation by performing relearning of the large language model M1 based on the common training data for summary creation, which is created from the activity data of each of the multiple business support functions. The common business support model M2A1 for summary creation has higher accuracy in summary creation than the large language model M1, but it cannot perform summary creation specific to a particular business support function.

[0103] For example, the relearning unit 103 creates a common business support model M2A2 for machine translation by performing relearning of the large language model M1 based on the common training data for machine translation, which is created from the activity data of each of the multiple business support functions. The common business support model M2A2 for machine translation has higher accuracy in machine translation than the large language model M1, but it cannot perform machine translation specific to a particular business support function.

[0104] For example, for each business support function, the training data acquisition unit 102 acquires specific training data that is specific to the business support function and is created based on the activity data of the business support function. In Modification 2, similar to Modification 1, a case where the activity data used for creating the specific training data and the activity data used for creating the common training data are the same is described, but these may be different.

[0105] For example, assume that there are two business support tasks: summary creation and machine translation. Further, assume that there are two business support functions: a thread function and a mail management function. In this case, for creating a specific business support model M2B1 that performs summary creation specific to the thread function, the training data acquisition unit 102 acquires specific training data created based on thread data. For creating a specific business support model M2B2 that performs summary creation specific to the mail management function, the training data acquisition unit 102 acquires specific training data created based on activity data indicating the text of the mails managed by the mail management function.

[0106] For example, for creating a specific business support model M2B3 that performs machine translation specific to the thread function, the training data acquisition unit 102 acquires specific training data created based on thread data. For creating a specific business support model M2B4 that performs machine translation specific to the mail management function, the training data acquisition unit 102 acquires specific training data created based on activity data indicating the text of the mails managed by the mail management function.

[0107] For example, for each business support function, the relearning unit 103 creates a specific business support model M2B, which is a business support model M2 specific to the business support function, by performing relearning of the common business support model M2A based on the specific training data of the business support function. The specific business support model M2B of Modification 2 is a model specialized for a specific business support task of a specific business support function. The specific business support model M2B can support the operations of users who utilize the specific business support function. The specific business support model M2B is the final model provided to users who utilize the specific business support function. The method of relearning based on specific training data is the same as the relearning based on the training data described in the embodiment.

[0108] For example, assume that there are two business support tasks: summary creation and machine translation. Further assume that there are two business support functions: the thread function and the mail management function. In this case, the relearning unit 103 creates a specific business support model M2B1 for performing summary creation specific to the thread function by performing relearning of the common business support model M2A1 for summary creation based on the specific training data created from thread data. The relearning unit 103 creates a specific business support model M2B2 for performing summary creation specific to the mail management function by performing relearning of the common business support model M2A1 for summary creation based on the specific training data created from the activity data of users who utilize the mail management function.

[0109] For example, the relearning unit 103 creates a specific business support model M2B3 for performing machine translation specific to the thread function by performing relearning of the common business support model M2A2 for machine translation based on the specific training data created from thread data. The relearning unit 103 creates a specific business support model M2B4 for performing machine translation specific to the mail management function by performing relearning of the common business support model M2A2 for machine translation based on the specific training data created from the activity data of users who utilize the mail management function.

[0110] In the above description, the case where there are two business support tasks and two business support functions has been described as an example. However, the number of business support tasks and business support functions may be arbitrary. For example, when there are three or more business support tasks, three or more common business support models M2A may be created. When there are three or more business support functions, three or more specific business support models M2B may be created for each business support task.

[0111] For example, the relearning unit 103 records each specific business support model M2B of a plurality of business support functions in the data storage unit 100. The relearning unit 103 transmits each specific business support model M2B of a plurality of business support functions to the business support server 20. The business support server 20 records each of the plurality of business support functions and the specific business support model M2B of the business support function in the data storage unit 200 in association with each other. When a user uses a certain business support function, the business support unit 201 supports the user's business based on the specific business support model M2B associated with the business support function.

[0112] When a user uses a business support function for which the specific business support model M2B has not been created, the business support unit 201 may support the user's business based on the common business support model M2A. In this case, the business support unit 201 may support the user's business based on the specific business support model M2B associated with another business support function instead of the common business support model M2A.

[0113] The relearning system 1 of Modification Example 2 creates a common business support model M2A by performing relearning of the large language model M1 based on common training data for each business support task. The relearning system 1 creates a specific business support model M2B, which is a business support model M2 specific to the business support function, by performing relearning of the common business support model M2A based on the specific training data of the business support function for each business support function. As a result, after creating the common business support model M2A based on more common training data, the relearning system 1 can create the specific business support model M2B specialized for the tendencies of individual business support functions, so that the business support model M2 specific to individual business support functions can be created efficiently. For example, the relearning system 1 can achieve highly accurate business support according to a specific business support task and a specific business support function.

[0114] [6-3. Modification Example 3] For example, in Modification Example 3, as another example of the multi-stage relearning described in Modification Example 1, multi-stage relearning according to business support functions and organizations will be described. The business support task of Modification Example 3 may be one.

[0115] For example, the training data acquisition unit 102 acquires common training data, which is common training data for a plurality of organizations, created based on the activity data of each of the plurality of organizations for each business support function of the business support system 2.

[0116] For example, assume that there are two business support functions: a thread function and a mail management function. In this case, the training data acquisition unit 102 acquires common training data created based on the activity data indicating the text input by users belonging to each of the plurality of organizations using the thread function for creating the common business support model M2A1 for the thread function. For example, the training data acquisition unit 102 acquires common training data created based on the activity data indicating the text input by users belonging to each of the plurality of organizations using the mail management function for creating the common business support model M2A2 for the mail management function.

[0117] For example, for each business support function, the relearning unit 103 creates a common business support model M2A common to multiple organizations by performing relearning of the large language model M1 based on common training data. Although the common business support model M2A of Modification 3 is a model specialized for the business support system 2, it is not a model specific to a particular organization. The common business support model M2A can generally support the operations of each of the multiple organizations. The common business support model M2A is an intermediate model for creating the specific business support model M2B described later. The method of relearning based on the common training data is the same as the relearning based on the training data described in the embodiment.

[0118] For example, assume that there are two business support functions: a thread function and a mail management function. In this case, the relearning unit 103 creates a common business support model M2A1 for the thread function by performing relearning of the large language model M1 based on common training data created from the activity data (activity data indicating the text input by the user in the thread function) of each of the multiple organizations. The common business support model M2A1 for the thread function can provide business support more specific to the thread function than the large language model M1, but it cannot provide business support specific to a particular organization.

[0119] For example, the relearning unit 103 creates a common business support model M2A2 for the mail management function by performing relearning of the large language model M1 based on common training data created from the activity data (activity data indicating the text input by the user in the mail management function) of each of the multiple organizations. The common business support model M2A2 for the mail management function can provide business support more specific to the mail management function than the large language model M1, but it cannot provide business support specific to a particular organization.

[0120] For example, the training data acquisition unit 102 acquires, for each organization, unique training data that is unique to the organization and is created based on the activities of the organization. In Modification 3, a case where the activity data used for creating the unique training data and the activity data used for creating the common training data are the same is described, but they may be different.

[0121] For example, assume that there are two business support functions: a thread function and a mail management function. In this case, for creating a unique business support model M2B1 for the thread function of a certain first organization, the training data acquisition unit 102 acquires unique training data created based on the activity data of the first organization. For creating a unique business support model M2B2 for the thread function of another second organization, the training data acquisition unit 102 acquires unique training data created based on the activity data of the second organization.

[0122] For example, for creating a unique business support model M2B3 for the mail management function of a certain first organization, the training data acquisition unit 102 acquires unique training data created based on the activity data of the first organization. For creating a unique business support model M2B4 for the mail management function of another second organization, the training data acquisition unit 102 acquires unique training data created based on the activity data of the second organization.

[0123] For example, the relearning unit 103 creates, for each organization, a unique business support model M2B that is a business support model M2 unique to the organization by performing relearning of the common business support model M2A based on the unique training data of the organization. The unique business support model M2B in Modification 3 is a model specialized for a specific organization. The unique business support model M2B is a model for a user belonging to a specific organization to use a specific business support function. The unique business support model M2B is the final model provided to users belonging to a specific organization. The method of relearning based on the unique training data is the same as the relearning based on the training data described in the embodiment.

[0124] For example, assume that there are two business support functions: a thread function and a mail management function. In this case, the relearning unit 103 creates a specific business support model M2B1 that can provide business support specific to the first organization by executing relearning of the common business support model M2A1 for the thread function based on the specific training data created from the activity data of the first organization. The relearning unit 103 creates a specific business support model M2B2 that can provide business support specific to the second organization by executing relearning of the common business support model M2A1 for the thread function based on the specific training data created from the activity data of the second organization.

[0125] For example, the relearning unit 103 creates a specific business support model M2B3 that can provide business support specific to the first organization by executing relearning of the common business support model M2A2 for the mail management function based on the specific training data created from the activity data of the first organization. The relearning unit 103 creates a specific business support model M2B4 that can provide business support specific to the second organization by executing relearning of the common business support model M2A2 for the mail management function based on the specific training data created from the activity data of the second organization.

[0126] In the above description, the case where there are two business support functions and two organizations is taken as an example for explanation. However, the number of business support functions and organizations can be arbitrary. For example, when there are three or more business support functions, three or more common business support models M2A may be created. When there are three or more organizations, three or more specific business support models M2B may be created for each business support function.

[0127] For example, the relearning unit 103 records the unique business support model M2B of each of the plurality of organizations in the data storage unit 100. The relearning unit 103 transmits the unique business support model M2B of each of the plurality of organizations to the business support server 20. The business support server 20 records in the data storage unit 200 by associating the organization ID of each of the plurality of organizations with the unique business support model M2B of the organization. When a user belonging to a certain organization uses a certain business support function, the business support unit 201 supports the user's business based on the unique business support model M2B for the business support function associated with the organization ID of the organization.

[0128] In addition, when a user belonging to an organization in which the unique business support model M2B has not been created logs in to the business support system 2, the business support unit 201 may support the user's business based on the common business support model M2A. In this case, the business support unit 201 may support the user's business based on the unique business support model M2B associated with the organization ID of another organization (for example, an organization with the same industry type or employee scale) instead of the common business support model M2A.

[0129] The relearning system 1 of Modification Example 3 creates the common business support model M2A by performing relearning of the large language model M1 based on the common training data. The relearning system 1 creates the unique business support model M2B by performing relearning of the common business support model M2A based on the unique training data of each organization for each organization. As a result, after the relearning system 1 creates the common business support model M2A based on more common training data, it can create the unique business support model M2B specialized for the tendency of each individual organization to which the users using a specific business support function belong, so that the business support model M2 unique to each individual organization can be created efficiently. For example, the relearning system 1 can realize highly accurate business support according to a specific business support function and a specific organization.

[0130] [6-4. Modification Example 4] For example, in Modifications 1 to 3, a plurality of business support models M2 such as a common business support model M2A and a specific business support model M2B are created. Each of these plurality of business support models M2 may be used for data augmentation of training data.

[0131] FIG. 9 is a diagram showing an example of a function realized in Modification 4. For example, the relearning terminal 10 in Modification 4 includes a training data creation unit 104. The training data creation unit 104 is realized by the control unit 11. The training data creation unit 104 creates new training data based on each of a plurality of training data and each of a plurality of business support models M2. For example, the training data creation unit 104 inputs the input part of the training data used for the relearning of each of the plurality of business support models M2 to each of the business support models M2. The business support model M2 outputs in response to the input part.

[0132] For example, the training data creation unit 104 acquires the input part of the training data input to the business support model M2 as the input part of the new training data. The training data creation unit 104 acquires the output from the business support model M2 as the output part of the new training data. The output of the new training data does not have to be the output from the business support model M2 itself. For example, data in which a part of the output from the business support model M2 is changed may be the output part of the new training data. The training data creation unit 104 creates a pair of the input part and the output part as the new training data. The training data creation unit 104 stores the new training data in the training database DB1.

[0133] The relearning unit 103 in Modification 4 executes further relearning of at least one of the plurality of business support models M2 based on the new training data. The relearning unit 103 may execute relearning of all the business support models M2 or may execute relearning of only some of the business support models M2 based on the new training data. The further relearning is different from the relearning in the embodiment and Modifications 1 to 3 in that new training data is used, but the method of relearning may be the same as that in the embodiment and Modifications 1 to 3.

[0134] The relearning system 1 of Modification Example 4 creates new training data based on each of a plurality of training data and each of a plurality of business support models M2. The relearning system 1 performs further relearning of at least one of the plurality of business support models M2 based on the new training data. Thereby, since the relearning system 1 can realize data augmentation using each of the plurality of business support models M2, effective relearning can be realized.

[0135] [6-5. Modification Example 5] For example, even if the relearning system 1 attempts to create a business support model M2 specific to a certain organization, there may not be a sufficient amount of activity data in that organization. Therefore, activity data of other organizations may be used to create a business support model M2 specific to an organization. In Modification Example 5, the organization for which the business support model M2 is to be created is referred to as the first organization. Another organization whose activity data is used in creating the business support model M2 specific to the first organization is referred to as the second organization.

[0136] The training data acquisition unit 102 of Modification Example 5 acquires first organization training data, which is training data created based on the activity data of the first organization, and second organization training data, which is training data created based on the activity data of the second organization related to the first organization. The second organization is another organization associated with the first organization. For example, the second organization is another organization having the same industry type or employee scale as the first organization.

[0137] It is assumed that in the training database DB1 of Modification Example 5, an organization ID of the organization to which the user who performed the activity indicated by the activity data used in creating the respective training data is associated with each piece of training data. Further, it is assumed that the association between the first organization and the second organization is defined in the data storage unit 100. For example, if the second organization is another organization of the same industry type as the first organization, the industry type of each organization is defined in the data storage unit 100. If the second organization is another organization of the same employee scale as the first organization, the employee scale of each organization is defined in the data storage unit 100.

[0138] The relearning unit 103 of Modification Example 5 creates a business support model M2 specific to the first organization based on the first organization training data and the second organization training data. For example, the relearning unit 103 creates a business support model M2 specific to the first organization by performing relearning of the large language model M1 based on the first organization training data and the second organization training data. When a common business support model M2A is created as in Modification Examples 1 to 3, the relearning unit 103 creates a specific business support model M2B for the first organization by performing relearning of the common business support model M2A based on the first organization training data and the second organization training data.

[0139] The relearning system 1 of Modification Example 5 creates a business support model M2 specific to the first organization based on the first organization training data and the second organization training data. Thereby, the relearning system 1 can efficiently create a business support model M2 specific to the first organization. For example, even if the first organization is a newly registered organization in the business support system 2 and there is not enough activity data, the relearning system 1 can use the second organization training data based on the activity data of the second organization, so as to improve the accuracy of the business support model M2 specific to the first organization.

[0140] [6-6. Modification Example 6] For example, the second organization in Modification Example 5 may be another organization with a longer usage period of the business support system 2 than the first organization. The data storage unit 100 of Modification Example 6 stores data indicating the usage period of each organization. The relearning terminal 10 can identify which organization corresponds to the second organization by referring to the data.

[0141] The training data acquisition unit 102 of Modification Example 6 acquires the first organization training data of the first organization newly registered in the business support system 2 and the second organization training data of the second organization whose usage period of the business support system 2 is longer than that of the first organization. The training data acquisition unit 102 does not acquire training data for creating the business support model M2 specific to the first organization based on the activity data of other organizations whose usage period of the business support system 2 is shorter than that of the first organization.

[0142] The relearning unit 103 of Modification Example 6 creates a business support model M2 specific to the first organization newly registered in the business support system 2. The relearning unit 103 creates a business support model M2 specific to the first organization based on the second organization training data of the second organization whose usage period of the business support system 2 is longer than that of the first organization, without using training data based on the activity data of other organizations whose usage period of the business support system 2 is shorter than that of the first organization. Although the method of acquiring the second organization training data is different from that of Modification Example 5, the relearning method based on the first organization training data and the second organization training data is the same as that of Modification Example 5.

[0143] The relearning system 1 of Modification Example 6 creates a business support model M2 specific to the first organization newly registered in the business support system 2 based on the first organization training data of the first organization newly registered in the business support system 2 and the second organization training data of the second organization whose usage period of the business support system 2 is longer than that of the first organization. Since there may be more activity data as the usage period is longer, the relearning system 1 can efficiently create a business support model M2 specific to the first organization newly registered in the business support system 2.

[0144] [6-7. Modification Example 7] For example, other data than the activity data to be processed may be input to the business support model M2. In Modification 7, as in the embodiment, the case where the business support task is the creation of a thread summary is taken as an example. For example, there may be keywords in the thread that the user wants to include in the summary. In this case, not only the thread data of the thread to be summarized but also the keywords specified by the user may be input to the business support model M2. In Modification 7, a business support model M2 that can handle such input is created.

[0145] The training data acquisition unit 102 of Modification 7 acquires the entire activity indicated by the activity data, a part selected from the activity data, and the training data shown. The part selected from the activity data is the part that particularly requires retraining. For example, the operating company of the business support system 2 selects a part from the entire activity and annotates it. The input part of the training data is a pair of the entire activity indicated by the activity data and the annotated part. The output part of the training data may be the same as in the embodiment.

[0146] The retraining unit 103 of Modification 7 creates the business support model M2 by performing retraining of the large language model M1 based on the training data indicating the above-mentioned whole and the above-mentioned part. Although the input part of the training data is different from that of the embodiment, the retraining method is the same as that of the embodiment. The retraining unit 103 performs retraining of the large language model M1 so that when the above-mentioned whole and the above-mentioned part indicated by the input part of the training data are input to the large language model M1, the large language model M1 outputs the output part of the training data.

[0147] For example, when a user uses the business support model M2, the user designates the part corresponding to the said part. When the business support task is creating a summary of a thread, the business support unit 201 inputs the thread data of the thread to be summarized and the keyword designated by the user from among the said threads into the business support model M2. The business support model M2 outputs summary data based on the embedded expressions of these texts. The summary indicated by the said summary data includes the keyword designated by the user.

[0148] Note that the part selected from the activity data is not limited to the above example. The said part may be a part corresponding to the business support task. For example, if the business support task is creating an answer to a question, the said part may be the content of the question. If the business support task is creating text, the said part may be a keyword to be included in the text. Otherwise, the said part may be anything that has some correlation with the output part of the training data.

[0149] The relearning system 1 of Modification Example 7 acquires the entire activity indicated by the activity data, a part selected from the activity data, and the training data indicated thereby. The relearning system 1 creates the business support model M2 by executing the relearning of the large language model M1 based on the training data indicating the whole and the part. Thereby, the relearning system 1 can create a more flexible business support model M2.

[0150] [6-8. Other Modification Examples] For example, the above modification examples may be combined.

[0151] For example, the functions described as being realized by the relearning terminal 10 may be realized by the business support server 20. The functions described as being realized by the relearning terminal 10 may be shared by a plurality of computers. [Explanation of Reference Numerals]

[0152] 1 Relearning system, 2 Business support system, 10 Relearning terminal, 11, 21, 31 Control unit, 12, 22, 32 Memory unit, 13, 23, 33 Communication unit, 20 Business support server, 30 User terminal, 14, 34 Operation unit, 15, 35 Display unit, N Network, 100 Data storage unit, 101 Large language model acquisition unit, 102 Training data acquisition unit, 103 Relearning unit, 104 Training data creation unit, 200 Data storage unit, 201 Business support unit, 300 Data storage unit, 301 Display control unit, 302 Operation reception unit, DB1 Training database, DB2 User database, M1 Large language model, M2 Business support model, M2A, M2A1, M2A2 Common business support model, M2B, M2B1, M2B2, M2B3, M2B4 Specific business support model, SC Business support screen.

Claims

1. A large language model acquisition unit that acquires a pre-trained large language model, A training data acquisition unit that acquires training data for re-training the large language model, the training data being created based on activity data indicating the user's activities performed in a business support system that supports the user's business, A re-training unit that creates a business support model specific to the business support system by performing re-training of the large language model based on the training data, A re-training system including the above.

2. The training data acquisition unit acquires a plurality of the training data created from different viewpoints, The re-training unit creates a plurality of the business support models by performing multi-stage re-training of the large language model based on the plurality of training data, The re-training system according to claim 1.

3. The training data acquisition unit acquires common training data, which is the training data common to a plurality of organizations and is created based on the activity data of each of the plurality of organizations for each business support task performed by the business support model, The re-training unit creates a common business support model, which is the business support model common to the plurality of organizations, by performing re-training of the large language model based on the common training data for each business support task, The training data acquisition unit acquires specific training data, which is the training data specific to each organization and is created based on the activity data of that organization, for each organization, The re-training unit creates a specific business support model, which is the business support model specific to each organization, by performing re-training of the common business support model based on the specific training data of that organization for each organization, The re-training system according to claim 2.

4. The training data acquisition unit acquires common training data, which is the training data common to the plurality of business support functions and is created based on the activity data of each of the plurality of business support functions included in the business support system, for each business support task performed by the business support model. The re-learning unit creates a common business support model, which is the business support model common to the plurality of business support functions, by performing re-learning of the large language model based on the common training data for each business support task. The training data acquisition unit acquires specific training data, which is the training data specific to each business support function and is created based on the activity data of the business support function, for each business support function. The re-learning unit creates a specific business support model, which is the business support model specific to each business support function, by performing re-learning of the common business support model based on the specific training data of the business support function for each business support function. The re-learning system according to claim 2.

5. The training data acquisition unit acquires common training data, which is the training data common to the plurality of organizations and is created based on the activity data of each of the plurality of organizations, for each business support function included in the business support system. The re-learning unit creates a common business support model, which is common to the plurality of organizations, by performing re-learning of the large language model based on the common training data for each business support function. The training data acquisition unit acquires specific training data, which is the training data specific to each organization and is created based on the activity of the organization, for each organization. The re-learning unit creates a specific business support model, which is the business support model specific to each organization, by performing re-learning of the common business support model based on the specific training data of the organization for each organization. The re-learning system according to claim 2, comprising:

6. The retraining system further includes a training data creation unit that creates new training data based on each of the plurality of training data and each of the plurality of business support models. The retraining unit performs further retraining on at least one of the plurality of business support models based on the new training data. The retraining system according to any one of claims 2 to 5, further comprising:

7. The training data acquisition unit acquires first organization training data, which is the training data created based on the activity data of a first organization, and second organization training data, which is the training data created based on the activity data of a second organization related to the first organization. The retraining unit creates the business support model specific to the first organization based on the first organization training data and the second organization training data. The retraining system according to any one of claims 1 to 5.

8. The training data acquisition unit acquires the first organization training data of the first organization newly registered in the business support system and the second organization training data of the second organization that has a longer usage period of the business support system than the first organization. The retraining unit creates the business support model specific to the first organization newly registered in the business support system. The retraining system according to claim 7.

9. The training data acquisition unit acquires the training data indicating the entire activity indicated by the activity data and a part selected from the activity data. The retraining unit creates the business support model by performing retraining on the large language model based on the training data indicating the whole and the part. The retraining system according to any one of claims 1 to 5.

10. Acquire a pre-trained large language model. Obtain training data for retraining the large language model, which is created based on activity data indicating the user's activities performed in a business support system that supports the user's business. Create a business support model specific to the business support system by performing retraining of the large language model based on the training data. Retraining method. **Claim 11** A large language model acquisition unit that acquires a pre-trained large language model. A training data acquisition unit that acquires training data for retraining the large language model, which is created based on activity data indicating the user's activities performed in a business support system that supports the user's business. A retraining unit that creates a business support model specific to the business support system by performing retraining of the large language model based on the training data. A program for causing a computer to function as such.

Images