Computer programs, information processing devices, and communication systems

Abstract

To provide computer programs, information processing devices, and communication systems that enable more accurate and convenient information presentation. [Solution] The computer program causes the computer to receive a first sentence in natural language via a predetermined communication application, identify related data for the first sentence, and provide the identified related data along with the first sentence to a first language model that has been trained to output a response sentence for a given natural language sentence. The program then outputs response data based on the response sentence output by the first language model via the predetermined communication application.

Description

Technical Field

[0001] The present invention relates to a computer program, an information processing apparatus, and a communication system that realize more accurate information presentation in business support.

Background Art

[0002] For business improvement in enterprises, there is a need to quickly and accurately extract the knowledge accumulated within the enterprise. In response to this, business support using AI (Artificial Intelligence) called a Large Language Model that outputs natural language in response to input in natural language has been proposed (Patent Document 1, etc.).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] When using a system that extracts data having a context similar to a query in natural language input from the knowledge accumulated within an enterprise using a large language model, depending on the way of inputting the query, there are cases where the output actually required by the user cannot be obtained. Technical skills are required for query setting. Also, in a system that returns a response to a question based on a large language model, it is often used in such a way that responses to a series of inquiries by one user are output.

[0005] An object of the present disclosure is to provide a computer program, an information processing apparatus, and a communication system that enable more accurate and more convenient information presentation.

Means for Solving the Problems

[0006] A computer program in one embodiment of the present disclosure causes a computer to receive a first sentence in natural language via a predetermined communication application, identify related data for the first sentence, provide the identified related data along with the first sentence to a first language model that has been trained to output a response sentence for a given natural language sentence, and output response data based on the response sentence output from the first language model via the predetermined communication application.

[0007] The computer program, information processing device, and communication system disclosed herein implement a service on a communication application that responds to queries using a language model. The communication application may be an existing application, and for example, an instance that responds using a language model may be associated with an account in the communication application. By querying this account, a user can receive responses that are similar to querying an expert within the company, providing information based on data within the organization. [Effects of the Invention]

[0008] According to this disclosure, users can obtain response sentences using a language model as responses to questions they make to other users on a communication application. [Brief explanation of the drawing]

[0009] [Figure 1] This is an overview diagram of the communication system disclosed herein. [Figure 2] This is a block diagram showing the configuration of an information processing device. [Figure 3] This is a schematic diagram of the first and second language models used by the information processing device. [Figure 4] This is a block diagram showing the configuration of a terminal device. [Figure 5] This is a flowchart illustrating an example of the procedure for sending and receiving messages. [Figure 6] An example of a message output by a terminal device is shown below. [Figure 7] Here is another example of a message output by a terminal device. [Figure 8] Here is another example of a message output by a terminal device. [Figure 9] Here is another example of a message output by a terminal device. [Figure 10] Here is another example of a message output by a terminal device. [Figure 11] Here is another example of a message output by a terminal device. [Modes for carrying out the invention]

[0010] This disclosure will be described in detail with reference to drawings illustrating embodiments thereof.

[0011] Figure 1 is an overview diagram of the communication system 100 of this disclosure. The communication system 100 is a system that enables multiple users to send and receive messages between terminal devices 2 each. The communication system 100 includes an information processing device 1, multiple terminal devices 2, and a data server 3.

[0012] Information processing device 1 is connected to a local network LN within an organization such as a company to which multiple users belong. A data server 3, which stores data shared within the organization, is connected to the local network LN, and communication is possible from information processing device 1 to data server 3. Information processing device 1 can read the data stored in data server 3.

[0013] The local network LN is connected to the network N, which is an external public communication network, via the communication device 4. The local network LN is a wired and / or wireless communication medium, and realizes communication connections and data transmission / reception from the information processing device 1 and the terminal device 2 that are authenticated within the organization. The local network LN may include a VPN (Virtual Private Network) over the public communication network N. In this case, it may accept communication connections from the terminal device 2 that is physically separated from other communication devices owned by the organization.

[0014] The communication system 100 can communicate with a server 6 outside the system that provides communication services via the network N. Each of the plurality of terminal devices (corresponding to the second computer) 2 can send and receive messages within a group corresponding to the organization based on the account data of the communication service. In the communication system 100 of the present disclosure, a specific account is also given to the information processing device (corresponding to the first computer) 1 on the communication service. The information processing device 1 can receive a message input from the terminal device 2 on the communication service and output a response message.

[0015] The communication service includes a chat service, a message transmission / reception service, a data sharing service, etc. In these services, one user can send and receive one-on-one messages with other users. In these services, group settings including a plurality of users are possible, and one user can send and receive messages with a plurality of other users. The message basically includes text. Image data such as a photo, drawing, table, or graph, voice data, etc. can be attached to the message. A moving image, a document file, etc. can be attached to the message. In the communication service, these data can be shared.

[0016] In the communication system 100 of the present disclosure, the information processing apparatus 1 uses a language model within the local network LN and creates a response sentence for the received message in natural language. At this time, the information processing apparatus 1 can output a response sentence using the data of the data server 3 as knowledge.

[0017] The data server 3 that can be referred to by the information processing apparatus 1 stores data that is set to be shareable within the local network LN. The data server 3 may be an on-premises server within the local network LN or a cloud server that can be referred to via the network N. The data stored in the data server 3 may be structured data described by common attribute items and specific data therefor. The data stored in the data server 3 also includes image data and unstructured data composed of text and images. For example, the unstructured data may use data used in presentations, PDF data used for distribution, and image data itself that may include videos. In the communication system 100 of the present disclosure, these unstructured data can also be used as intellectual property (knowledge) shared within the organization. The information processing apparatus 1 can read and write the data of the data server 3 via the API for the database provided by the data server 3.

[0018] As will be described later, the information processing apparatus 1 not only uses the first language model M1 as it is, but also executes a process of analyzing a sentence corresponding to a query given to the first language model M1 using the second language model M2, and outputs a response sentence with higher accuracy. Thereby, for example, when a message of A is output from the terminal device 2 of one user, it is received by a specific account belonging to the group and the terminal devices 2 of other users via the server 6, and a response sentence of B is output from the information processing apparatus 1 corresponding to the specific account, and is output by the terminal devices 2 of the users belonging to the group. Further, it is possible to continue the conversation by further sending a message to the response sentence from the terminal devices 2 of other users belonging to the group and share the response from the information processing apparatus 1.

[0019] The following describes the configuration and processing details of the information processing device 1, which enables accurate information processing in response to user responses and in returning responses on the communication service.

[0020] Figure 2 is a block diagram showing the configuration of the information processing device 1. The information processing device 1 comprises a processing unit 10, a storage unit 11, and a communication unit 12. The information processing device 1 uses a server computer. In the following description, the information processing device 1 is described as being composed of a single server computer, but it may also be configured such that multiple server computers are connected via a local network LN and perform distributed processing to function as a single information processing device 1. The information processing device 1 functions as an on-premise server, but it may also function as a cloud server while ensuring security.

[0021] The processing unit 10 includes one or more processors such as a CPU (Central Processing Unit), MPU (Micro-Processing Unit), or GPU (Graphics Processing Unit). The processing unit 10 also includes memory, which is a temporary storage medium such as SRAM (Static Random Access Memory) or DRAM (Dynamic Random Access Memory). The processing unit 10 may be configured as a single hardware (SoC: System On a Chip) integrating the processor, memory, storage unit 11, and communication unit 12. The processing unit 10 reads the information processing program P1 stored in the storage unit 11 into memory and executes it, thereby causing a general-purpose computer to perform various processes described later, and to function as the information processing device 1 of this disclosure.

[0022] The storage unit 11 is a relatively large-capacity non-temporary storage medium such as a hard disk or flash memory. The storage unit 11 stores the program (program product) necessary for the processing unit 10 to execute processing, and reference setting data. The program product includes an information processing program P1, a communication application program (hereinafter referred to as a communication app) P21 for using the communication service, a first language model M1, and a second language model M2. Part or all of the first language model M1 and the second language model M2 may be stored on an external server that provides language models, and the processing unit 10 may be configured to obtain the processing results of the first language model M1 or the second language model M2 via network N or local network LN.

[0023] The configuration data stored in the memory unit 11 includes data for the communication service to be used. As described above, the communication service is selected by the organization from one of the following: a group chat service, a message sending / receiving service, or a data sharing service. The configuration data includes account data for the communication service set by the organization. The account data includes the account name and password, or authentication data. The account data is set so that the information processing device 1 can be recognized by users within the organization as the account data. The processing unit 10 can log in to the communication service via the communication application P21 based on the account data. The communication application P21 may include a web client.

[0024] The configuration data stored in the memory unit 11 includes settings for the role and behavior of a specific account on the communication service being used. For example, it may be set whether a particular account is in a neutral "robot" position, a "supervisor" position over users within the organization, an account with special privileges to access all data, or an account that corresponds to other users and can only access some data. The configuration data may also include settings for the technical fields in which a particular account is an expert. The configuration data may also include settings for the tone (way of expression) of responses from a particular account.

[0025] The information processing program P1, communication application P21, first language model M1, and second language model M2 stored in the memory unit 11 may be obtained by the processing unit 10 reading the information processing program P9, communication application P8, first language model M91, and second language model M92 stored in the computer-readable non-temporary storage medium 9 and storing them in the memory unit 11. Alternatively, the information processing program P1, communication application P21, first language model M1, and second language model M2 may be obtained by the processing unit 10 downloading them from a download server via the communication unit 12 and storing them in the memory unit 11.

[0026] The communication unit 12 enables communication via the local network LN and network N. The communication unit 12 may be a wired network card or a wireless communication device for Wi-Fi. The communication unit 12 may also be a wireless communication device connected to a carrier network. The processing unit 10 can send and receive data with the data server 3 and terminal device 2 via the communication unit 12.

[0027] Figure 3 is an overview diagram of the first language model M1 and the second language model M2 used by the information processing device 1. Both the first language model M1 and the second language model M2 use a language model. The first language model M1 and the second language model M2 may also be large language models. The first language model M1 and the second language model M2 may include an image recognition model using a Vision Transformer or the like that outputs the features of the input image data as text. The first language model M1 and the second language model M2 are basically trained models that convert a given natural language text (query) into vector data and output a response sentence with appropriate features in natural language for a sentence (weighted word group) that has features represented by the vector data. The first language model M1 also accepts input of data that should be referenced as knowledge from the data server 3 in response to the query.

[0028] The first language model M1 is positioned downstream of the second language model M2, and by receiving data output from both the first and second language models M2, it enables the output of more accurate responses to queries.

[0029] As described above, the first language model M1 is a language model trained to output the text of a response sentence corresponding to a given text. The first language model M1 may be a model selected by the organization using the communication system 100 from among multiple models trained with different algorithms. Multiple first language models M1 trained with different algorithms may be made available for use, and the processing unit 10 may determine which algorithm's first language model M to use depending on the content of the sent and received messages before continuing processing.

[0030] The second language model M2, like the first language model M1, uses a language model to convert a given natural language sentence into vector data and is a model trained to output a sentence (or simply a weighted word set) that corresponds to the meaning or context of the input from the features represented by the vector data. The second language model M2 may also output image data such as photographs, figures, tables, or graphs. For example, the second language model M2 is trained to extract word sets that represent the background or context of a given natural language sentence and complement the given natural language sentence, and to output a sentence using those word sets. By utilizing the second language model M2, the information processing device 1 performs a function to refine the input sentence by repeatedly asking itself questions about the given natural language sentence so that it becomes a query that can obtain an appropriate answer from the first language model M1. Within the second language model M2, the processing unit 10 may further feed the answer sentence output from the language model part back to the language model and check for deviations in meaning, repeating this process until it outputs a sentence that appropriately explains the meaning, background, or premise.

[0031] For example, given a natural language sentence such as, "Please suggest a menu with 3 dishes. The salt content should be 10g or less, and the vegetables should be 700g," the second language model M2 is trained to output a sentence that expresses the gist of the user's request, based on the words "menu," "salt," and "vegetables," meaning, "The user is requesting a recipe for a healthy, vegetable-centered menu. The recipe should be low in salt, based on the target salt intake per meal for adults, and high in vegetables, based on the target vegetable intake per meal for adults."

[0032] The second language model M2, when outputting a sentence (including image data as needed) that corresponds to the main idea or context using a given natural language sentence, is trained to extract word groups similar to the background or context of the given sentence from the configured knowledge (in this case, data stored in data server 3). In other words, the second language model M2 identifies the common knowledge that is assumed when a user converses with other users within their organization from the data stored in data server 3, and outputs a sentence that corresponds to the main idea or context.

[0033] The information processing device 1 uses the first language model M1 and the second language model M2 shown in Figure 3 to provide the received message (first sentence) to the second language model M2 as needed, and then provides the original message alone, or combined with the output from the second language model M2 (second sentence), to the first language model M1 to obtain a response. Details of the process will be described later.

[0034] Figure 4 is a block diagram showing the configuration of terminal device 2. Terminal device 2 is a smartphone, tablet device, or personal computer. Terminal device 2 comprises a processing unit 20, a storage unit 21, a communication unit 22, a display unit 23, an operation unit 24, and an audio input / output unit 25.

[0035] The processing unit 20 includes one or more processors such as CPUs, MPUs, and GPUs. The processing unit 20 also includes memory, which is a temporary storage medium such as SRAM or DRAM. The processing unit 20 may be configured as a single hardware (SoC: System On a Chip) integrating the processor, memory, storage unit 21, and communication unit 22.

[0036] The storage unit 21 is a relatively large-capacity non-volatile storage area such as an SSD or flash memory. The storage unit 21 stores the program (program product) necessary for the processing unit 20 to execute processing, and reference configuration data. The program product includes a communication application P22. The communication application P22 may also include a web client. The communication application P22 may be downloaded by the processing unit 20 from a download server via the communication unit 22 and stored in the storage unit 21. The communication application P22 may also be read by the processing unit 20 from a communication application P72 stored on the storage medium 7 and stored in the storage unit 21.

[0037] The communication unit 22 enables communication with the communication service server 6 via the network N. The communication unit 22 enables communication with the data server 3 or information processing device 1 within the local network LN. The communication unit 22 may be a wired network card or a WiFi wireless communication device. The communication unit 22 may also be a wireless communication device connected to a carrier network. The processing unit 20 can send and receive data for the communication service provided by the server 6 using the communication unit 22.

[0038] The display unit 23 is a display such as a liquid crystal display or an organic EL (Electro-Luminescence) display. The display unit 23 may also be a touch panel-integrated display. The processing unit 20 displays a communication screen obtained by the server 6's functions on the display unit 23 and can display text entered by the user, as well as images including photographs, diagrams, tables, or graphs.

[0039] The operation unit 24 is a user interface that can input and output to the processing unit 20, and includes, for example, a keyboard and a pointing device such as a mouse. The operation unit 24 may also include physical buttons, switches, physical dials, or a touch panel built into the display unit 23.

[0040] The audio input / output unit 25 includes a speaker and a microphone. The processing unit 20 or the audio input / output unit 25 itself may have a speech recognition function that converts audio collected using the microphone into text data. The processing unit 20 can output sound effects and voice using the speaker.

[0041] The processes performed in the communication system 100 configured in this way will now be described. Figure 5 is a flowchart showing an example of the procedure for sending and receiving messages.

[0042] The following processing procedure is initiated when a user uses terminal device 2 to input a message to information processing device 1. The following processing procedure is executed when information processing device 1 determines that a user has designated a corresponding account as the recipient of a message, and that user sends the message. The following processing procedure is also executed when a message is sent from any user belonging to a group that is configured to include the account of information processing device 1.

[0043] When the processing unit 10 of the information processing device 1 receives a message (step S101), it extracts the text contained in the received message (corresponding to the first sentence) (step S102). The processing unit 10 stores the extracted text (first sentence) as a log in the storage unit 11, associating it with the account of the message sender's terminal device 2 and time information (date and time) (step S103).

[0044] The processing unit 10 determines whether the message received in step S101 contains an instruction to use the second language model M2 (step S104). The instruction may be expressed, for example, as a script or tag that starts with a specific character. The processing in step S104 may be replaced by processing based on the information processing program P1 of the processing unit 10 itself to determine whether or not to use the second language model M2.

[0045] If the processing unit 10 determines that the instruction to use the second language model M2 is included (S104: YES), it reads the text of messages previously sent from the same account that were stored within a predetermined time (step S105). In step S105, the processing unit 10 may read the text of past messages only if the text extracted in step S102 contains expressions indicating past messages or expressions indicating previously recorded data, or it may read the text of messages from the indicated period. For example, the predetermined time is stored in the storage unit 11 as 3 hours. This makes it possible to maintain the context of a conversation about a continuous topic. Also, if the text extracted in step S102 contains phrases that refer to past messages beyond a predetermined time, such as "yesterday," "a month ago," or "last year," the processing unit 10 may read the text of messages from the account from the corresponding period.

[0046] The processing unit 10 provides the text extracted in step S102 and the text read in step S105 to the second language model M2 (step S106). If the received message has image data attached, the processing unit 10 provides the image data along with the text to the second language model M2 (step S107). If the received message does not have image data attached, the processing unit 10 skips the process in step S107.

[0047] The processing unit 10 stores the text output from the second language model M2 (corresponding to the second sentence) as a log in the storage unit 11, associating it with the account and time information of the message sender's terminal device 2 (step S108). The second language model M2 is configured to perform a query for data from the data server 3 during its internal processing. In step S108, the processing unit 10 may obtain and store the reference location of the queryed data (including the file name, directory, and folder location) from the second language model M2.

[0048] The data queried by the second language model M2 (data to supplement the background or assumptions of the message) includes data about the organization to which the user who sent the message belongs. The processing unit 10 may refer to the user's responsibilities, positions, etc., within the organization from the personnel data in the data server 3, or it may store roles such as "sales," "development," "research," and "public relations" in the storage unit 11, associated with account data, and refer to the data from there.

[0049] The processing unit 10 retrieves relevant data from the data server 3 as needed based on the output from the second language model M2 (step S109). In step S109, the processing unit 10 retrieves the reference destination of the data output from the second language model M2. In step S109, the processing unit 10 may also identify the relevant data itself based on the meaning-indicating text output from the second language model M2. Step S109 corresponds to the RAG (Retrieval-Augmented Generation) function. By using the relevant data, it is possible to refer to organizational terminology and past cases stored in the data server 3 and output response sentences from specific accounts in the organizational context, and furthermore, it is expected that the occurrence of hallucination in responses from the first language model M1 can be suppressed.

[0050] The processing unit 10 creates a query for the first language model M1 from the text obtained from the second language model M2, the acquired related data, and the image data attached to the message in step S101 (step S110). When creating the query in step S110, if the message received in step S101 has image data attached, the processing unit 10 provides the image data to the Embedding module corresponding to the first language model M1 to obtain it as vector data and adds it to the query. If the message received in step S101 does not have image data attached, it does not need to be used to create the query.

[0051] The processing unit 10 provides the query created in step S110 to the first language model M1 (step S111). The processing unit 10 obtains the response statement output from the first language model M1 (step S112). The processing unit 10 sends the response data, including the obtained response statement, to the server 6 on the communication application P21 as a response to the message received in step S101, using the account data assigned to it (step S113). The response data may include the response statement, the file name of the reference to related data, and image data. As a result, the response from the information processing device 1 to the message sent in step S101 is notified to the terminal device 2 via the communication service server 6.

[0052] If step S104 determines that there is no instruction to use the second language model M2 (S104: NO), the processing unit 10 may proceed to step S110 and directly provide the received message to the first language model M1 to obtain a response.

[0053] The functions realized by the communication system 100 using the first language model M1 and the second language model M2 will be explained with specific examples of the processing procedure shown in Figure 5. Figure 6 shows an example of a message output by the terminal device 2. Figure 6 shows a message screen 230 based on the communication application P22. The message screen 230 in Figure 6 shows an example of what is displayed on the display unit 23 of the terminal device 2 corresponding to the user "XX" account.

[0054] The message screen 230 includes a recipient list screen 231 that shows a list of accounts and groups of other users to whom the message is addressed. In the example in Figure 6, the recipient list screen 231 includes icons for recipients who can send and receive messages using the communication app P22, a specific account "JJJ" corresponding to the information processing device 1, another user "AAA", the specific account "JJJ", and the group "SSS" which includes the other user "AAA" and yet another user "BBB". The icon for the specific account "JJJ" shows a specific account corresponding to the information processing device 1 and a specific icon to indicate that it is an account that responds based on AI.

[0055] The message screen 230 includes an input section 232 for entering the message to be sent. In the input section 232, the user can input text or image data using the operation section 24. The message screen 230 includes a send button 233 for sending the text and / or image data entered in the input section 232 to the communication service server 6. In the message screen 230 shown in Figure 6, the sent and received messages are displayed in speech bubble-shaped objects 234. The objects 234 are displayed in chronological order of when they were sent and received, arranged vertically on the screen. The display manner of sent and received messages is not limited to this. Since it is used with a communication application P22 that corresponds to an existing chat or data sharing service, the display manner may be adapted to match the display manner of those applications.

[0056] In the example shown in Figure 6, the message screen 230 is a screen that shows a one-to-one conversation between a user with account "XX" and a specific account "JJJ" on the information processing device 1. First, a text message from user "XX" that reads "Please tell me the release schedule for product Q" is displayed.

[0057] The processing unit 10 of the information processing device 1 receives the message shown in Figure 6 as a message addressed to a specific account "JJJ" corresponding to itself. The processing unit 10 provides the text "Please tell me the release schedule for product Q" from user "XX" to the second language model M2. Through the processing of the processing unit 10 itself, or through the functions of the second language model M2, it is determined whether the user of account "XX" is in charge of "development" or "publication," and the second language model M2 outputs text to supplement the message text. For example, the second language model M2 outputs a sentence (second sentence) indicating that "the user in charge of development is requesting the schedule of the development staff before and after the release of product Q." The output from the second language model M2 does not have to be in the form of a sentence; it may also be in the form of a weighted word group included in the query provided to the first language model M1.

[0058] The processing unit 10 of the information processing device 1 may further provide the sentence "The user in charge of development is requesting the schedule of the development staff before and after the release of product Q" to the second language model M2 and repeat the processing until an appropriate second sentence is obtained. For example, the second language model M2 may further upgrade the output to something like "The user in charge of development is requesting the schedule of the development staff and external press before and after the release of product Q, as well as the document that serves as the basis for the latest schedule." The processing unit 10 of the information processing device 1 provides the output obtained from the second language model M2 to the first language model M1 to obtain a response sentence.

[0059] Figure 7 shows another example of a message output by terminal device 2. Figure 7 shows an example where a response message from a specific account "JJJ" corresponding to information processing device 1 is output to the message screen 230 shown in Figure 6. The response message from the specific account "JJJ" is output based on the response message obtained when the message from user "XX" using the second language model M2 described above, along with its meaning, is provided to the first language model M1. In the example in Figure 7, the response output is "The schedule related to user XX for the release of product Q is as follows. The following data can be accessed from the data server."

[0060] As shown in the example in Figure 7, the user can view, on the communication app P21, the schedules of the message sender rather than the schedules of all staff members before and after the release of product Q. In this case, schedules of external press releases and other events critical to the development schedule are supplemented. Also, in the example shown in Figure 7, the information processing device 1 provides information on the location where the document containing the relevant schedule is stored, preventing the first language model M1 from presenting an incorrect schedule.

[0061] As a result, users of terminal device 2 can, while using the communication app P22 to send and receive messages to share information and communicate with other users on a daily basis, obtain knowledge-based responses from the information processing device 1 by simply specifying a particular account as the recipient.

[0062] The information processing device 1, by using the second language model M2, can reflect the true meaning of the user's question, the premise or background of the question, and the underlying context of the conversation exchanged between users in the query to the first language model M1, thereby enabling the creation of a more appropriate response.

[0063] Figure 8 shows another example of a message output by terminal device 2. Similar to Figure 6, Figure 8 shows an example displayed on the display unit 23 of terminal device 2 corresponding to user "XX"'s account. In the example in Figure 8, the message screen 230 shows a conversation within the group "SSS" which includes user "XX", a specific account "JJJ", and other users "AAA" and "BBB". In the message screen 230 of Figure 8, the terminal device 2 of user "XX" sends request text along with image data to the specific account and the other users. The request is exemplified as "JJJ, output a plot using this image material". The image data includes illustrations of two characters and handwritten text about the characters.

[0064] The processing unit 10 of the information processing device 1 receives this message as a message addressed to a specific account, "JJJ". Since the conversation takes place within the group "SSS", which includes users "XX", "AAA", and "BBB", the processing unit 10 refers to the history of past messages, including their account data, and the "responsibilities" within the organization of the group "SSS", from the storage unit 11. For example, the processing unit 10 can identify that it is the "product Q development group" from the data server 3 or the user affiliation data for each project stored in the storage unit 11.

[0065] The processing unit 10 provides the second language model M2 with the text "Output a plot using the material from this image" and the transmitted image data. Through the processing of the processing unit 10 itself, or through the functions of the second language model M2, it is recognized that the image based on the image data contains an illustration of a character and that the character's setting information is written as text. As a result, the second language model M2 outputs a sentence (second sentence) indicating, for example, "A user in charge of developing product Q is requesting the output of a story plot using two characters as material." By using the second language model M2, the processing unit 10 outputs text that confirms that the "plot" is a plot for a "story." If there is a prior discussion within the group "SSS" about planning the production of an animation to promote product Q, or if there is a track record of creating a similar "story" using characters in a project related to product Q within the group "SSS" in the past, the processing unit 10 can more easily confirm that it is a plot for a "story" and reduce the possibility of misinterpreting the intent. The processing unit 10, which uses the second language model M2, may repeatedly provide the outputted second sentence to the second language model M2, thereby further refining the second sentence by making it a self-questioning process.

[0066] Figure 9 shows another example of a message output by terminal device 2. Figure 9 shows an example where a response message from a specific account "JJJ" corresponding to information processing device 1 is output to the message screen 230 shown in Figure 8. The response message from the specific account "JJJ" is based on a response message obtained when the message from user "XX" using the second language model M2 described above, along with its meaning, is provided to the first language model M1. In the example in Figure 9, the response is output as follows: "The plot of the story using the characters in the image as material is as follows: Chapter 1 "The Door to the Journey," the protagonist in the image sets out on a journey. Chapter 2 "The Kingdom," the protagonist encounters a mysterious character in the image. ...The adventure of two people traveling between the past and the future unfolds."

[0067] Figure 10 shows another example of a message output by terminal device 2. Figure 10 shows an example where a message corresponding to a question from another user "AAA" is output to the message screen 230 shown in Figure 9. User "AAA" asks, "JJJ, how many years ago is 'the past'?"

[0068] The processing unit 10 of the information processing device 1 receives a message from user "AAA" as shown in Figure 10, addressed to a specific account "JJJ" that corresponds to it. The processing unit 10 provides the text "How many years ago is 'the past'?" from user "AAA" and the conversation history of the group "SSS" up to that point to the second language model M2. For example, the second language model M2 outputs a sentence (second sentence) indicating that "user 'AAA' is requesting specific examples of 'the past' included in the plot of the output story." The processing unit 10 of the information processing device 1 provides the output obtained from the second language model M2 to the first language model M1 to obtain a response. Alternatively, the processing unit 10 may provide the first language model M1 with a question following the response sentence from "JJJ" shown in Figure 9, without going through the second language model M2, to obtain output.

[0069] Figure 11 shows an example where a response message from a specific account "JJJ" corresponding to the information processing device 1 is output to the message screen 230 shown in Figure 10. For example, the response from the specific account "JJJ" is "This is England 500 years ago."

[0070] As shown in Figures 8-11, the processing of the information processing device 1 of this disclosure allows for responses that reflect the context within the group, even when other users add questions to a response from a specific account "JJJ," rather than in a one-to-one exchange. This enables users to share information based on organizational knowledge using the first language model M1 and the second language model M2 with multiple users, while minimizing resistance to implementation by using the communication app P22, which is used to send and receive messages to share information and communicate with other users on a daily basis.

[0071] The embodiments disclosed above are illustrative in all respects and not restrictive. The scope of the present invention is indicated by the claims, and all modifications within the meaning and scope equivalent to the claims are included. [Explanation of Symbols]

[0072] 1. Information Processing Device 10 Processing Unit 11 Storage section P1 Information Processing Program (Computer Program) M1 First Language Model M2 Second Language Model 2 Terminal devices 20 Processing Units 21 Memory section 23 Display section 24 Control section P21, P22 Communication app

Claims

1. On the computer, The first sentence, written in natural language, is received via a designated communication application. Identify the relevant data for the first sentence above, The identified relevant data is provided along with the first sentence to a first language model that has been trained to output a response sentence to a given natural language sentence. The response data based on the response sentence output from the first language model is output via the predetermined communication application. A computer program that executes a process.

2. To the aforementioned computer, The first sentence received is given to a second language model that has been trained to output a second sentence in natural language that corresponds to the meaning of the given input sentence based on analysis. The second sentence output from the second language model is acquired as part or all of the associated data. A computer program according to claim 1 that causes a process to be executed.

3. To the aforementioned computer, The received first sentence is given to a second language model, which is trained to output a second sentence that optimizes the response sentence from the first language model in response to a given natural language input sentence. The second sentence output from the second language model is provided to the first language model as a replacement for the first sentence. A computer program according to claim 1 that causes a process to be executed.

4. To the aforementioned computer, Along with the first sentence, the specified communication application accepts image data of photographs, diagrams, tables, or graphs. The image data is provided to the second language model along with the first sentence to obtain the associated data. The computer program according to claim 2.

5. The computer receives the first sentence via the predetermined communication application running on another computer with which it is connected. A computer program according to any one of claims 1 to 4 that causes a process to be executed.

6. The computer receives a message intended for a specific account on the predetermined communication application as the first sentence. A computer program according to claim 5 that causes a process to be executed.

7. To the aforementioned computer, Create a group that includes the specific account on the predetermined communication application and accounts of different users. Within the group chat of the created group, multiple messages from the accounts of the different users are each received as the first sentence. Of the multiple messages mentioned above, the earlier messages in chronological order are used as the associated data for the later messages in chronological order. A computer program according to claim 6 that causes a process to be executed.

8. The first sentence, written in natural language, is received via a designated communication application. Identify the relevant data for the first sentence above, The identified relevant data is provided along with the first sentence to a first language model that has been trained to output a response sentence to a given natural language sentence. The response data based on the response sentence output from the first language model is output via the predetermined communication application. An information processing device equipped with a processing unit.

9. The first computer and the second computer, which are connected to each other in communication, each store account data on a predetermined communication application. The first computer is, The first sentence in natural language is received by the second computer when the predetermined communication application is started. Identify the relevant data for the first sentence above, The identified relevant data is provided along with the first sentence to a first language model that has been trained to output a response sentence to a given natural language sentence. The response data based on the response sentence output from the first language model is output to the second computer via the predetermined communication application. Communication system.

Images