Information processing system and information processing program

Abstract

An information processing system (1A) comprises: a request reception unit (11) for receiving a request regarding an operation to be performed in the use of a machine tool; an input generation unit (12) for generating an input dataset according to the request by referring to at least one of a specification database for containing information representing specifications of the information processing system (1A) or specifications of the machine tool or a non-machine-tool device which are external devices for the information processing system (1A), a setting parameter related to a setting of the information processing system (1A) or the setting of the external devices for the information processing system (1A), and retention data retained in a storage region which can be accessed by the information processing system (1A); and a response presentation unit (15) for presenting a response, to a request, created on the basis of an output by an artificial intelligence (4) into which the transmitted input dataset has been inputted.

Description

Information processing system and information processing program

[0001] The present disclosure relates to an information processing system and an information processing program.

[0002] In recent years, the declining birthrate and aging population have led to a shrinking working population, which has become a social issue. Therefore, various industries are making efforts to improve labor productivity by using information processing systems that utilize information technology (IT) to reduce the burden on workers and improve work efficiency. However, if the use of information processing systems requires proficiency and experience, it takes a long time and a great deal of effort to master the system, which may prevent the full benefits of improving labor productivity. For this reason, there is a demand for information processing systems that do not require prior knowledge or experience to use.

[0003] One way to realize an information processing system that does not require prior knowledge or experience is to provide a dedicated screen or user interface. For example, Patent Document 1 discloses a system that uses a dedicated screen to guide the user through the input of necessary data in an interactive format, allowing even beginners to easily utilize functions.

[0004] Furthermore, in terms of versatility in use, generative AI (Artificial Intelligence) technology has been developed and spread in recent years. For example, services that provide information to users or assist them in their work using natural language chat, which is one of the highly versatile input methods, are becoming more widely used. Patent Document 2 discloses a system that provides information related to a specific service using a chatbot that interacts with users.

[0005] Given the above background, there is a growing demand for information processing systems that can provide highly versatile usability using natural language chat, etc., rather than providing multiple dedicated screens or multiple user interfaces for each function or use case.

[0006] Patent No. 4271159 Patent No. 7329585

[0007] However, with the technology of Patent Document 1, it is difficult to create a dedicated screen for each specific function in a general-purpose and comprehensive manner while considering all possible users, and it is difficult to provide general-purpose ease of use for all possible functions. Furthermore, since a large number of dedicated screens are required for each function, there are problems in that it takes a long time to build the system and the cost of developing the system is high.

[0008] Furthermore, as disclosed in the above-mentioned Patent Document 2, for example, conventional dialogue response systems using natural language and AI with a language model can provide general-purpose ease of use by using natural language, but it is known that, for example, variations in input can lead to large variations in output quality. Specifically, even in sentences with similar meanings, the output may change if the word order is changed or words are changed to synonyms. In other words, conventional dialogue response systems have lower stability and reproducibility of output in response to input than systems that do not include AI. In other words, when used in fields where errors are not tolerated, such as machine tools, conventional dialogue response systems have a problem in that they cannot achieve a target quality that is usable due to variations in output quality.

[0009] The present disclosure has been made in consideration of the above, and aims to provide an information processing system that provides general-purpose ease of use in natural language dialogue responses using artificial intelligence while enabling the suppression of variation in output quality.

[0010] In order to solve the above-mentioned problems and achieve the objectives, the information processing system of the present disclosure comprises a request receiving unit that receives requests for work to be performed when using a machine tool; a specification database that stores information representing the specifications of the information processing system or the specifications of a machine tool or a device other than a machine tool that is an external device to the information processing system; an input generation unit that generates an input data set in response to the request by referring to at least one of setting parameters related to the settings of the information processing system or the settings of a device external to the information processing system and retained data retained in a memory area accessible to the information processing system; and an answer presentation unit that presents an answer to the request created based on the output of an artificial intelligence to which the transmitted input data set has been input.

[0011] The information processing system according to the present disclosure has the effect of providing general-purpose ease of use while suppressing variation in output quality in natural language dialogue responses utilizing artificial intelligence.

[0012]

[0013] An information processing system and an information processing program according to an embodiment will be described in detail below with reference to the accompanying drawings.

[0014] Embodiment 1. Fig. 1 is a diagram showing a first configuration example of an information processing system according to embodiment 1. Fig. 1 shows an information processing system 1A, which is the first configuration example of an information processing system according to embodiment 1. The information processing system 1A is configured in a terminal device 2A. Fig. 1 also shows the terminal device 2A and an artificial intelligence 4 external to the information processing system 1A. The artificial intelligence 4 is realized by a computer system external to the information processing system 1A.

[0015] The terminal device 2A includes a request receiving unit 11, an input generating unit 12, a transmitting / receiving unit 13, an answer generating unit 14, and an answer presenting unit 15. The request receiving unit 11 receives a request for an operation to be performed when using the machine tool. The input generating unit 12 generates an input data set according to the request received by the request receiving unit 11.

[0016] The transmission / reception unit 13 communicates with devices external to the terminal device 2A. The terminal device 2A communicates with the artificial intelligence 4 through the transmission / reception unit 13. The transmission / reception unit 13 transmits the input dataset generated by the input generation unit 12 to the artificial intelligence 4. The transmission / reception unit 13 receives a first answer, which is an output by the artificial intelligence 4 to which the input dataset has been input.

[0017] The answer generation unit 14 generates a second answer, which is an answer to the request, using the first answer received by the transmission / reception unit 13. The answer presentation unit 15 presents the second answer generated by the answer generation unit 14.

[0018] 2 is a diagram showing an example of a hardware configuration for realizing a terminal device 2A included in the information processing system 1A shown in FIG. 1. The terminal device 2A is realized by a computer system including a processing circuit 50, a communication device 51, an input device 54, and a display device 55. The processing circuit 50 includes a processor 52 and a memory 53. The processing circuit 50 is a circuit on which the processor 52 executes software. The processing functions of the request receiving unit 11, the input generating unit 12, the transmitting / receiving unit 13, the answer generating unit 14, and the answer presenting unit 15 are each realized by software, firmware, or a combination of software and firmware.

[0019] The software or firmware is written as a program and stored in memory 53. In processing circuit 50, processor 52 reads and executes the program stored in memory 53, thereby realizing the processing functions of terminal device 2A. That is, processing circuit 50 includes memory 53 for storing an information processing program, which is a program that results in the processing of terminal device 2A being executed. It can also be said that the information processing program stored in memory 53 causes a computer to execute the procedures and methods of terminal device 2A.

[0020] The processor 52 is a CPU (Central Processing Unit), a central processing unit, a processing unit, an arithmetic unit, a microprocessor, a microcomputer, a processor, or a DSP (Digital Signal Processor). The memory 53 is, for example, a non-volatile or volatile semiconductor memory such as a RAM (Random Access Memory), a ROM (Read Only Memory), a flash memory, an EPROM (Erasable Programmable Read Only Memory), or an EEPROM (Electrically Erasable Programmable Read Only Memory), a magnetic disk, a flexible disk, an optical disk, a compact disk, a minidisk, or a DVD (Digital Versatile Disc).

[0021] The communication device 51 communicates with devices external to the terminal device 2A. The communication function of the transmitter / receiver 13 is realized by the communication device 51. The communication device 51 communicates via Ethernet (registered trademark), a wireless LAN, Wi-Fi (registered trademark), or the like.

[0022] The input device 54 is a device that is operated by a user of the information processing system 1A and into which information is input. The input device 54 includes, for example, a keyboard, a mouse, a keypad, or a touch panel. The function of receiving information in the request receiving unit 11 is realized by the input device 54.

[0023] The display device 55 is a device that displays information. The display device 55 is, for example, an LCD (Liquid Crystal Display) or an organic EL (Electro-Luminescence) display. The function of displaying information in the answer presentation unit 15 is realized by the display device 55.

[0024] The hardware configuration realizing the terminal device 2A is not limited to the example shown here, and may be modified as appropriate. The terminal device 2A may include an audio input unit and an audio output unit. The audio input unit is, for example, a microphone. The audio output unit is, for example, a speaker. A typical example of the terminal device 2A is a personal computer. The terminal device 2A may also be a terminal such as a smartphone or a tablet. The terminal device 2A may also be a numerical control device, which is a control device for a machine tool.

[0025] 1, the components of the information processing system 1A are provided in a single device, that is, a terminal device 2A. The information processing system according to the first embodiment is not limited to a system in which the components of the information processing system are provided in a single device. The components of the information processing system may be distributed across two or more devices that can function together.

[0026] FIG. 3 is a diagram showing a second configuration example of the information processing system according to the first embodiment. FIG. 3 shows information processing system 1B, which is the second configuration example of the information processing system. Information processing system 1B is configured with a terminal device 2B and a back-end device 3B. That is, in information processing system 1B, the components of information processing system 1B are distributed between terminal device 2B and back-end device 3B. FIG. 3 shows terminal device 2B, back-end device 3B, and artificial intelligence 4 external to information processing system 1B.

[0027] The terminal device 2B includes a request receiving unit 11, a transmission / reception unit 13, and an answer presentation unit 15. The back-end device 3B includes an input generation unit 12, a transmission / reception unit 13, and an answer generation unit 14. In the information processing system 1B, the components that serve as an interface with the user are provided in the terminal device 2B, and the other components are provided in the back-end device 3B. In the information processing system 1B, the terminal device 2B and the back-end device 3B each include the transmission / reception unit 13.

[0028] The request receiving unit 11 of the terminal device 2B receives requests for work to be performed when using the machine tool, similar to the request receiving unit 11 of the terminal device 2A shown in Fig. 1. The transmitting / receiving unit 13 of the terminal device 2B communicates with the back-end device 3B. The terminal device 2B communicates with the back-end device 3B through the transmitting / receiving unit 13. The transmitting / receiving unit 13 of the terminal device 2B transmits information indicating the request received by the request receiving unit 11 to the back-end device 3B.

[0029] The transmitter / receiver 13 of the backend device 3B communicates with each of the terminal device 2B and the artificial intelligence 4. The transmitter / receiver 13 of the backend device 3B receives information indicating a request transmitted from the terminal device 2B. The input generation unit 12 of the backend device 3B generates an input data set corresponding to the request based on the information indicating the request received by the transmitter / receiver 13. The transmitter / receiver 13 of the backend device 3B transmits the input data set generated by the input generation unit 12 to the artificial intelligence 4. The transmitter / receiver 13 of the backend device 3B receives a first answer, which is an output by the artificial intelligence 4 to which the input data set has been input.

[0030] The answer generation unit 14 of the backend device 3B generates a second answer, which is an answer to the request, using the first answer received by the transmission / reception unit 13. The transmission / reception unit 13 of the backend device 3B transmits the second answer generated by the answer generation unit 14 to the terminal device 2B. The transmission / reception unit 13 of the terminal device 2B receives the second answer transmitted from the backend device 3B. The answer presentation unit 15 of the terminal device 2B presents the second answer received by the transmission / reception unit 13.

[0031] The transceiver 13 of the terminal device 2B and the transceiver 13 of the back-end device 3B are connected to each other so that they can communicate with each other using various known communication means. The communication means may be either wired communication or wireless communication. Ethernet may be used for wired communication. Wired communication may be either serial communication using a Universal Serial Bus (USB) or parallel communication. Wi-Fi or Bluetooth (registered trademark) may be used for wireless communication.

[0032] Among the components of the information processing system 1B, the components that serve as the interface with the user are provided in the terminal device 2B, and the other components are provided in the back-end device 3B, thereby making the terminal device 2B lightweight.

[0033] The terminal device 2B can be realized by a hardware configuration similar to that shown in FIG. 2. The hardware configuration realizing the terminal device 2B may include an audio input unit or an audio output unit. As described above, the terminal device 2B can be configured to be lightweight, and is therefore suitable for devices for which lightweight design is important, such as smartphones, tablets, and wearable devices. The terminal device 2B may also be a personal computer. The back-end device 3B can be realized by a configuration similar to that of the hardware configuration shown in FIG. 2, excluding the input device 54 and the display device 55. Examples of the back-end device 3B include a personal computer or a server device.

[0034] FIG. 4 is a diagram showing a third configuration example of an information processing system according to the first embodiment. FIG. 4 shows an information processing system 1C, which is the third configuration example of an information processing system. The information processing system 1C is configured with two terminal devices 2C1 and 2C2 and a back-end device 3C. That is, in the information processing system 1C, the components of the information processing system 1C are distributed among the two terminal devices 2C1 and 2C2 and the back-end device 3C. FIG. 4 shows the two terminal devices 2C1 and 2C2, the back-end device 3C, and an artificial intelligence 4 external to the information processing system 1C.

[0035] Terminal device 2C1 includes a request receiving unit 11 and a transmission / reception unit 13. Terminal device 2C2 includes a transmission / reception unit 13 and an answer presentation unit 15. Back-end device 3C includes an input generation unit 12, a transmission / reception unit 13, and an answer generation unit 14. Back-end device 3C has the same configuration as back-end device 3B shown in FIG. 3. In information processing system 1C, components that are interfaces with users are distributed to each of terminal devices 2C1 and 2C2, and components other than the interface are provided in back-end device 3C. In information processing system 1C, terminal device 2C1, terminal device 2C2, and back-end device 3C each include a transmission / reception unit 13.

[0036] The request receiving unit 11 of the terminal device 2C1 receives requests for work to be performed when using the machine tool, similar to the request receiving unit 11 of the terminal device 2A shown in Fig. 1. The transmitting / receiving unit 13 of the terminal device 2C1 communicates with the back-end device 3C. The terminal device 2C1 communicates with the back-end device 3C through the transmitting / receiving unit 13. The transmitting / receiving unit 13 of the terminal device 2C1 transmits information indicating the request received by the request receiving unit 11 to the back-end device 3C.

[0037] The transmission / reception unit 13 of the backend device 3C communicates with each of the terminal device 2C1, the terminal device 2C2, and the artificial intelligence 4. The transmission / reception unit 13 of the backend device 3C receives information indicating a request transmitted from the terminal device 2C1. The input generation unit 12 of the backend device 3C generates an input data set corresponding to the request based on the information indicating the request received by the transmission / reception unit 13. The transmission / reception unit 13 of the backend device 3C transmits the input data set generated by the input generation unit 12 to the artificial intelligence 4. The transmission / reception unit 13 of the backend device 3C receives a first answer which is an output by the artificial intelligence 4 to which the input data set has been input.

[0038] The answer generation unit 14 of the backend device 3C generates a second answer, which is an answer to the request, using the first answer received by the transmission / reception unit 13. The transmission / reception unit 13 of the backend device 3C transmits the second answer generated by the answer generation unit 14 to the terminal device 2C2. The transmission / reception unit 13 of the terminal device 2C2 receives the second answer transmitted from the backend device 3C. The answer presentation unit 15 of the terminal device 2C2 presents the second answer received by the transmission / reception unit 13.

[0039] The transmitting / receiving unit 13 of the terminal device 2C1, the transmitting / receiving unit 13 of the terminal device 2C2, and the transmitting / receiving unit 13 of the back-end device 3C are connected to each other so as to be able to communicate with each other by various known communication means. The communication means may be either wired communication or wireless communication, as in the case of the information processing system 1B shown in FIG.

[0040] Of the components of the information processing system 1C, the components that serve as an interface with users are provided in terminal devices 2C1 and 2C2, and the other components are provided in back-end device 3C. Furthermore, the request receiving unit 11 and answer presentation unit 15, which serve as interfaces with users, are separated into terminal device 2C1 and terminal device 2C2. Terminal device 2C1 can be configured to specialize in the function of receiving requests. Terminal device 2C2 can be configured to specialize in the function of presenting answers. This allows each terminal device 2C1 and 2C2 to be lightweight.

[0041] Each of the terminal devices 2C1 and 2C2 can be realized by a hardware configuration similar to the hardware configuration shown in FIG. 2 . The hardware configuration realizing each of the terminal devices 2C1 and 2C2 may include an audio input unit or an audio output unit. For example, the terminal device 2C1 may be a headset, and the terminal device 2C2 may be smart glasses. In this case, a request is input to the request receiving unit 11 by speech from a user wearing the headset. Furthermore, an answer is presented by the answer presenting unit 15 through a display on the smart glasses worn by the user. The user can input a request and confirm the answer to the request without using their hands. This allows the information processing system 1C to provide the user with extremely light and easy operability.

[0042] The manner in which the components of the information processing system are distributed is not limited to the second or third configuration example described above, and may be any other manner.

[0043] Next, an operation procedure of the information processing system according to the first embodiment will be described. Here, the operation procedure of the information processing system 1A will be described using the information processing system 1A shown in Fig. 1 as an example. Fig. 5 is a flowchart showing an example of the operation procedure of the information processing system according to the first embodiment.

[0044] In step S1, the request receiving unit 11 receives a request for work to be performed when using the machine tool. In step S2, the input generating unit 12 generates an input data set corresponding to the request based on the request received in step S1. In step S3, the transmitting / receiving unit 13 transmits the input data set generated in step S2 to the artificial intelligence 4. The artificial intelligence 4 outputs a first answer in response to the input data set.

[0045] In step S4, the transmitter / receiver 13 receives the first answer output by the artificial intelligence 4. In step S5, the answer generator 14 generates a second answer using the first answer received in step S4. In step S6, the answer presenter 15 presents the second answer generated in step S5. With this, the information processing system 1A completes the operation according to the procedure shown in FIG.

[0046] In the information processing system 1A, the terminal device 2A performs operations according to the procedure from step S1 to step S6. In the information processing system 1B shown in Fig. 3, the terminal device 2B and the back-end device 3B perform operations similar to those according to the procedure shown in Fig. 5. In the information processing system 1C shown in Fig. 4, the two terminal devices 2C1 and 2C2 and the back-end device 3C perform operations similar to those according to the procedure shown in Fig. 5.

[0047] Next, each component of the information processing system according to embodiment 1 will be described. Below, each component of the information processing system 1A will be described using the information processing system 1A shown in Fig. 1 as an example. Note that the description of each component of the information processing system 1A also applies to each component of the information processing systems 1B and 1C.

[0048] <Request Receiving Unit 11> The request receiving unit 11 receives requests for work performed when using a machine tool. In the first embodiment, the machine tool includes a cutting machine, a grinding machine, an electric discharge machine, a laser machine, an additive manufacturing (AM) machine, or the like. There are also no particular limitations on the configuration of each of these machines. For example, the cutting machine may be a machining center, a lathe, a multi-tasking machine, or the like. Furthermore, work performed when using a press machine, an injection molding machine, an industrial robot, or the like may be the subject of a request.

[0049] The use of such a wide variety of industrial machines requires specialized knowledge of the machines themselves, or of the controllers or peripheral devices that control the machines. The requirements in the first embodiment cover all work that is performed based on such specialized knowledge.

[0050] More specifically, a request received by the request receiving unit 11 is represented by information representing an operation to be performed when using the machine tool. The information includes a pair of verb information representing at least one action and object information representing one or more objects that are the targets of the action. Note that a request may combine one piece of verb information with multiple pieces of object information. A request may also include multiple pieces of verb information.

[0051] For example, in a request to "create a ladder program that does ...," "create" is the verb information, and "ladder program that does ..." is the object information. Also, in a request to "What is the parameter that sets the maximum speed?", it can be considered that "what?" has been omitted after "is." In this case, "what is?" is the verb information, and the object information corresponding to this verb information is "the parameter that sets the maximum speed." Alternatively, "what is ...?" can be read as "answer ...," or "answer ..." can be used as the verb information.

[0052] There are no particular limitations on the format of the request. The request may be written in natural language using text data, i.e., character strings, or may be written using voice information. The request may also be written using information indicating a button operation or information indicating a screen operation. In this case, the content of the button or screen operation may be associated in advance with the content of the request, so that the content of the request can be determined based on the operated button or the content of the screen operation. In this way, the request receiving unit 11 may be able to understand the content of the request and convert the input request into request information.

[0053] When a request is composed of audio information, more specifically, audio data obtained from a sound collection device such as a microphone is converted into text information by a transcription process. When a request is received by a button operation or a screen operation, the content of the request indicated by the button or operation is stored in a database or table in advance, and the content of the request corresponding to the button operation or screen operation is read from the database or table to understand the content of the request.

[0054] FIG. 6 is a first diagram showing an example of a screen displayed by the request receiving unit 11 included in the information processing system according to the first embodiment. FIG. 7 is a second diagram showing an example of a screen displayed by the request receiving unit 11 included in the information processing system according to the first embodiment. The button labeled "AI Assist" at the bottom right of the screen shown in FIG. 6 is a button that accepts an operation to start the request receiving unit 11. When the "AI Assist" button is pressed as shown in FIG. 6, a pop-up for inputting a request is displayed on the screen as shown in FIG. 7. Note that the handprint shown in FIG. 6 indicates that the button is being pressed.

[0055] The area for inputting a request may be displayed on the screen at all times. However, as shown in Figures 6 and 7, the area for inputting a request may be displayed as a pop-up as needed, thereby improving the operability of the screen when not accepting a request. Furthermore, it becomes easier to detect the operation content of the request accepting unit 11 from the button operation history or the screen operation history. In other words, by using the operation of pressing the "AI Assist" button as a trigger, it becomes easier to determine that subsequent operations are operations related to the request.

[0056] As shown in FIG. 7 , the pop-up displays a system-side speech bubble that shows the content of the conversation between the information processing system 1A and the user. In the example shown in FIG. 7 , the speech bubble displays the character string "Please select your request from the list below," which is a message that prompts the user to input a request. The pop-up shown in FIG. 7 displays a button for accepting questions for the information processing system 1A and a button for accepting work requests to the information processing system 1A. When the user presses the "Question" button, the request accepting unit 11 accepts the question. When the user presses the "Request" button, the request accepting unit 11 accepts the work request.

[0057] FIG. 8 is a third diagram showing an example of a screen displayed by the request receiving unit 11 included in the information processing system according to the first embodiment. FIG. 8 shows an example of a pop-up that is displayed when a request consisting of text data is received. The pop-up shown in FIG. 8 is displayed when the "AI Assist" button is pressed from the state shown in FIG. 6. The pop-up shown in FIG. 8 is an example of a pop-up that is displayed when a request is received in chat format from the beginning when the request receiving unit 11 is started. In the pop-up shown in FIG. 8, a field for inputting text data is displayed in the speech bubble on the user's side.

[0058] Fig. 9 is a fourth diagram showing an example of a screen displayed by the request receiving unit 11 included in the information processing system according to embodiment 1. Fig. 9 shows an example of a pop-up that is displayed when a request is input by voice. The pop-up shown in Fig. 9 is displayed when the "AI Assist" button is pressed in the state shown in Fig. 6.

[0059] 9, a button with a microphone graphic is displayed in the pop-up. When the button is pressed, the request receiving unit 11 enters a state in which it is able to receive voice. When the button is pressed again after the request receiving unit 11 enters a state in which it is able to receive voice, the request receiving unit 11 stops receiving voice. This allows the request receiving unit 11 to prevent malfunction due to unintended voice input.

[0060] Alternatively, by changing the display of the graphic representing the microphone only when sound is detected, it is possible to easily confirm whether sound is being detected by visually checking the graphic. Note that the manner in which the display of the graphic representing the microphone is changed is arbitrary. For example, the manner in which the display of the graphic is changed may be changed by changing the brightness or color of the graphic.

[0061] 9, a character string indicating the transcription result of the input voice is displayed in a balloon on the user's side, allowing the user to confirm whether the content of the spoken request has been correctly accepted.

[0062] Fig. 10 is a diagram showing a first example of a transition of a screen displayed by the request receiving unit 11 included in the information processing system according to the first embodiment. Fig. 10 shows an example of a transition of a screen when the "Question" button is pressed in the pop-up shown in Fig. 7. When the "Question" button is pressed as shown in the upper part of Fig. 10, the display of the pop-up transitions to the display shown in the middle part of Fig. 10. As shown in the middle part of Fig. 10, the character string "Please select your question from the list below", which is a message prompting the user to select the content of a question, is displayed in the system-side balloon part of the pop-up.

[0063] Also, several buttons representing options for the question, such as "Regarding machine tools" or "Regarding numerical control," are displayed below the balloon shown in Fig. 10. When the "Regarding numerical control" button is pressed in the state shown in the middle of Fig. 10, the pop-up display transitions to the display shown in the lower part of Fig. 10.

[0064] As shown in the lower part of Fig. 10, the system-side bubble in the pop-up displays the following message, "This is about numerical control, right? Please enter your question." This message prompts the user to enter a question. The user-side bubble displays a field for entering text data indicating the question. This allows the request receiving unit 11 to understand that a question about numerical control has been entered as a request, and that a question about numerical control has been entered as text data.

[0065] Fig. 11 is a diagram showing a second example of the transition of a screen displayed by the request receiving unit 11 included in the information processing system according to the first embodiment. Fig. 11 shows an example of the transition of a screen when the "Request" button is pressed in the pop-up shown in Fig. 7. When the "Request" button is pressed as shown in the upper part of Fig. 11, the display of the pop-up transitions to the display shown in the middle part of Fig. 11. As shown in the middle part of Fig. 11, the character string "Please select the content of your request from the list below" is displayed in the balloon on the system side in the pop-up, which is a message prompting the user to select the content of the request.

[0066] Additionally, several buttons representing options for the request content, such as "Create a program" or "Create and output data," are displayed below the balloon shown in Fig. 11. When the "Create a program" button is pressed in the state shown in the middle of Fig. 11, the pop-up display transitions to the display shown in the lower part of Fig. 11.

[0067] As shown in the lower part of Figure 11, the system-side bubble in the pop-up displays the text "You want to create a program. Please select a target," which is a message prompting the user to select a program to be created. The pop-up also displays several buttons indicating options for the program to be created. In the example shown in the lower part of Figure 11, a "Ladder Program" button and a "Machining Program" button are displayed. When either the "Ladder Program" button or the "Machining Program" button is pressed, the request receiving unit 11 can understand that a request has been made to create the program selected by pressing the button.

[0068] Fig. 12 is a diagram showing a third example of the transition of a screen displayed by the request receiving unit 11 included in the information processing system according to the first embodiment. Fig. 12 shows an example of the transition of a screen when the "Request" button is pressed in the pop-up shown in Fig. 7. The pop-up display shown in the middle section of Fig. 12 is the same as the pop-up display shown in the middle section of Fig. 11. In Fig. 12, it is assumed that the "Create / Output Data" button is pressed in the state shown in the middle section of Fig. 12. When the "Create / Output Data" button is pressed, the pop-up display transitions to the display shown in the lower section of Fig. 12.

[0069] As shown in the lower part of Figure 12, the system-side speech bubble in the pop-up displays the text "Data creation / output, right? Please select the target," which is a message prompting the user to select the data to be created or output. The pop-up also displays several buttons indicating the options for the data to be created or output. In the example shown in the lower part of Figure 12, the buttons for "machining program," "tool model," "work procedure manual," and "workpiece model" are displayed. When one of these buttons is pressed, the request receiving unit 11 can understand that a request has been made to create or output the data selected by pressing the button.

[0070] Inputting a request in natural language through chat or voice allows the request receiving unit 11 to grasp the details of the request with greater versatility. However, for example, when a request is made to create a machining program, the request receiving unit 11 can display a list of function names to allow selection, or select from a list of requests previously input, thereby making it possible to receive detailed request content through screen or button operations only.

[0071] The request receiving unit 11 converts the received request into data called request information and passes it to the input generating unit 12. The request information is data that makes at least the content of the received request available to the input generating unit 12. The request information includes one or more pieces of verb information and one or more pieces of object information that are paired with the verb information.

[0072] The request information is information extracted from a request expressed in the various expression formats described above. The format of the request information is not particularly limited. The request information may be in the form of a simple sentence in which words extracted from the request are simply arranged in the order in which they were extracted. Alternatively, the request information may be in a structured format in which words are arranged according to a certain rule. The certain rule may be, for example, clarifying the relationship between a verb and an object, or a relationship such as parallelism or contrast, through the order of words. Alternatively, the request information may simply hold a character string indicating the received request.

[0073] As described above, the request receiving unit 11 creates request information by extracting verb information and object information from the received request. The request receiving unit 11 passes the created request information to the input generating unit 12.

[0074] <Input Generation Unit 12> The input generation unit 12 generates an input dataset that is an input to the artificial intelligence 4, based on the request information received from the request reception unit 11. The input generation unit 12 generates an input dataset according to the request indicated in the request information by referring to at least one of a specification database in which information representing the specifications of the information processing system 1A or the specifications of a device external to the information processing system 1A is stored, setting parameters related to the settings of the information processing system 1A or the settings of a device external to the information processing system 1A, and retained data retained in a storage area accessible by the information processing system 1A.

[0075] The input generation unit 12 is provided with request information expressed in natural language. For this reason, the request information input to the input generation unit 12 may contain problems that hinder obtaining the desired answer, such as variations in expression, missing or insufficient necessary information, or the inclusion of errors. Artificial intelligence 4 that inputs and outputs natural language often works by predicting which words are likely to appear before or after a certain word and then creating an output in response to the input. Language processing is not performed by accurately reading the context, etc. For this reason, the above-mentioned problems are expected to reduce the stability and reproducibility of the output in response to the input, significantly affecting the quality of the output obtained from the artificial intelligence 4.

[0076] In the first embodiment, in order to suppress the adverse effects of such problems, an input data set is generated by the input generation unit 12. The generation of the input data set by the input generation unit 12 will be described below using an example in which a machining program is requested to be created.

[0077] The elements that make up a machining program include G-codes or M-codes that specify the functions of the numerical control device, S-codes or F-codes that specify various speeds, axis commands that specify the axes to be operated or the coordinates of the target position, and T-codes that specify the tool numbers to be used. Other elements that make up a machining program include information that specifies the path number or argument information corresponding to the G-code to be used. To create a machining program as intended, the correct information for these elements must be selected and specified in the correct order according to the specifications. However, this information varies depending on the manufacturer of the numerical control device, and even within the same manufacturer, it may vary depending on the series or settings. For this reason, it is not easy to obtain the correct output from a request input expressed in natural language.

[0078] Therefore, the input generating unit 12 solves this problem by referring to a specification database, setting parameters, and stored data. Note that, hereinafter, the specification database will be referred to as a "specification DB."

[0079] <Specification DB> The specification DB referenced by the input generation unit 12 is a database that stores information about the specifications of the information processing system 1A and information about the specifications of external devices or external equipment that are the targets of information processing by the information processing system 1A. An external device or external equipment is, for example, a machine tool or an artificial intelligence 4. Information can be added and deleted from the specification DB at any time. However, the content of the information stored in the specification DB is not frequently changed by the user of the information processing system 1A.

[0080] The specification DB stores, for example, specification information or manual information. The specification information or manual information includes, for example, information showing the basic usage of various functions, information about errors or alarms, or definitions and explanations of terms. More specifically, the specification DB may store, for example, a table in which functions and codes of G codes or M codes, which are commands in a machining program for a machine tool, are associated with each other by machine tool manufacturer, by numerical control device manufacturer, by numerical control device series, or by parameter setting.

[0081] The specification DB stores information about the specifications of each numerical control device by manufacturer or series. For example, the information stored in the specification DB includes information in a table or database that describes the correspondence between the components and functions of the above-mentioned components of the machining program. The information stored in the specification DB may also include examples of commands in the machining program for each function.

[0082] <Setting Parameters> The setting parameters referenced by the input generation unit 12 are parameters related to the settings of the information processing system 1A, or parameters related to the settings of an external device or external equipment that is the target of information processing by the information processing system 1A. An external device or external equipment is, for example, a machine tool or an artificial intelligence 4. One example of setting parameters is a parameter used when controlling a machine tool. The setting parameters include one or more of the type of machine tool, information indicating the manufacturer of the machine tool, and information indicating the manufacturer of the machine tool's control device. In the information processing system 1A, the specification DB to be referenced is switched depending on the set value of the setting parameter.

[0083] For example, the setting parameters may include information indicating the type of machine tool, such as a machining center, lathe, laser processing machine, or electric discharge machine. The information indicating the type of machine tool is used to switch the command format, i.e., the command format, in function commands such as G-code or M-code. This is because different functions are used for different types of machine tools, or the importance of the functions used for different types of machine tools varies, resulting in different command formats for different types of machine tools.

[0084] The setting parameters may also include information about the axes of the machine tool to be controlled, more specifically, the names of the axes or the number of axes. The axis command is a combination of an alphabetic character specifying the axis and the numerical coordinates of the target position. For example, if the target position for the X axis is 100.0 mm, the axis command would be "X100.0." Correctly creating such an axis command requires information about what axes the machine tool has and in what direction the axes are attached. Information about what axes the machine tool has can also be said to indicate what axes are not provided, i.e., what axes should not be the subject of commands.

[0085] <Retained Data> The retained data referenced by the input generation unit 12 is data stored in a storage area accessible to the information processing system 1A. Retained data is data that is not only added or deleted but also changed more frequently than the data in the specification DB. Examples of retained data include machining programs, ladder programs, tool data, work data, work instructions, and internal data (coordinate values, variable values, or state variables).

[0086] For example, suppose the retained data contains values ​​indicating the current coordinates of each axis. The current coordinate values ​​of each axis can be broadly divided into command values ​​and feedback values. Command values ​​are values ​​that indicate the target position created by the control software. Feedback values ​​are values ​​that are obtained when the motor actually operates in response to a command, and information from the encoder or sensor attached to the motor is fed back to the control system.

[0087] The input generation unit 12 can grasp the current state of the machine tool by referring to this information indicating the current coordinates of each axis. This information is necessary when the user implicitly desires to create a machining program for operations "from the current position."

[0088] Alternatively, the input generator 12 can use this information indicating the current coordinates of each axis when determining whether a user request is appropriate in light of the current position. For example, if the current position along a certain axis is near the end of the range of movement, a request indicating movement to a position further ahead may be determined to be inappropriate.

[0089] <Input Data Set> An input data set is information that is directly input to the artificial intelligence 4, and is information that is composed of one piece of data or a combination of two or more pieces of data. An input data set is mainly composed of two types of data: a prompt and a data file. An input data set may be information that consists of only one of a prompt and a data file, or a combination of a prompt and a data file. An input data set may also be a combination of multiple prompts and multiple data files.

[0090] <Prompt> In generative AI, such as ChatGPT (Chat Generative Pre-trained Transformer), or large language models (LLMs) used in generative AI, prompts are used as input to the AI. Prompts are written in natural language. Prompts include questions or requests for the AI.

[0091] <Data files> The data files included in the input data set include one or more of a machining program, a ladder program, setting parameters, a machine tool operation procedure manual, shape data for each of the tools and jigs attached to the machine tool, and shape data for the workpiece to be machined by the machine tool.

[0092] Large-scale language models exist that can handle data files. For example, ChatGPT, using a plug-in called Code Interpreter, can handle not only text files but also files written in programming language formats such as Python, JavaScript (registered trademark), and HTML, as well as MS Office files and image files such as JPG and PNG. Furthermore, AI that generates images or three-dimensional data (hereinafter referred to as "3D data") can generate 3D data from text data or two-dimensional image data. Examples of such AI include NVIDIA's GET3D, Stable Dreamfusion, and Text-to-CAD. Including data files in the input dataset allows for effective utilization of such artificial intelligence. Note that in the first embodiment, there are no particular limitations on the format or size of the data file. The format and size of the data file can be selected according to the desired artificial intelligence.

[0093] <Function of the Input Generator 12> As a basic process, the input generator 12 adds additional sentences to the prompt to improve the accuracy of the answer. In artificial intelligence that uses natural language as input and output as described above, it is known that the accuracy of the output changes depending on the prompt, and methods for improving the accuracy of the output are known empirically.

[0094] For example, by clarifying the role of the artificial intelligence 4, that is, by clarifying the position of the artificial intelligence 4 that provides the answer, the accuracy of the output of the artificial intelligence 4 can be improved. Specifically, adding content such as "You are an experienced machine tool operator" to the prompt can be expected to be generally effective when giving the artificial intelligence 4 requests necessary for using the machine tool.

[0095] As another example, a study called EmotionPrompt (Large Language Models Understand and Can Be Enhanced by Emotional Stimuli, Cheng Li et al.) found that adding emotional expressions to prompts improves the accuracy of LLM output. Specific examples of emotional expressions include "This is very important to my career" or "Take pride in your work and give it your best." Adding such statements to prompts that elicit a sense of responsibility and self-efficacy from an emotional perspective is a good example.

[0096] Next, the input generation unit 12 refers to the setting parameters and adds to the prompt an explanation of the situation in which the information processing system 1A or an external device or equipment (e.g., a machine tool or the artificial intelligence 4) is placed. By inputting detailed background knowledge or situation settings to the artificial intelligence 4 through the prompt, it is expected that the probability of an irrelevant answer being output will be reduced. For example, it is conceivable to add information indicating the type of machine tool in question to the prompt.

[0097] The input generation unit 12 then extracts relevant specific examples from the information stored in the specification DB and adds the specific examples to the prompt. By adding reference specific examples to the prompt for the content for which the artificial intelligence 4 is requested to respond, it is expected that the accuracy of the response will improve. In particular, if the content of the request is content that is expected to be in a predetermined form or format, including a specific example written in that form or format in the prompt increases the likelihood that the response from the artificial intelligence 4 will be in accordance with the expected form or format.

[0098] For example, if the request is for the creation of a machining program, it is conceivable to include in the prompt a specific program example that conforms to the format of the machining program. In this case, the input generation unit 12 refers to the specification DB and describes in the prompt a program example that conforms to the functional specifications. This is expected to prevent an answer that violates the functional specifications from being output.

[0099] Then, the input generation unit 12 selects data files to be included in the input dataset from the stored data. As described above, when utilizing the artificial intelligence 4 that can handle data files, the input generation unit 12 selects data to be included in the input dataset from the stored data. The data selection is performed according to the content of the request. For example, if part of the work content is changed in an existing work manual, when creating a work manual that reflects the change, the existing work manual that serves as the basis is selected.

[0100] In this way, the input generating unit 12 generates an input data set corresponding to the request based on the request information.

[0101] <Transmission / reception unit 13> The transmission / reception unit 13 transmits the input data set generated by the input generation unit 12 to the artificial intelligence 4. The transmission / reception unit 13 receives a first answer, which is an output by the artificial intelligence 4 to which the input data set has been input. Note that in a configuration without the input generation unit 12, the transmission / reception unit 13 transmits the request information to the artificial intelligence 4 as is.

[0102] The first response is a response corresponding to the request and is data including at least text information written in a natural language or a data file configured in a specific form or format. The first response may include both text information and a data file.

[0103] Transmission and reception by the transmitter / receiver 13 may be achieved using any known method as long as it is possible to input the necessary information to the artificial intelligence 4 and obtain the output from the artificial intelligence 4. Transmission and reception may be achieved, for example, by a wired electrical communication connection such as Ethernet, or by a wireless communication connection such as Wi-Fi or Bluetooth. Furthermore, in a configuration in which the artificial intelligence 4 is configured on the same hardware as the information processing system 1A, transmission and reception may be achieved by software-based data input / output processing via a storage unit such as a memory shared by the information processing system 1A and the artificial intelligence 4.

[0104] In the first embodiment, there is no restriction on where the artificial intelligence 4 is configured. As long as the transmission / reception unit 13 can transmit an input data set to the artificial intelligence 4 and the transmission / reception unit 13 can receive the output of the artificial intelligence 4, the artificial intelligence 4 may be configured inside the information processing system 1A or outside the information processing system 1A.

[0105] Furthermore, in the first embodiment, various forms of artificial intelligence can be used as the artificial intelligence 4. The artificial intelligence 4 may be configured using an algorithm such as Transformer, BERT (Bidirectional Encoder Representations from Transformers), or GPT, or may be configured by combining a plurality of algorithms including these. Furthermore, in the first embodiment, the information processing system 1A can also utilize various artificial intelligence services external to the information processing system 1A. For example, the information processing system 1A may utilize ChatGPT, Bing, or Bard, or may use a smaller language model, and may further utilize these by adding additional learning.

[0106] <Answer Generation Unit 14> The answer generation unit 14 generates a second answer, which is an answer to the request, using the first answer received by the transmission / reception unit 13. When generating the second answer, the answer generation unit 14 performs processes of verifying basic rules, verifying functional specifications, verifying operation, and converting the second answer into an answer format when presenting the second answer. The answer generation unit 14 does not need to perform all of these processes, but rather performs any one or a combination of two or more of them. Alternatively, the answer generation unit 14 may perform all of these processes.

[0107] <Verification of Basic Rules> The answer generation unit 14 verifies the formal aspects of the first answer as verification of the basic rules. Here, the formal verification refers to mechanically checking whether the superficial aspects of the first answer conform to certain rules, regardless of the content of the first answer.

[0108] For example, for machining programs, checking basic grammar corresponds to verifying the basic rules. If the machining program is in EIA / ISO format, a line feed code (End Of Block: EOB) must be inserted at the end of each block, often represented by ";". In addition, an End Of Record (EOR), usually represented by "%", must be inserted at the beginning and end of the program. Furthermore, only alphanumeric characters may be used in programs; double-byte characters found in some languages, or characters from other languages, are not accepted. These characters are permitted when they are written as comments that serve as notes rather than as parts to be executed, i.e., when they are enclosed in symbols indicating comment areas such as "()".

[0109] For data files, checking the data format corresponds to the basic rule verification. There are various data formats, but they are generally identified by the extension or by the information contained in the header area at the beginning of the data. It is verified whether the data format identified by these methods matches the requested data type.

[0110] <Verification of Functional Specifications> While the verification of the basic rules described above was a verification of the formal aspects, the verification of functional specifications involves more specifically verifying the content of the first answer. The first answer includes the functions of the information processing system 1A and content related to various functions provided by external devices or external equipment (e.g., machine tools) that are the targets of information processing by the information processing system 1A. Verification of functional specifications involves verifying whether this content is correct in light of the specification information of the functions. The answer generation unit 14 may verify the functional specifications by referring to a specification DB.

[0111] For example, for machining programs, it is verified whether there are any problems with the usability of the commands included in the machining programs, in light of the usage environment, such as parameter settings, etc. For ladder programs, it is verified whether there are any problems with the usability of the commands or devices used in the ladder programs.

[0112] For each of the verification of basic rules and the verification of functional specifications, items that violate the rules may be kept in list form as check items, and such a list of prohibited items (hereinafter referred to as an "NG list") may be used for verification. Specifically, a method of detecting a specific character string or a specific type of data file may be considered. In addition, items on this NG list may be added, deleted, or changed as appropriate or sequentially. Furthermore, the list that serves as the NG list may be added, deleted, or changed as appropriate or sequentially.

[0113] Here, we will explain a more specific example of functional specification verification. AI can encounter a problem known as hallucination, where the AI ​​responds with incorrect or inaccurate information as if it were correct. One method that takes advantage of this problem is to train the AI ​​to learn malicious code and have it output answers that include the malicious code, which has been reported to be capable of executing the malicious code (Diving Deeper into AI Package Hallucinations (Lasso Security)).

[0114] Using the NG list, it is possible to detect information about known malicious code that has been made public as described above and prevent the execution of the malicious code, and by updating the NG list when new information is made public, it is possible to respond quickly.It is also possible to prevent the risk of the AI ​​4 responding with information that should not be made public to the user or operations that should be prohibited from a safety perspective.

[0115] <Verification of Operation> In the verification of operation, among the contents included in the first answer, the contents that can be executed as software are operated either actually or virtually using a simulation, etc., and it is verified whether there are any problems with the contents. For example, for machining programs, many products are equipped with a program check or simulation function, and operation can be verified by using these.

[0116] <Conversion to Answer Format When Presenting a Second Answer> The purpose of converting to an answer format when presenting a second answer is to make the format of the second answer more suitable for the user by changing the format of expression while leaving the content of the information contained in the first answer unchanged. While it is possible to present the first answer as the answer as is, depending on the user's desired situation, the user may be required to perform additional tasks such as editing, duplicating, or fine-tuning. Therefore, here, the answer generation unit 14 generates the second answer by changing only the format of expression of the first answer. The answer generation unit 14 ultimately converts the format into one suitable for presenting the answer to the user.

[0117] Here, conversion of the expression format refers to changing the display method, such as the screen position, size, or color, when presenting an answer on the answer presentation unit 15; visual representation methods, such as whether to display a list of character strings, display in a table format, or display in figures or 3D graphics; and electronic representation methods, such as the data format when outputting as text data or a data file. Furthermore, conversion of the expression format may also include changing the communication format, such as whether to communicate the answer visually or audibly, i.e., by voice.

[0118] Alternatively, the display position of the answer may be specified by coordinates on the screen. The display position of the answer may be specified, for example, at the current position of the cursor, at the position of a specific window, or after a character string that is a specific keyword. For example, for a machining program, the display position of the answer may be specified by selecting one of the following: outputting the created program as a program file, inputting the created program at a predetermined position on the screen like an MDI (Manual Data Input) window, or inputting the created program after the current position of the cursor.

[0119] Fig. 13 is a first diagram for explaining the conversion of the answer format by the answer generation unit 14 of the information processing system according to the first embodiment. Fig. 14 is a second diagram for explaining the conversion of the answer format by the answer generation unit 14 of the information processing system according to the first embodiment. Fig. 13 shows a state when a request to create a machining program for a tool center point control command is made by operating a pop-up on the screen. Fig. 14 shows a state in which the machining program created in response to such a request is input following the current position of the cursor.

[0120] Fig. 15 is a third diagram for explaining the conversion of the answer format by the answer generation unit 14 of the information processing system according to the first embodiment. Fig. 16 is a fourth diagram for explaining the conversion of the answer format by the answer generation unit 14 of the information processing system according to the first embodiment. Fig. 15 shows a state when a request to create a continuation of a machining program currently being created is made by operating a pop-up on the screen. Fig. 16 shows a state when a machining program created in response to such a request is input as a continuation of a machining program that has been created up to that point.

[0121] Furthermore, if a problem is found in each verification performed by the answer generation unit 14, the answer generation unit 14 may stop the presentation of answers by the answer presentation unit 15. In this case, the information processing system 1A can achieve the effect of suppressing the presentation of answers with quality problems. Alternatively, the answer generation unit 14 may cause the answer presentation unit 15 to present an answer and may also alert the user by notifying the user that a problem was found in the verification as a verification result. The answer generation unit 14 may also present a second answer by adding information about the verification result to the first answer and displaying it.

[0122] Alternatively, the answer generation unit 14 may delete only the part of the answer that was found to be problematic in the verification, and use the remaining part as the second answer. For example, when the above-mentioned NG list is used, by deleting only the information that should not be presented as an answer, the information processing system 1A can present as an answer parts that have no quality problems, while avoiding the risk of presenting parts that have problems.

[0123] Furthermore, the answer generation unit 14 may be able to distinguish the portions of the processing program created by the artificial intelligence 4. Since it is difficult to guarantee 100% reproducibility or quality of the output of the artificial intelligence 4, there may be a desire to check the portions created by the artificial intelligence 4 before the processing program is finally used. By making it possible to distinguish the portions created by the artificial intelligence 4, it becomes clear which parts need to be checked, and it is possible to improve the explainability and transparency of the system using the artificial intelligence 4.

[0124] The answer generation unit 14 converts the display to make it possible to distinguish the part created by the artificial intelligence 4. In other words, the conversion to the answer format showing the second answer includes a conversion to make it possible to distinguish the part created by the artificial intelligence 4.

[0125] The manner of display conversion to make the portion created by the artificial intelligence 4 distinguishable is arbitrary. Specific examples include, for character strings, changing the color or thickness of the characters, highlighting the relevant portion, surrounding the relevant portion with a frame, or underlining the relevant portion. For data, it is possible to embed information indicating that the data was created by the artificial intelligence 4 as metadata, or to clearly indicate in the file name of the data that it was created by the artificial intelligence 4. The information processing system 1A can clearly indicate to the user the portion of the answer presented to the user that was created by the artificial intelligence 4, by displaying the portion created by the artificial intelligence 4 in a manner that makes it possible to distinguish the portion.

[0126] Furthermore, if the part created by the artificial intelligence 4 includes a numerical value, the answer generation unit 14 may further convert the numerical value.

[0127] The language model-based AI 4 predicts what words are likely to follow a given word by learning based on the associations between words. In this case, since numerical information is not as dependent on the context as words, it is known that the accuracy of the numerical information output by the generation AI is lower than that of words. For this reason, it can be said that the numerical information output by the AI ​​4 needs to be carefully checked by the user.

[0128] To encourage the user to check more carefully, the answer generation unit 14 may, for example, convert the numerical information created by the artificial intelligence 4 so that it stands out more than other parts. Specific examples include making the characters of the numerical information stand out by combining various display modes, such as making the characters thicker, highlighting the characters and then making them thicker, or using a color complementary to the color of the highlighter as the character color.

[0129] Alternatively, when a numerical value is included in the portion created by the artificial intelligence 4, the answer generation unit 14 may replace the numerical value with a specific symbol or specific character, or may delete the numerical value. In other words, when a numerical value is included in the portion created by the artificial intelligence 4, conversion to an answer format indicating a second answer may include replacing the numerical value with a specific symbol or specific character, or deleting the numerical value.

[0130] For example, simply deleting a numerical value included in a portion created by the artificial intelligence 4 may result in the deleted numerical value being overlooked. In this case, the answer generation unit 14 replaces the numerical value with a specific symbol or specific character, making it easier for the user to recognize that confirmation and correction by the user is required. Note that such conversion is not necessarily performed uniformly for all numerical values ​​included in a portion created by the artificial intelligence 4. The answer generation unit 14 may selectively convert only numerical values ​​that meet specific conditions by following specific rules.

[0131] Here, a specific example of the conversion of numerical values ​​included in the portion created by the artificial intelligence 4 will be described. For example, suppose the artificial intelligence 4 outputs a machining program such as "G90 G01 X100.0 F1000;." In such a program, the numerical value following "G" is a numerical value specifying a G-code and can be said to have a one-to-one correspondence with a function. Therefore, the numerical value following "G" in the program output by the artificial intelligence 4 can be said to be highly reliable among the numerical information. The numerical value specifying an M-code, i.e., the numerical value following "M," can also be said to be highly reliable, just like the numerical value following "G." In contrast, the numerical value following "X" and the numerical value following "F" represent the axis position and movement speed, respectively. The numerical value following "X" and the numerical value following "F" can each take any value, making accurate prediction difficult. Therefore, the numerical value following "X" and the numerical value following "F" in the program output by the artificial intelligence 4 can be said to be less reliable than the numerical value of the G-code or the numerical value of the M-code.

[0132] Therefore, the answer generation unit 14 converts, for example, a program such as "G90 G01 X100.0 F1000;" into "G90 G01 X F0;". In the converted program, the number following the "X" has been deleted. The user then modifies this converted program by adding the correct number for the X code.

[0133] In this way, because the numerical value of the X code is deleted from the converted program, even if the converted program is executed as is, the execution of the program will be interrupted due to an error. If an invalid value is written as the numerical value of the X code, the invalid value may be overlooked, causing the machine tool to attempt to move to an unexpected position, which may cause problems. By performing a conversion that deletes the numerical value of the X code using the answer generation unit 14, it is possible to prevent such problems from occurring.

[0134] Furthermore, in the converted program, the unusual symbol "o" is inserted in place of the number following "F." The user modifies this converted program by deleting the "o" and adding the correct numerical value of the F code. The symbol inserted in place of the numerical value may be a specific symbol that is not normally used in programs, and may be a symbol other than "o." Alternatively, a specific letter may be inserted in place of the numerical value instead of a symbol. By inserting a specific symbol or letter in place of the numerical value of the F code through conversion in the answer generation unit 14, the user can easily recognize that the symbol or letter needs to be replaced with the numerical value of the F code.

[0135] Alternatively, the information processing system 1A may automatically verify, correct, or modify the numerical portion output by the artificial intelligence 4. For example, when verifying the operation of a machine tool by executing a program, if a problem occurs with the position or speed in the direction of a certain axis, the information processing system 1A may use the verification results of the operation to modify the numerical portion output by the artificial intelligence 4.

[0136] It is not necessary for all of the above processes to be performed by the answer generation unit 14; it is sufficient to perform at least one of them. Two or more of the above processes may be performed in combination, or all of the above processes may be performed. Depending on the content of each process performed, it is possible to obtain the effect of improving the accuracy of the second answer through verification, and the effect of improving user convenience by converting the answer format.

[0137] Furthermore, for example, the answer generation unit 14 may perform only basic rule verification and use the first answer without changing the format as the second answer, or the answer generation unit 14 may perform neither verification nor change the format of the first answer as the second answer.

[0138] The above processes in the answer generation unit 14 may be performed in any order. Furthermore, at least one of the above processes may be performed multiple times. For example, the answer generation unit 14 may format the answer by converting the format, then perform each verification process, and then convert the format again. For example, when creating a processing program using the artificial intelligence 4, explanatory text other than the main body of the processing program may be added to the created processing program. In such cases, if verification is performed while leaving such explanatory text, it may be deemed inappropriate. The answer generation unit 14 can avoid such problems by extracting only the parts to be verified in advance.

[0139] The answer generator 14 outputs the second answer generated as described above to the answer presenter 15 .

[0140] <Answer presentation unit 15> The answer presentation unit 15 presents the second answer generated by the answer generation unit 14. In other words, the answer presentation unit 15 presents an answer to a request that is created based on the output of the artificial intelligence 4 to which the input data set has been input.

[0141] The answer presentation unit 15 presents the second answer using a display means such as a display provided in the terminal device 2A. Note that the information processing system according to the first embodiment may omit the answer generation unit 14. When the information processing system is configured so as to omit the answer generation unit 14, the answer presentation unit 15 presents the first answer output by the artificial intelligence 4 as is.

[0142] When the request received by the request receiving unit 11 is for creating or outputting a data file, the information processing system 1A stores the data file in a predetermined storage area. In this case, the answer presenting unit 15 presents information indicating that the data file has been saved.

[0143] As described above, the series of roles played by the components of the information processing system 1A enables the information processing system 1A to provide a more accurate answer to a request input to the information processing system 1A. The information processing system 1A can provide a more accurate answer than when an answer is obtained by simply inputting a request to the artificial intelligence 4.

[0144] <Specific example> The series of operations of the information processing system according to the first embodiment explained above will be explained using a specific example. Here, a case will be taken as an example in which a user inputs a request to "change tool at T08" to the information processing system. If the information processing system according to the first embodiment is not applied, such a request will be input as is to the artificial intelligence 4. In this case, since the request is unclear, the artificial intelligence 4 may output a response such as, for example, "Understood. We will change tool at T08. Please wait a moment."

[0145] In the information processing system according to the first embodiment, first, "replace" is determined as a verb and "tool" is detected as an object by the request receiving unit 11. The request receiving unit 11 outputs request information including the original input request "replace tool at T08", the verb information "replace", and the object information "tool".

[0146] Next, the input generating unit 12 searches the specification DB using the keyword "tool change." As a result, "M06" is found as an M-code command for tool change, and information "Tt M06" is detected as a program example.

[0147] 17 is a diagram showing an example of a prompt included in the input data set in the information processing system according to Embodiment 1. By performing the above-described processing by the request receiving unit 11 and the input generating unit 12, for example, simply configuring the prompt of the input data set using information such as that shown in FIG. 17 can be expected to bring the output from the artificial intelligence 4 closer to the correct answer, "T08 M06."

[0148] Next, we will explain a case where a user inputs a request for creating a ladder program into an information processing system. While there are various methods for writing ladder programs, we will use the ST language as an example. ST language is a language defined in IEC 61131-3. Using ST language, ladder programs or sequence programs can be written using control syntax such as expressions using operators (e.g., *, / , +, -, =), conditional statements for branching, loop statements for repetition, calls to user-defined function blocks (FBs), calls to other vendor-defined functions, and comments containing full-width characters. Because ST language is similar to a high-level language, it is expected to be easier to process using artificial intelligence (AI) than ladder languages.

[0149] Figure 18 is the first diagram for explaining an example in which the creation of a ladder program is requested to the information processing system according to embodiment 1. For example, suppose a request to create a ladder program as shown in Figure 18 is input as a request. If the information processing system according to embodiment 1 is not applied, such a request will be input as is to the artificial intelligence 4. In this case, the "ladder program" indicated in the request is generally assumed to be a program in a ladder diagram, and therefore may not be processable by the text-based artificial intelligence 4.

[0150] In the information processing system according to the first embodiment, first, the request receiving unit 11 determines "create" as a verb and detects "ladder program" as an object. The request receiving unit 11 outputs request information including the entire original request that was input and a combination of the verb information "create" and the object information "ladder program."

[0151] Next, the input generating unit 12 searches the specification DB using the keyword "ladder program." As a result, "ST language" is detected.

[0152] Fig. 19 is a second diagram for explaining an example in which the information processing system according to embodiment 1 is requested to create a ladder program. Fig. 20 is a third diagram for explaining an example in which the information processing system according to embodiment 1 is requested to create a ladder program. Through the above-described processing by the request receiving unit 11 and the input generating unit 12, for example, the phrase "in ST language" as shown in Fig. 19 is supplemented as a prompt for the input data set. By supplementing the phrase "in ST language," for example, the possibility of obtaining an answer describing a specific program as shown in Fig. 20 increases.

[0153] The answer generation unit 14 also performs a process of deleting the explanatory sentences at the beginning and end of the answer shown in Fig. 20. Furthermore, the answer generation unit 14 performs processes such as checking whether labels can be used or verifying grammar for the program portion of the answer shown in Fig. 20. The information processing system can present the program desired by the user by presenting a second answer after the answer generation unit 14 has performed these processes.

[0154] Next, we will explain a case where a user inputs a request for a change to a work procedure manual into an information processing system. A work procedure manual is a document created for the purpose of standardizing the procedures for using machine tools. A work procedure manual generally describes procedures and precautions for work using machine tools and the processes before and after, such as advance preparation (setup) and cleanup, using text and diagrams. Work procedure manuals are sometimes used as paper documents, but they can also be created and referenced as electronic files such as Word, Excel, or Portable Document Format (PDF) files.

[0155] The purpose of creating work procedures is to standardize work quality regardless of knowledge or skill by having operators who use machine tools follow these work procedures. Even if there is a slight change or modification to the procedure, it is desirable to revise the work procedure manual so that it matches the actual work, but changing the electronic file can be time-consuming for operators.

[0156] For example, suppose a user inputs a request to the information processing system saying, "Please change work procedure manual_No.xxx as follows [change location / change content] and create work procedure manual_No.xxx-1." The [change location / change content] may include information that identifies the change location in the work procedure manual that is the target of the change, such as "add a note to work procedure 1 that says to wear protective equipment," and information that specifies the change content.

[0157] In the information processing system according to the first embodiment, first, the request receiving unit 11 detects "change" and "create" as verbs, and "working manual_No.xxx" and "working manual_No.xxx-1" as objects. The request receiving unit 11 outputs request information including the original request that was input, and a set of verb information "change" and "create," and object information "working manual_No.xxx" and "working manual_No.xxx-1."

[0158] Next, the input generation unit 12 searches the retained data for the keywords "work procedure manual" and "creation." As a result, "work procedure manual_No.xxx" is detected as existing data, and "work procedure manual_No.xxx" is set as a data file of the input data set. The original request content is set in the prompt. The transmission / reception unit 13 transmits the input data set generated by the input generation unit 12 in this manner to the artificial intelligence 4. The artificial intelligence 4 is assumed to be an artificial intelligence capable of inputting and outputting data files. The transmission / reception unit 13 receives the new work procedure manual, which is the output of the artificial intelligence 4 in response to the input. In this way, the information processing system obtains, as a first answer, a new work procedure manual in which the desired change locations have been changed with the desired change content.

[0159] The answer generation unit 14 generates a second answer by changing the name of the file acquired as the first answer to "Work Procedure Manual_No.xxx-1." The answer presentation unit 15 displays the file, notifies the user that the file has been created, or displays the location where the file is stored. This completes a series of responses to requests by the information processing system.

[0160] Next, a description will be given of an example of outputting a data file by the information processing system according to embodiment 1. Here, a case will be described in which a setting parameter file, which is a file that collects setting parameters, is output as a data file.

[0161] One example of a case in which output of a setting parameter file is desired is when the current operating environment of a machine tool needs to be saved as a backup during machine tool maintenance. In such a case, all setting parameters may be targeted for saving, but there may also be cases in which saving only specific setting parameters is desired. Depending on the user's wishes, requests such as "save all current setting parameters," "save parameters related to acceleration / deceleration settings," or "save X-axis parameters" are input to the request receiving unit 11. In response to such requests, the request receiving unit 11 determines "save" as verb information and detects "xxx (setting) parameters" as object information. Here, "xxx" refers to information related to the setting parameters requested to be saved, such as "acceleration / deceleration settings," and is information that changes depending on the input request.

[0162] The input generation unit 12 detects all the setting parameters. The input generation unit 12 converts the detected setting parameters into data in a data format that can be handled by the artificial intelligence 4, such as text data. The input generation unit 12 sets a setting parameter file consisting of the converted setting parameters as a data file of the input dataset. The input generation unit 12 also searches for information corresponding to the above "xxx" by referencing the specification DB. For example, if "xxx" is "acceleration / deceleration setting relationship," keywords such as acceleration / deceleration or acceleration / deceleration time constant are detected. Information for specifying specific parameters corresponding to the detected keywords, such as parameter numbers or parameter names, is included in the prompt that constitutes the input dataset. This allows the output by the artificial intelligence 4 to be limited to only the setting parameters indicated in the prompt from the setting parameter file set as the data file of the input dataset.

[0163] Furthermore, the data structure of the setting parameter file often employs a table format in which each parameter is a row and the parameter setting target, such as the X-axis or Y-axis, is a column, with the parameter setting value stored at the intersection of the row and column. For this reason, including a keyword such as "X-axis" in the prompt can be expected to inform the artificial intelligence 4 which row value in the setting parameter file should be extracted. Furthermore, to further improve the accuracy of the output of the artificial intelligence 4, the input generation unit 12 may specify the data structure of the setting parameters in the prompt.

[0164] The transmitting / receiving unit 13 transmits the input data set created by the above processing by the input generating unit 12 to the artificial intelligence 4. The transmitting / receiving unit 13 receives the setting parameter file, which is the first answer output from the artificial intelligence 4. This allows the information processing system to obtain the setting parameter file desired by the user.

[0165] The answer generation unit 14 generates a second answer by assigning a file name to the setting parameter file that is the first answer. The answer generation unit 14 may assign a name such as "X-axis setting parameter" that is created based on the contents of "xxx" as the file name. Alternatively, the answer generation unit 14 may assign the date and time of request execution as the file name to make it possible to distinguish between multiple setting parameter files.

[0166] Finally, the information processing system displays the file name of the saved setting parameter file, notifies the user that saving of the setting parameter file has been completed, or notifies the user of the location where the setting parameter file has been saved, by the answer presenting unit 15. In this way, the information processing system completes the process for the request.

[0167] <Effects> According to the first embodiment, the information processing system includes a request receiving unit 11 that receives a request for work performed when using the machine tool; an input generating unit 12 that generates an input data set corresponding to the request by referencing at least one of a specification DB, setting parameters, and stored data stored in a storage area accessible to the information processing system; and an answer presenting unit 15 that presents an answer to the request created based on the output of the artificial intelligence 4 to which the transmitted input data set has been input. Compared to when the request is input directly to the artificial intelligence 4, the information processing system can improve the stability and reproducibility of the output in response to the input, thereby improving the accuracy of the output from the artificial intelligence 4. The information processing system can achieve general-purpose ease of use by using natural language. As a result, the information processing system achieves a first effect of being able to suppress variation in output quality while providing general-purpose ease of use in dialogue responses in natural language using the artificial intelligence 4.

[0168] Further, the answer presentation unit 15 presents the second answer generated by the answer generation unit 14. The requests accepted by the request acceptance unit 11 include at least one of creation of a machining program, creation of a ladder program, creation or output of a data file related to a machine tool or work using the machine tool, and response to a question related to a machine tool or work using the machine tool. The information processing system does not simply present the output obtained from the artificial intelligence 4 as an answer, but rather the answer generation unit 14 performs at least one of verification of basic rules, verification of functional specifications, verification of operation, and conversion of the answer format to a second answer. This achieves a second effect in that the information processing system can present a more accurate answer to the user.

[0169] The first effect described above is achieved by including the input generation unit 12, and the second effect described above is achieved by including the answer generation unit 14. These two effects are obtained independently of each other. That is, even if the information processing system does not include one of the input generation unit 12 and the answer generation unit 14, the effect provided by the other of the input generation unit 12 and the answer generation unit 14 can be obtained without any problems. Furthermore, if both the input generation unit 12 and the answer generation unit 14 are included, the effects provided by each of the input generation unit 12 and the answer generation unit 14 are not diminished. In this case, the effects provided by each of the input generation unit 12 and the answer generation unit 14 are synergistically combined, making it possible to present more accurate answers to the user.

[0170] The setting parameters include one or more of the type of machine tool, information indicating the manufacturer of the machine tool, and information indicating the manufacturer of the machine tool's control device. In the information processing system 1A, the specification DB to be referenced is switched depending on the set value of the setting parameters. This limits the range of specification DBs referenced and the range of specification DB searches in the information processing system 1A, thereby enabling more efficient processing. Furthermore, by referencing more appropriate specification information, the accuracy of the output of the artificial intelligence 4 can be improved.

[0171] The data file also includes one or more of the machining program, ladder program, setting parameters, operating procedure manual for the machine tool, shape data for each of the tools and jigs attached to the machine tool, and shape data for the workpiece to be machined by the machine tool. By including one or more of the machining program, ladder program, setting parameters, operating procedure manual, shape data for each of the tools and jigs, and shape data for the workpiece in the input to the artificial intelligence 4, the accuracy of the output from the artificial intelligence 4 can be further improved.

[0172] Furthermore, the conversion to the answer format showing the second answer includes a conversion that makes it possible to distinguish the portion created by the artificial intelligence 4. This allows the information processing system to clearly indicate to the user the portion of the answer to be presented to the user that was created by the artificial intelligence 4.

[0173] In the above, the artificial intelligence 4 is assumed to be a device external to the information processing systems 1A, 1B, and 1C. In the first embodiment, the artificial intelligence 4 may be a device built into the information processing systems 1A, 1B, and 1C. In the second and subsequent embodiments, the artificial intelligence 4, which is assumed to be provided outside the information processing systems, may also be provided inside the information processing systems rather than outside them.

[0174] Embodiment 2. Fig. 21 is a diagram showing a first configuration example of an information processing system according to embodiment 2. Fig. 21 shows an information processing system 1D, which is the first configuration example of an information processing system according to embodiment 2. The information processing system 1D is configured in a terminal device 2D. Fig. 21 also shows the terminal device 2D and an artificial intelligence 4 external to the information processing system 1D. In embodiment 2, the same components as those in embodiment 1 above are assigned the same reference numerals, and the following mainly describes configurations that differ from embodiment 1.

[0175] Terminal device 2D has the same configuration as terminal device 2A shown in Fig. 1. Furthermore, terminal device 2D has a notification information creation unit 21 and an information notification unit 22. The notification information creation unit 21 creates notification information including one or more pieces of information from the content of the processing by the input generation unit 12, the content of the processing by the answer generation unit 14, the input data set, and the first answer. The information notification unit 22 notifies the user of the notification information. Terminal device 2D can be realized by a hardware configuration similar to the hardware configuration shown in Fig. 2.

[0176] 21 , the components of the information processing system 1D are provided in a single device, that is, a terminal device 2D. The information processing system according to the second embodiment is not limited to one in which the components of the information processing system are provided in a single device. The components of the information processing system may be distributed across two or more devices that can function together.

[0177] Fig. 22 is a diagram showing a second configuration example of an information processing system according to embodiment 2. Fig. 22 shows an information processing system 1E, which is the second configuration example of an information processing system. The information processing system 1E is configured with a terminal device 2E and a back-end device 3E. That is, in the information processing system 1E, the components of the information processing system 1E are distributed between the terminal device 2E and the back-end device 3E.

[0178] Terminal device 2E has a configuration similar to that of terminal device 2B shown in Fig. 3 and an information notification unit 22. Back-end device 3E has a configuration similar to that of back-end device 3B shown in Fig. 3 and a notification information creation unit 21. Terminal device 2E can be realized by a hardware configuration similar to that shown in Fig. 2. Back-end device 3E can be realized by a configuration similar to that shown in Fig. 2, excluding the input device 54 and the display device 55. Note that the manner in which the components of the information processing system are distributed is not limited to the manner shown in Fig. 22 and is arbitrary.

[0179] Next, each component of the information processing system according to embodiment 2 will be described. Below, each component of the information processing system 1D will be described using the information processing system 1D shown in Fig. 21 as an example. The description of each component of the information processing system 1D also applies to the information processing system 1E shown in Fig. 22. Here, the description of the same content as in embodiment 1 will be omitted.

[0180] <Notification information creation unit 21> The notification information creation unit 21 creates notification information including one or more pieces of information from the content of the processing by the input generation unit 12, the content of the processing by the answer generation unit 14, the input data set, and the first answer.

[0181] As described in the first embodiment, the input generation unit 12 generates prompts to be included in the input data set by processing such as adding additional sentences to improve accuracy of the request information, adding an explanation of the situation by referring to the setting parameters, and adding specific examples related to the request by referring to the specification DB. The input generation unit 12 also selects data files to be included in the input data set from the stored data.

[0182] The notification information creation unit 21 includes the content of such processing by the input generation unit 12 as an action item in the notification information. The notification information creation unit 21 may also include detailed information on the specific action content in the notification information as detailed information on the action item. The method of expressing the content or detailed information of the processing in the notification information may be to describe the actual action content in detail, or to use a description that is somewhat abstracted to make it easy to understand or to conceal technical know-how.

[0183] For example, as described in the first embodiment, to clarify the position of the artificial intelligence 4, it is assumed that the prompt is supplemented with a statement such as "You are an experienced machine tool operator." In this case, the notification information may include information indicating the action item, such as "Clarify role based on basic prompting techniques," and detailed information about the action item, such as "Add a statement that you are an experienced machine tool operator." If the content of the action is to be an abstract description, the notification information may include information indicating the action item, such as "Add a statement that allows the artificial intelligence to recognize its role," and detailed information about the action item, such as "Add a statement that allows the artificial intelligence to recognize its role."

[0184] Fig. 23 is a first diagram showing an example of notification information created by the notification information creation unit 21 included in the information processing system according to the second embodiment. Fig. 23 shows an example of notification information notified by the information notification unit 22 regarding the content of processing by the input generation unit 12. In the notification information shown in Fig. 23, action items and detailed information are associated with each other. In the example shown in Fig. 23, a confirmation button for displaying more specific content of the notification information is displayed for each action item. By pressing the confirmation button, additional text indicating the specific content of the action item is displayed.

[0185] The notification information indicating the content of the processing by the answer generation unit 14 is also created in the same manner as the notification information indicating the content of the processing by the input generation unit 12. The processing by the answer generation unit 14 includes three verifications: verification of basic rules, verification of functional specifications, and verification of operation, as well as conversion to an answer format when presenting a second answer. The notification information includes information indicating the content of the verification when the verification is performed, information indicating details of the verification content, and the result of the verification. Furthermore, when the answer format is converted, the notification information includes information indicating the format before conversion and information indicating the format after conversion.

[0186] For example, if a "basic grammar check of the processing program" is performed as the verification of the basic rules, the notification information will include information such as a "check of line feed codes" or a "check of character types" that indicates details of the verification content. The notification information may also include information that indicates details of the verification results. For example, the notification information may include information that indicates parts of the processing program that were found to be inappropriate by the verification as information that indicates details of the verification results.

[0187] FIG. 24 is a second diagram illustrating an example of notification information created by the notification information creation unit 21 included in the information processing system according to the second embodiment. FIG. 24 illustrates an example of notification information notified by the information notification unit 22 regarding the content of processing performed by the answer generation unit 14. In the notification information illustrated in FIG. 24, the performed processing items and their implementation statuses are associated with each other. The implementation status may be represented, for example, by information simply indicating whether the processing item has been performed or not. Alternatively, the information indicating the implementation status may be a percentage representation of the progress, as illustrated in FIG. 24. Furthermore, the verification processing items are also associated with the verification results. In the example illustrated in FIG. 24, a confirmation button is displayed for each processing item to display more specific content of the notification information. Pressing the confirmation button displays additional text indicating the specific content of the processing item. The specific content of the processing item may, for example, be information indicating the parts that were determined to be inappropriate by the verification.

[0188] For each of the input data set and the first answer, the input data set or the first answer may be included in the notification information as is, or the input data set or the first answer may be summarized or compressed so that only general information is included in the notification information.

[0189] Fig. 25 is a third diagram showing an example of notification information created by the notification information creation unit 21 included in the information processing system according to embodiment 2. Fig. 26 is a fourth diagram showing an example of notification information created by the notification information creation unit 21 included in the information processing system according to embodiment 2. Fig. 27 is a fifth diagram showing an example of notification information created by the notification information creation unit 21 included in the information processing system according to embodiment 2. Fig. 28 is a sixth diagram showing an example of notification information created by the notification information creation unit 21 included in the information processing system according to embodiment 2.

[0190] The "Command statement for AI" button shown in Fig. 25 is a button for displaying the prompt in the input data set. When the "Command statement for AI" button is pressed, a prompt such as that shown in Fig. 26 is displayed. This allows the user to check the contents of the prompt included in the input data set.

[0191] The "Input data for artificial intelligence" button shown in FIG. 25 is a button for displaying the data files in the input dataset. When the "Input data for artificial intelligence" button is pressed, an overview of the data files included in the input dataset is displayed, as shown in FIG. 27. In the example shown in FIG. 27, the data file overview displays the type of data file and the file name of the data file. This allows the user to check the overview of the data files included in the input dataset.

[0192] The "Raw data output from AI" button shown in Fig. 25 is a button for displaying the first answer. When the "Raw data output from AI" button is pressed, the first answer as shown in Fig. 28 is displayed. This allows the user to confirm the content of the first answer output by the AI ​​4.

[0193] The contents of the information included in the notification information have been described above, but the notification information may include only one type of information, two or more types of information, or all types of information.

[0194] <Information notification unit 22> The information notification unit 22 receives the notification information created by the notification information creation unit 21 and notifies the user of the notification information. The information notification unit 22 notifies the user of the notification information by, for example, displaying the notification information in a pop-up displayed on the screen.

[0195] There are no particular limitations on the manner in which the information notification unit 22 provides notification. The information notification unit 22 may provide notification each time notification information is created by the notification information creation unit 21, or may provide notification at a predetermined timing in the answer generation flow. In the first case in which notification is provided each time notification information is created, as the processing of the input generation unit 12 progresses, notification information corresponding to the completed processing content is created and a notification is provided. In this case, the user can understand the content of the processing being performed in the information processing system while waiting for the processing to create an answer to the request. Alternatively, in the second case in which notification is provided at a predetermined timing in the answer generation flow, the user can know the results of the verification each time the verification processing in the answer generation unit 14 is completed.

[0196] 23 and displays an action item for "add text" in the pop-up when the input generation unit 12 performs a process for adding text. Furthermore, the information notification unit 22 displays an action item for "change text" in the pop-up when the input generation unit 12 performs a process for further changing text. These are examples of the first case in which the information notification unit 22 notifies the user of notification information every time notification information is created.

[0197] 24 and displays the processing items for the "basic rule" in the pop-up when the answer generation unit 14 has completed verification of the basic rule. Furthermore, the information notification unit 22 displays the processing items for the "functional specification" in the pop-up when the answer generation unit 14 has completed verification of the functional specification. These are also examples of the first case in which the information notification unit 22 notifies the user of notification information every time the notification information is created.

[0198] In the second case, the information notification unit 22 notifies the user of the processing details of the input generation unit 12 and information related to the input dataset, for example, when the processing of the input generation unit 12 is completed and the input dataset is generated. This allows the user to check the progress. In the second case, the pop-up shown in FIG. 23 or 25 is displayed when the processing of the input generation unit 12 is completed and the input dataset is generated.

[0199] <Effects> According to the second embodiment, the information processing system includes a notification information creation unit 21 that creates notification information including one or more pieces of information selected from the content of processing performed by the input generation unit 12, the content of processing performed by the answer generation unit 14, the input data set, and the first answer, and an information notification unit 22 that notifies the user of the notification information. This allows the information processing system to present the basis for the answer or the process leading up to the answer to the user, allowing the user to recognize the correctness of the answer and the evidence supporting the answer. Therefore, the information processing system can improve the user's sense of satisfaction with the presented answer, thereby improving the reliability of the information processing system.

[0200] Embodiment 3. Fig. 29 is a diagram showing a first configuration example of an information processing system according to embodiment 3. Fig. 29 shows an information processing system 1F, which is a first configuration example of an information processing system according to embodiment 3. The information processing system 1F is configured in a terminal device 2F. Fig. 29 shows the terminal device 2F and an artificial intelligence 4 external to the information processing system 1F. In embodiment 3, the same components as those in embodiment 1 or 2 above are assigned the same reference numerals, and the description will mainly focus on configurations that differ from embodiment 1 or 2.

[0201] The terminal device 2F has the same configuration as the terminal device 2A shown in Fig. 1. Furthermore, the terminal device 2F has a request determination unit 31, a partial request generation unit 32, a database selection unit 33, an order determination unit 34, an information complementation unit 35, and a request management unit 36. The information complementation unit 35 is provided in the input generation unit 12. Hereinafter, the database selection unit will be referred to as a "DB selection unit." The terminal device 2F can be realized by a hardware configuration similar to the hardware configuration shown in Fig. 2.

[0202] 29, the components of the information processing system 1F are provided in a single device, that is, a terminal device 2F. The information processing system according to the third embodiment is not limited to one in which the components of the information processing system are provided in a single device. The components of the information processing system may be distributed across two or more devices that can function together.

[0203] Fig. 30 is a diagram showing a second configuration example of an information processing system according to embodiment 3. Fig. 30 shows an information processing system 1G, which is the second configuration example of the information processing system. The information processing system 1G is configured with a terminal device 2G and a back-end device 3G. That is, in the information processing system 1G, the components of the information processing system 1G are distributed between the terminal device 2G and the back-end device 3G.

[0204] The terminal device 2G has a configuration similar to that of the terminal device 2B shown in Fig. 3. The back-end device 3G has a configuration similar to that of the back-end device 3B shown in Fig. 3, and includes a request determination unit 31, a partial request generation unit 32, a DB selection unit 33, an order determination unit 34, an information complementation unit 35, and a request management unit 36. The terminal device 2G can be realized by a hardware configuration similar to that shown in Fig. 2. The back-end device 3G can be realized by a configuration similar to that shown in Fig. 2, excluding the input device 54 and the display device 55. Note that the manner in which the components of the information processing system are distributed is not limited to the manner shown in Fig. 30 and is arbitrary.

[0205] Next, an overview of each component of the information processing system according to embodiment 3 will be described. Below, each component of the information processing system 1F will be described using the information processing system 1F shown in Fig. 29 as an example. The description of each component of the information processing system 1F also applies to the information processing system 1G shown in Fig. 30. Here, the description of the same content as in embodiment 1 or 2 will be omitted.

[0206] The request determination unit 31 determines the request category included in the request information and creates request category information. The request category information is referenced by both the input generation unit 12 and the request management unit 36. The information processing system 1F switches the processing when generating an input data set depending on the request category. This is expected to further improve accuracy.

[0207] The partial request generation unit 32 generates one or more partial requests based on the request. The partial request generation unit 32 may generate partial requests based only on the request information, but by using request category information to generate partial requests, it is possible to generate a set of partial requests of a single category from a request that spans multiple categories. Converting a request into partial requests enables the input generation unit 12 to generate an input data set for each partial request. This can further simplify the content of the request to the artificial intelligence 4, and is expected to prevent a deterioration in the accuracy of the output of the artificial intelligence 4. Note that, hereinafter, converting a request into multiple partial requests refers to generating multiple partial requests based on the request.

[0208] The partial request generation unit 32 converts a request into multiple partial requests based on the decomposition of the original request. Alternatively, the partial request generation unit 32 may convert a request into multiple partial requests by searching for information contained in the specification DB for information to be referenced by the input generation unit 12. In this case, the partial request can be said to be a request to the artificial intelligence 4 to search and extract information from the specification DB that is appropriate for creating an input data set corresponding to the request entered by the user. Hereinafter, a partial request generated by such information search will be referred to as a knowledge search partial request.

[0209] Furthermore, the partial request generator 32 may convert the request into multiple partial requests by including information obtained as a result of the search in the input data set. In other words, it uses the answers to the knowledge search partial requests to create an input data set corresponding to the original request and the partial requests obtained by decomposing the original request. This concept is known as Retrieval-Augmented Generation (RAG).

[0210] Next, the differences between the processing contents of the input generation unit 12 and the first embodiment will be outlined. In the third embodiment, the general processing contents of generating an input data set based on request information are the same as those in the first embodiment. In the third embodiment, processing by the DB selection unit 33 and the information complementation unit 35, which are components related to the processing by the input generation unit 12, and the above-mentioned request determination unit 31 and partial request generation unit 32 are added.

[0211] The DB selection unit 33 selects the specification DB to be referenced according to the setting parameters. This limits the range of specification DBs to be referenced and the range of specification DBs to be searched, making it possible to improve processing efficiency. Furthermore, by referencing more appropriate specification information, it is expected that the accuracy of the output of the artificial intelligence 4 will be improved. Note that in cases where the above-mentioned RAG is used, the DB selection unit 33 may select the specification DB using a response to a knowledge search partial request.

[0212] The information complementing unit 35 complements information that is missing in the request. The information complementing unit 35 may also add a new partial request. The information complementing unit 35 determines missing information or extracts additional information using the specification DB selected by the DB selection unit 33. This makes it possible to complement information that has been overlooked by the user in the request. Alternatively, it makes it possible to complement information that is implicitly assumed by the user but is unknown to the artificial intelligence 4. This is expected to improve the accuracy of the output of the artificial intelligence 4.

[0213] The order determination unit 34 determines the order of all partial requests, including the requests added by the information supplementation unit 35. The order determination unit 34 generates partial request order information, which is information indicating the determined order. By linking the input data set corresponding to each partial request with the partial request order information corresponding to each partial request, the partial request order information does not need to be included in the input data set. In other words, the partial request order information may be managed as information separate from the input data set.

[0214] The information processing system 1F includes a partial request generation unit 32, an order determination unit 34, and an information supplementation unit 35, and thereby generates a plurality of partial requests based on a request, determines the order of the plurality of partial requests, and supplements information that is missing from the request. Note that the information processing system 1F is not limited to performing all of generating a plurality of partial requests based on a request, determining the order of the plurality of partial requests, and supplementing information that is missing from the request. The information processing system 1F may perform one or more of generating a plurality of partial requests based on a request, determining the order of the plurality of partial requests, and supplementing information that is missing from the request.

[0215] This concludes the description of the additional configuration related to the input generation unit 12. Next, the content of the processing added to the input generation unit 12 will be described.

[0216] When the request category information is output by the request determination unit 31, the input generation unit 12 switches the generation process of the input data set according to the request category indicated in the request category information, as described above. Specifically, for example, if the request category information indicates that the request is "related to a machining program," the input generation unit 12 refers to a specification DB related to the machining program and performs processing such as adding specific examples of the program format to the prompt. This is expected to improve the accuracy of the output of the artificial intelligence 4.

[0217] When multiple partial requests are generated, the input generation unit 12 generates an input data set for each partial request. The input generation unit 12 also sets a destination artificial intelligence 4 or external processing (such as any software or web application) for each input data set. This destination does not have to be one. Multiple destinations may be set. For example, even if the content of the partial request is the same, multiple artificial intelligences 4 with different models may be set as destinations. By comparing the output results of the multiple artificial intelligences 4, the input generation unit 12 can select an output result that is more suitable as an answer. Alternatively, by majority voting of the multiple output results, the input generation unit 12 can select an output with a higher degree of accuracy as an answer.

[0218] The prompt may include at least one or more command information. Additionally, the prompt may include example information or context information. This information may be implicitly included in the prompt by being embedded in a sequence of sentences. By constructing the prompt in a structured manner, this information is explicitly stated, which is expected to improve the accuracy of the output of the artificial intelligence 4.

[0219] When there are multiple input datasets, the request management unit 36 ​​manages the processing of the input datasets in cooperation with the transmission / reception unit 13. The request management unit 36 ​​manages the processing of the input datasets so that the input datasets for each partial request are processed in the order indicated in the partial request order information. In doing so, the request management unit 36 ​​controls the transmission / reception unit 13 so that the input datasets are transmitted to the destinations set for each input dataset.

[0220] For example, in a certain input data set, a response may be required for a partial request that is processed in a sequence preceding the partial request corresponding to the input data set. Such a constraint is hereinafter referred to as an "order constraint." When an order constraint exists, the request management unit 36 ​​waits for reception of a response for the partial request that is processed in the sequence preceding the partial request, and then causes the transmitting / receiving unit 13 to start transmission processing after attaching information indicating the response to the input data set.

[0221] If there is no such order constraint and the partial requests have different destinations, the request management unit 36 ​​may control the transceiver 13 to start sending the next partial request without waiting for a response to the partial request processed in the previous order. This allows the information processing system 1F to process each partial request in parallel, enabling efficient processing.

[0222] After all of the first partial answers have been received, the obtained first answers to the partial requests (hereinafter referred to as "first partial answers") may be integrated together and output as a first answer to the answer generation unit 14. In this way, it is possible to avoid the first partial answers being determined to be inappropriate in various verification processes in the answer generation unit 14.

[0223] Furthermore, when each first partial answer is an answer to a partial request of a different request category, the first partial answers are independent from each other, and the verification process in the answer generation unit 14 may also be independent. In such a case, the first partial answers may be output as is to the answer generation unit 14, and after the answer generation unit 14 performs a verification process on each first partial answer, the answer generation unit 14 may convert the answer format to integrate the first partial answers into a second answer. In this way, the information processing system 1F can perform the processing of the transmitter / receiver 13 and the processing of the answer generation unit 14 in advance from the part of the first answer for which the answer by the artificial intelligence 4 has been completed, thereby enabling efficient processing.

[0224] In this way, the second answer in the third embodiment is generated. Next, the details of each of the above components will be described. Note that the description of the contents that overlap with the first or second embodiment will be omitted.

[0225] <Request Determination Unit 31> The request determination unit 31 extracts characteristics of a request from information included in the request. The request determination unit 31 calculates the degree of conformance to a category type based on the extracted characteristics. The request determination unit 31 determines the request category based on the calculated degree of conformance. As a result, the request determination unit 31 determines the category of the request and creates request category information.

[0226] Here, a requirement category is a classification of requirements distinguished by one or both of the type of reference information required to satisfy the requirement and the type of output from the artificial intelligence 4. The reference information required to satisfy the requirement is a concept that collectively expresses the above-mentioned specification DB, setting parameters, and retained data. The type of output from the artificial intelligence 4 refers to the type of data, such as text data, image data, or 3D data.

[0227] Specific examples of request categories include creation of a machining program, creation of a ladder program, creation or output of a data file related to a machine tool or work using a machine tool, or responses to questions related to a machine tool or work using a machine tool. Furthermore, creation or output of a data file can be further subdivided according to the type of data file in question. Similarly, responses to questions can also be subdivided according to the type of question. For example, questions can be divided into general questions and specialized questions. Because the information to be referenced differs between specialized information related to machine tools and other general information, the information processing system 1F can more appropriately respond to requests by subdividing responses according to the type of question.

[0228] Examples of information included in a request include character string information in natural language, or information indicating the results of a screen operation or button operation. Examples of features extracted from the information included in a request include the presence or absence of a specific keyword, the frequency of occurrence of a certain word, or a combination of specific words and verbs. Keywords may be detected by referring to a keyword list prepared in advance for each category type. Furthermore, information indicating the target of the operation is a feature of information indicating the result of a screen operation or button operation. For example, when a screen operation such as that shown in FIG. 11 is performed, information such as "create" and "processing program" is considered to be a feature.

[0229] The suitability for a category type is an evaluation value calculated based on the above-mentioned characteristics of the information included in the request and the criteria for determining the category type. For example, the evaluation value may be increased by 1 when a specific keyword is included, or the frequency of occurrence of a specific word may be used as the evaluation value. The text information included in the information included in the request may be vectorized, and the cosine similarity calculated using the word vectors that serve as criteria for determining the category type or text vectors expressed as a linear combination thereof may be used as the suitability.

[0230] When the information included in the request is information indicating the results of a screen operation or button operation, the degree of conformance may be calculated based on whether the information indicating the target of the operation matches a pre-designed key operation or button. For example, if the buttons are operated in the order shown in Figure 12 and the button for selecting a machining program is pressed last, the category of the request is likely to be creation and output of a machining program, or creation or output of a machining program. In this case, the request determination unit 31 may calculate a high degree of conformance for the request category of creation or output of a machining program.

[0231] Furthermore, the request determination unit 31 may determine multiple request categories for one request based on the result of the request category determination. For example, if the determination result in the request determination unit 31 does not determine one request category, the request category with the highest suitability may be used as the determination result, or all request categories with suitability exceeding a certain threshold may be used as the determination result.

[0232] However, if multiple request categories are determined as the determination result, it is desirable that multiple partial requests be generated by the partial request generation unit 32 (described later). In this case, the request determination unit 31 may determine the request category for each of the multiple partial requests. By repeating the process in the request determination unit 31 in this manner, one partial request can be associated with one request category.

[0233] The information processing system 1F can generate an input data set according to the request category by determining the request category using the request determination unit 31. This allows the information processing system 1F to prevent information that is less relevant to the request category from being included in the input data set, thereby improving the accuracy of the output by the artificial intelligence 4.

[0234] <Partial Request Generator 32> The more complex the content of the request information, the more likely it is that part of the request will be ignored by the artificial intelligence 4. This can lead to a deterioration in accuracy, such as an inability to obtain an appropriate answer. Furthermore, if a request includes content that spans multiple request categories, and the request is input to the artificial intelligence 4 as is, there is also a possibility that part of the request will be ignored by the artificial intelligence 4. Therefore, the information processing system 1F according to the third embodiment uses the partial request generator 32 to convert a request into multiple partial requests. Note that the following mainly describes the case where the conversion process from a request to multiple partial requests in the partial request generator 32 is a process of dividing the request into multiple partial requests. The partial request generator 32 may also perform a conversion to add knowledge search partial requests using the RAG technique described above.

[0235] Whether or not the partial request generator 32 generates multiple partial requests is determined based on the characteristics of the information included in the request. Specific examples of the characteristics include the request including a sentence with multiple verbs, the request including information indicating multiple request categories, or the sentence including multiple keywords.

[0236] An example of a case where a request includes a sentence with multiple verbs is a case where the request includes a sentence such as "Create a machining program for xxx and output it to yyy." In this case, since the sentence includes two verbs, "create" and "output," the partial request generation unit 32 determines that the request is a decomposable request. Based on the request, the partial request generation unit 32 generates a partial request, "Create a machining program for xxx," and another partial request, "Output the created machining program to yyy."

[0237] Furthermore, each of the two words "processing program" and "output" contained in the above sentence can be a keyword. "Processing program" is a keyword suggesting that the request is related to the creation of a processing program. "Output" is a keyword suggesting that the request is related to the output of some kind of file. From this, the partial request generation unit 32 can generate two partial requests by searching for sentence breaks before and after these two keywords, in the same way as in the above case where attention is focused on verbs.

[0238] The partial request generation by the partial request generation unit 32 is not limited to one time. For example, if the partial request obtained by the first generation is a first partial request, the partial request generation unit 32 can determine whether a partial request can be generated from the first partial request. If the partial request generation unit 32 determines that a partial request can be generated from the first partial request, it may obtain a second partial request, which is a partial request generated a second time, from the first partial request. The partial request generation unit 32 repeats the partial request generation process until it is determined that further partial request generation is unnecessary for all partial requests, and ends the partial request generation when it is determined that further partial request generation is unnecessary for all partial requests.

[0239] The information processing system 1F may have the artificial intelligence 4 perform the partial request generation process, which is realized by the partial request generator 32 in the above example. As in the above example, this process is realized by including in the partial request a higher-level request that generates a partial request based on the original request when it is determined that the partial request can be generated. Also, by preparing a partial request that refers to the output result of the higher-level request, a response to the original request can be generated in a series of processing flows.

[0240] <DB Selection Unit 33> The DB selection unit 33 selects a specification DB to be referenced by the input generation unit 12 according to the setting parameters. For example, there are various differences in the G-codes and M-codes, which are the functional commands of a machining program, depending on conditions such as the type of machine tool, the manufacturer of the machine tool, or the manufacturer of the numerical control device that controls the machine tool. Even if all of these conditions are the same, there may be setting parameters that can change the command method. Therefore, the DB selection unit 33 can more accurately grasp the functional specifications by switching the specification DB to be selected according to the setting parameters.

[0241] Furthermore, even if two machine tools have the same conditions, the peripheral devices provided with the machine tools may differ, or the fixtures provided with the machine tools may differ. In such cases, the configuration information of the target machine tool can be acquired by referencing the setting parameters. By referencing the required specification DB based on the acquired configuration information and excluding unnecessary specification DB information, it is possible to prevent referencing incorrect information. For example, one specific example is to switch whether or not device information or M-codes for controlling external transport devices such as a gantry loader, pallet changer, or pallet pool system are included in the reference targets, depending on whether or not they are present.

[0242] By including the DB selection unit 33, the information processing system 1F can select a specification DB according to the information of the machine tool, numerical control device, or artificial intelligence 4 that the user expects to be the target of the request. The information processing system 1F can select a specification DB that is more suited to the usage environment of the machine tool, numerical control device, or artificial intelligence 4 as the specification DB to be referenced when generating the input data set, thereby improving the accuracy of the information or data included in the input data set. This allows the information processing system 1F to improve the accuracy of the output from the artificial intelligence 4.

[0243] <Input Generation Unit 12> Here, we will explain the features of the input generation unit 12 in embodiment 3. Note that the general role of the input generation unit 12, which is to create an input data set consisting of a prompt and a data file in response to a request, is the same as in embodiment 1, so explanation will be omitted.

[0244] The input generating unit 12 changes the input data generation process by switching one or more of the specification DB to be referenced, the setting parameters, and the stored data based on the request category information.

[0245] <Input Data Set> One of the features of the input data set in embodiment 3 is that the prompt contains command information, example information, or context information. Another feature of the input data set in embodiment 3 is that the data file is any one of a machining program, a ladder program, setting parameters, a work procedure manual, 3D model data of a tool, and 3D model data of a workpiece. Here, the definition of the information contained in the prompt will be described.

[0246] <Command Information> Command information is information indicating the content of an instruction included in a request. As described above, a request is converted into request information by the request receiving unit 11. Request information is data that makes at least the content of the request available to the input generating unit 12, and is defined as data including one or more verb information and one or more object information paired with the verb. Therefore, command information, like request information, is defined as data including one or more verb information and one or more object information paired with the verb information. As can be seen from this definition, basically, request information converted from a request input by a user corresponds to command information. However, if a request is converted into multiple partial requests or if a request is added, the verb information and object information included in the additional request corresponds to command information.

[0247] <Example Information> Example information is information in which input and output are paired, and includes a desired output format, specific examples, and specific examples of content related to the request. Here, input and output refer to the input to the artificial intelligence 4 and the output expected for the input. Taking a request for a machining program as an example, if information including the keyword "tool center point command" is input, the output is expected to include the keyword "G43.5." A desired output format or specific example is a pair of input and output similar to the request content, and by adding this pair of information to the input to the artificial intelligence 4, it is attached to a prompt to induce an output pattern.

[0248] When answering questions or creating machining programs or ladder programs, output written in text is required. When creating data files, output is required in the file format of each data file or in the format of each data file. Also, even if the output is written in text, when answering questions, output in conversational text is required. Also, when creating machining programs, explanatory text is not required, and output written in the format specified for the machining program is required. When creating ladder programs, output written in a dedicated format is also required.

[0249] In the case of an artificial intelligence 4 using a language model, unless otherwise specified, it is likely that an answer will be output in a conversational format. For example, if the request is for the creation of a machining program, one block of machining program commands will be extracted from the specification DB and used as a specific example of the machining program. Also, if the request is for the creation of a ladder program, an example of a single-function function will be extracted from the specification DB and used as a specific example of the ladder program. This allows the information processing system 1F to guide the output from the artificial intelligence 4 into an output format that conforms to the specific example. In the case of a request for the creation of a machining program, a specific program example that conforms to the format of the machining program corresponds to example information.

[0250] A specific example of content related to a request is a pair of information related to the request, such as a keyword, and an input and expected output when the information related to the request is input. For example, if the keywords included in the request include words such as "tool change" and "return to origin," the pair of "tool change" and "M6" and the pair of "return to origin" and "G28" are specific examples of content related to the request. In other words, this means teaching the artificial intelligence 4 information on the parts necessary to satisfy the request.

[0251] <Contextual Information> Contextual information includes background, prerequisite knowledge, or constraints that are prerequisites for a command. The background and prerequisite knowledge for a command are information that is desirable to have in order to accurately understand and execute the command, such as technical terminology, know-how, and tacit knowledge. Because language models are trained using commonly available linguistic information, they are expected to be trained and accurately understand commonly known information. In contrast, specialized information related to the specific field of machine tools has a smaller amount of information compared to the overall commonly available linguistic information, and may not have been sufficiently trained. Therefore, by explaining the definitions and meanings of technical terms that are characteristic of the field, or terms that have unique meanings in the field, it is expected that the content of the command will be accurately understood.

[0252] For example, since the word "program" is generally likely to be interpreted as referring to a program in a computer programming language such as C or Python, a specific example would be to add an explanation to the prompt as contextual information indicating prior knowledge, such as, "A machining program is a set of instructions for specifying the shape, dimensions, and machining method of a workpiece. Programs are written in specific formats called G-code or M-code."

[0253] Furthermore, while the word "block" means one line in the context of a machining program for a machine tool, it may be understood to mean a physical block in a general context. For this reason, a specific example would be to add an explanation such as "In a machining program, a block is one line of the program" to the prompt as contextual information indicating prerequisite knowledge.

[0254] For example, information indicating whether a machine tool is a cutting machine, a laser machine, or an electric discharge machine, or, if the machine tool is a cutting machine, information indicating whether the cutting machine is a machining center or a lathe, corresponds to the background or prerequisite knowledge that is the premise of a command in terms of context information.

[0255] Furthermore, restrictions are items that must be observed when creating a response to a request, and include information that corresponds to either a requirement that must be met or a prohibition that must never be implemented. To put it more simply, the information included in restrictions can be considered information equivalent to basic rules or functional specifications verified by the response generation unit 14. For example, in the case of a machining program, information such as "Add a '%' at the beginning and end of the program" is a restriction and corresponds to a requirement among restrictions. Similarly, information such as "Specific G-codes must not be written in the same block" is a restriction and corresponds to a prohibition among restrictions. Including such restrictions in the prompt in advance is expected to reduce the possibility that the artificial intelligence 4 will output a response that violates basic rules or functional specifications.

[0256] The example information and context information can be selected by keyword search using only keywords included in the requirement information. If requirement category information is available, the example information or context information can be selected more efficiently and accurately. For example, the specification DB may be divided into requirement categories, and information corresponding to the requirement category may be assigned like a label to the information included in the specification DB. This allows the search for example information or context information to be narrowed down to the specification DB corresponding to the requirement category.

[0257] As described above, in the third embodiment, by including information such as command information, example information, or context information in the prompt of the input data set, the artificial intelligence 4 can accurately grasp the content of the request when creating a response to the request. This is expected to improve the accuracy of the output from the artificial intelligence 4.

[0258] Command, example, or context information may be included in the request if the user has carefully entered the request, but what makes the third embodiment even more effective is that it clearly associates this information with headings, tags, etc., and that it structures the information so that it can be accurately identified and associated by the artificial intelligence 4.

[0259] The input data set may include one or more of command information, context information, and example information. By explicitly specifying one or more of command information, context information, and example information and including structurally constructed prompts in the input data set, the information processing system 1F can improve the accuracy of the information or data input to the artificial intelligence 4 and improve the accuracy of the output from the artificial intelligence 4.

[0260] FIG. 31 is a first diagram showing an example of a prompt included in an input data set in an information processing system according to embodiment 3. FIG. 31 shows an example in which information such as command information, example information, or context information is simply included in the prompt as a series of sentences. Even when information such as command information, example information, or context information is simply included in the prompt as a series of sentences, improvement in the accuracy of the output from the artificial intelligence 4 can be expected. However, a feature of embodiment 3 is that further improvement in accuracy can be expected by explicitly stating this information in the prompt as a structural description.

[0261] 32 is a second diagram showing an example of a prompt included in an input data set in an information processing system according to embodiment 3. In the example shown in FIG. 32, context information is written after the tag "###Instruction###." Example information is written after the tag "###Example###." Command information is written after the tag "###Question###." This makes it clear which parts of the text information included in the prompt correspond to which roles.

[0262] Furthermore, by including in the context information "Please follow the format of Example when answering," it is possible to impose constraints on the output format using example information described with Example tags. In this way, describing each piece of information in association with an identifier such as a tag to create a structured prompt that clearly indicates relationships such as references and dependencies between pieces of information is one of the features of the processing by the input generation unit 12 of embodiment 3. Note that the notation shown in FIG. 32 is merely an example, and any notation structured to clearly indicate the correspondence between the roles of command information, example information, and context information and the sentences can be used, and various modifications of the notation are possible.

[0263] <Information Complement Unit 35> The information complement unit 35 determines whether information is missing in the request based on the content of the request information or partial request. If the information complement unit 35 determines that information is missing in the request, it complements the missing information. This process complements the information required for the input data set, and is expected to bring the output from the artificial intelligence 4 closer to satisfying the user's request.

[0264] The information complementing unit 35 detects missing information by comparing the contents of the request information or partial request with the knowledge master. The knowledge master is a collection of data (knowledge data) on standard operating procedures assumed for each request, and is a database configured so that corresponding knowledge data can be searched using information or elements contained in the request as a search key. There are no particular restrictions on the format of the knowledge master database. The database format may be a simple table-type data structure or may be configured using database software such as SQL (Structured Query Language). It is desirable for knowledge data to be created and stored in a subdivided manner for each request, but it may also be created and stored at a larger granularity, for example, by request category.

[0265] When a lack of information is detected, the missing information is identified based on the difference with the corresponding knowledge data in the knowledge master. The identified missing information is added to any input data set by the input generation unit 12 depending on its type. More specifically, the identified missing information is added as an additional request to the original request information or a request before or after the partial request, or to the input data set corresponding to the original request information or the partial request. Even more specifically, the identified missing information is added to one or more of the command information, example information, and context information in the input data set. The above is the function of the information completion unit 35, and the information completion unit 35 will be further explained using a specific example.

[0266] When creating a machining program, when creating a command block for a single line (one block), it is necessary to consider not only that command block alone, but also the contents of the commands in the program preceding that command block. Machining program commands include modal commands, whose status is preserved even after the block in which they are issued, and amodal commands, whose status is not preserved. Depending on the command status of a modal command, some functions may become unavailable or may behave differently. If these circumstances are not taken into consideration, the system may issue a warning (alarm) when the command block is executed, making it impossible to execute the machining program. Or, the execution of the machining program may result in unexpected behavior.

[0267] The above is an example of knowledge data. When comparing with a request, the information complementing unit 35 searches for blocks before the part of the machining program to be created by the request, grasps the modal state, and checks for information that is missing from the original request. The search keys for checking this request against the knowledge master are "create machining program," "add to existing program," and "create one block," and the knowledge master can be configured so that the above knowledge data can be searched.

[0268] If the comparison results in insufficient information, it is determined that information needs to be supplemented. In this case, the information supplementing unit 35 extracts the information that needs to be supplemented from the knowledge data and adds a partial request corresponding to the extracted information. Alternatively, the information supplementing unit 35 adds command information, example information, or context information that is information corresponding to the extracted information.

[0269] The information complemented by the information complementing unit 35 will be described below with a specific example. There is a command (for example, G188 / G189) that switches the program format. In a lathe-type machine tool, once G188 is commanded, the program format becomes that of a machining center-type machine tool. For this reason, for example, in the above example, if this command exists before the part where the machining program is to be created, context information stating "the part where the program is to be created will be created using a machining center-type command" is added.

[0270] Another example is a function called constant peripheral speed control. This function increases the spindle speed so that the peripheral speed is constant depending on the machining position (radius from the spindle center). Therefore, there is a risk that the spindle speed may increase too much near the spindle center. Therefore, when using this function, it is recommended to use a command to clamp the maximum spindle speed in advance. Therefore, if there is no speed setting command for clamping the spindle before the machining program is created in response to a request to "create a machining program with constant peripheral speed control," the content of "create a machining program with a spindle clamp speed setting command" may be added as a command content or as a partial request.

[0271] <Destination Information> The input generation unit 12 sets destination information for each input data set corresponding to each of the multiple partial requests, depending on the characteristics of the multiple partial requests. The destination information can include, in addition to the artificial intelligence 4, an artificial intelligence 4 different from the artificial intelligence 4, or a software function providing unit other than the artificial intelligence 4.

[0272] The destination information may include information that can identify the destination in accordance with the communication method used by the transmitter / receiver 13. Here, the destination does not simply refer to an entity that provides resources for running the artificial intelligence 4 or software, such as a computer, server, or cloud, but also includes applications running on those entities. For example, if the destination exists on an IP (Internet Protocol) network, the destination information may include information such as an IP address or port number. In this case, the IP address corresponds to information for identifying the entity, and the port number corresponds to information for identifying the application.

[0273] The destination information may be uniquely determined for each input data set. Alternatively, multiple destination information candidates may be configured as a list, and one piece of destination information in the list may be identified by an identifier such as a label. Alternatively, such an identifier may be set in the input data set.

[0274] Furthermore, multiple destinations may be set for one input data set, and multiple identifiers may be set for one input data set. In the list, a group identifier that groups together multiple pieces of destination information may be further associated, and the group identifier may be set for the input data set, thereby setting multiple pieces of destination information for the input data set.

[0275] As mentioned above, when multiple partial requests are generated for different request categories, it is possible that the characteristics of the AI ​​4 that is best suited to responding to each partial request will differ, even if the partial requests were originally generated from a single request. In this case, it is necessary to select a different AI 4 as the destination for each partial request depending on the characteristics of the partial request.

[0276] Here, the different artificial intelligences 4 as transmission destinations include, for example, an artificial intelligence 4 that receives natural language as input and generates responsive conversational sentences, and an artificial intelligence 4 that receives natural language as input but can also receive data such as images and generates images, 3D models, videos, etc. In other words, from the perspective of differences in the input / output systems of the artificial intelligences 4, the different artificial intelligences 4 are included in the different artificial intelligences 4.

[0277] The different artificial intelligences 4 as destinations are not limited to those mentioned above. For example, artificial intelligences 4 of the same sentence generation system but with different language model formats as their base, or artificial intelligences 4 of the same language model but with different numbers of parameters or versions, are included in the category of different artificial intelligences 4. Examples of artificial intelligences 4 that are based on different language model formats include GPT, BERT, and ELMo (Embeddings from Language Models). Furthermore, even if two identical artificial intelligences 4 are configured with different operating resources (such as servers, computers, or the number of allocated CPUs or GPUs), they are also included in the category of different artificial intelligences 4.

[0278] Furthermore, some partial requests may contain content for which it is not necessarily optimal to request a response from the artificial intelligence 4. For example, when creating a machining program, if a CAD (Computer-Aided Design) model is used as input and a machining program is created to machine the shape shown in the CAD model, it is more reasonable to use CAM (Computer-Aided Manufacturing) software. In such cases, it is necessary to set a software function providing unit, such as software other than the artificial intelligence 4 or a web application, as the destination of the partial request.

[0279] The artificial intelligence 4 or application to be the destination may be provided not only outside the information processing system but also inside the information processing system. Figure 33 is a diagram showing a configuration example in which the destination is provided inside the information processing system in embodiment 3.

[0280] The information processing system 1H shown in FIG. 33 includes a terminal device 2H and a back-end device 3H. The terminal device 2H has a configuration similar to that of the terminal device 2G shown in FIG. 30. The back-end device 3H has a configuration similar to that of the back-end device 3G shown in FIG. 30. Furthermore, the back-end device 3H includes an artificial intelligence 42 and software 51. Each of the artificial intelligence 42 and software 51 is a transmission destination provided inside the information processing system 1H. The transmitter / receiver 13 of the back-end device 3H is connected to each of the artificial intelligence 41, the artificial intelligence 43, the software 52, and the web application 6. In FIG. 33, the web application is referred to as "Web App." Each of the artificial intelligence 41, the artificial intelligence 43, the software 52, and the web application 6 is a transmission destination provided outside the information processing system 1H. Each of the software 51, the software 52, and the web application 6 functions as a software function providing unit.

[0281] When the information processing system 1H is linked to an artificial intelligence 4 or information search system that specializes in specialized knowledge, or an artificial intelligence 4 or information search system that uses internal information that is not appropriate to be disclosed to the public, it is desirable that the artificial intelligence 4 or information search system be provided within the information processing system 1H, like the artificial intelligence 42 and software 51. This can reduce the risk of information leakage, etc.

[0282] As described above, the input generation unit 12 sets transmission according to the characteristics of each input data set corresponding to a partial request.

[0283] Next, specific destinations will be described using examples. For example, a specific destination for programming would be an artificial intelligence 4 such as LLM if the programming language is a common language such as C or Python. It is expected that ladder programs written in ST language can also be supported as destinations. On the other hand, in the case of machining programs, if the machining program is generated from a CAD model, CAM software would be a specific destination. If the machining program is generated from text, similar to programming, an artificial intelligence 4 such as LLM would be a specific destination.

[0284] Next, when performing an information search, artificial intelligence 4 such as LLM can handle general information, but is not suitable for searching niche, highly specialized information. In such cases, options for the destination include artificial intelligence 4 based on a "small language model" that has been trained to collect only specialized knowledge, or a RAG configured to be able to access a database of specialized knowledge. Also, database software or applications that have databases of specialized knowledge built using SQL or the like and that are searchable can also be options for the destination.

[0285] As can be seen from the above explanation, the destination can be selected according to the task, i.e., the type of request. For example, if the above identifier groups are assigned to programming, information retrieval, data creation, etc., it becomes possible to select the destination according to each request category.

[0286] <Order Determination Unit 34> When there are multiple partial requests, the order determination unit 34 determines the order of the multiple partial requests. When there are no partial requests and the request is a single request, processing by the order determination unit 34 is not necessary.

[0287] The order determination unit 34 basically determines the order of partial requests in the same order as the original requests. For example, the order needs to be determined again for partial requests added by the information supplementation unit 35. In this case, whether the supplemented information is to be placed before or after the target partial request is determined depending on whether it supplements prerequisite information or conditions for the partial request to be supplemented, or whether it adds information or conditions to the results obtained for the partial request. The order information determined in this way is output as partial request order information.

[0288] <Request management unit 36> The request management unit 36 ​​controls the transmission of input data sets to the artificial intelligence 4 for each input data set corresponding to each of the multiple partial requests. Specifically, the request management unit 36 ​​controls the order in which the multiple partial requests are processed and the progress management of the processing. The request management unit 36 ​​also controls the output by the transmission / reception unit 13 to the answer generation unit 14 of the first partial answer received by the transmission / reception unit 13 for each partial request.

[0289] The request management unit 36 ​​controls the order in which the partial requests are processed so that the input data sets corresponding to the partial requests are processed in an order based on the partial request order information. In this case, the request management unit 36 ​​controls the transmission / reception unit 13 to transmit the input data sets to the destinations set for each input data set. If there is an order constraint, the request management unit 36 ​​waits to receive a response to the partial request, and then assigns the response information to the input data set before causing the transmission processing by the transmission / reception unit 13 to start.

[0290] If there is no such order constraint and the partial requests have different destinations, the request management unit 36 ​​may control the transceiver 13 to start transmitting the next partial request without waiting for the reception of a response to the partial request processed in the previous order. This allows the information processing system 1F to process each partial request in parallel, enabling efficient processing. After all first partial answers have been received, the first partial answers may be integrated and output to the answer generation unit 14 as a first answer. This prevents the first partial answers from being determined to be inappropriate in various verification processes in the answer generation unit 14.

[0291] Furthermore, if each first partial answer is an answer to a partial request in a different request category, the first partial answers are independent of each other, and the verification process in the answer generation unit 14 may also be independent. In such a case, the first partial answers may be output as is to the answer generation unit 14, and after the answer generation unit 14 performs a verification process on each first partial answer, the answer generation unit 14 may convert the answer format to integrate the first partial answers into a second answer. In this way, the information processing system 1F can perform the processing of the transmitter / receiver 13 and the processing of the answer generation unit 14 in advance from the part of the first answer for which the answer by the artificial intelligence 4 has been completed, thereby enabling efficient processing.

[0292] The information processing system 1F includes the request management unit 36, which allows transmission and reception to be performed in consideration of order constraints between partial requests, and also allows transmission and reception to be performed in parallel when there are no order constraints. This enables the information processing system 1F to generate more advanced answers and shorten the time required to generate answers.

[0293] <Answer generation unit 14> Based on the request category information, the answer generation unit 14 changes the processing in one or more of the verification of basic rules, verification of functional specifications, verification of operation, and conversion of the answer format to a second answer.

[0294] <Specific Example> The series of operations of the information processing system 1F described above will be explained using a specific example. For example, changes to the ladder program may be required depending on the situation. A machine tool already installed and operating in a factory may be modified, such as by adding or removing an axis, or changes may be made to connected peripheral devices, such as measuring instruments and associated measurement systems, or to transport devices and transport systems such as gantry loaders, pallet changers, and pallet pools. In such cases, not only hardware changes but also system and software changes are required. From the perspective of the ladder program, these devices will not operate correctly unless the contact information corresponding to the changes and the function blocks that operate in accordance with that contact information are also changed.

[0295] In this situation, suppose a request is input saying, "I want to delete the XXX function block in the existing ladder program and replace it with the YYY function block." First, the request receiving unit 11 detects "delete" and "replace" as verb information, and then detects "XXX function block" and "YYY function block" as object information for the verb information.

[0296] Next, the partial request generator 32 generates a first partial request, "Delete the XXX function block of the existing ladder program" from "Delete," and a second partial request, "Replace with the YYY function block." The request determiner 31 also sets the creation of a ladder program as the request category information based on the keyword "ladder program." The DB selector 33 selects a specification DB related to the ladder program.

[0297] The input generation unit 12 extracts information indicating the processing content of the YYY function block from the selected specification DB. The input generation unit 12 also selects an existing ladder program from the stored data as a data file. The first input data set corresponding to the first partial request includes the information "Delete the XXX function block of the existing ladder program" in the prompt, and the existing ladder program as a data file. Note that the prompt may include other information and text, but the description thereof will be omitted here.

[0298] The second input data set corresponding to the second partial request includes the information "Replace with YYY function block" as a prompt. The second input data set also includes a first partial answer to the first input data set as a data file. In practice, the first partial answer is included as a reference, not as a data file. Here, including the first partial answer as a data file refers to including the first partial answer as a reference. Furthermore, the same artificial intelligence 4 is set as the destination of the first input data set and the destination of the second input data set.

[0299] Due to the dependency between the first input data set and the second input data set, the request management unit 36 ​​suspends transmission of the second input data set until the transceiver unit 13 receives the first partial answer for the first input data set. After the first partial answer is received, the request management unit 36 ​​starts transmission of the second input data set, which includes the first partial answer as a data file.

[0300] When the second partial answer for the second input data set is received, the answer generation unit 14 performs a function check process corresponding to the YYY function on the target ladder program. If the result of the function check process is satisfactory, the information processing system 1F presents the second partial answer as the second answer via the answer presentation unit 15.

[0301] In addition to ladder programs, there are also cases where modifications to a machine tool control device are desired using executable software modules written in a high-level language such as C and compiled as necessary by a compiler or the like. For example, some commercially available machine tool control devices have customizable open functions and areas available to users of the control device, including the machine tool manufacturer. The information processing system 1F according to the third embodiment can also be used when utilizing such customization functions. The general flow is similar to that of the ladder program described above, so a description thereof will be omitted, and only the characteristic parts of this specific example will be described.

[0302] In this case, the requirement might be, for example, "Create XXX processing in C, compile the source, and add the resulting object file to the control device." When the requirement is converted into multiple partial requirements in the same manner as in the ladder program above, the requirement is divided into three partial requirements. The three partial requirements are the first partial requirement, "Create XXX processing in C," the second partial requirement, "Compile the source," and the third partial requirement, "Add the object file to the control device."

[0303] The first partial request is not explained here because, although it is a ladder program and written in C, it is similar to the above-mentioned ladder program in terms of programming by the artificial intelligence 4. The keyword "compile" for the second partial request sets the software function provider, specifically compiler software, as the destination rather than the artificial intelligence 4. The keyword "add-on" for the third partial request sets the above-mentioned customization function as the destination software function provider.

[0304] In addition, by predefining keywords that are linked to specific functions, such as "add-on," the processing content in the information processing system 1F, such as setting the destination, can be performed with predefined content.

[0305] For each first partial answer created in the above manner, the answer generator 14 may perform, for example, static analysis of the C language source code, error checking of the compilation results, or error checking upon loading. Furthermore, the answer generator 14 may perform operation verification by virtually executing processing.

[0306] Next, an example of the operation of the information processing system 1F in a case where a question is input as a request to the request receiving unit 11 will be described. For example, suppose that a question such as "What is the M code for normal rotation of the spindle?" or "What is the M code for stopping the coolant?" is input as a request to the request receiving unit 11. First, the request receiving unit 11 determines "What (is)?" as a verb, and detects "M code for normal rotation of the spindle" or "M code for stopping the coolant" as the object.

[0307] Next, the input generation unit 12 references specifications related to the M-code from the specification DB and includes the M-code for normal spindle rotation or the M-code for coolant stop as example information in the prompt. At this time, the input generation unit 12 may also reference setting parameters and switch between the specification DB to be referenced and the M-code specifications based on information indicating the type of machine tool, the manufacturer of the machine tool, or the manufacturer of the machine tool control device. This allows the input generation unit 12 to include more accurate example information in the prompt.

[0308] Here, a description of the subsequent operations in this case will be omitted. By generating an input data set by including the correct answer information stored in the specification DB in the prompt, the information processing system 1F is expected to significantly reduce the possibility that the artificial intelligence 4 will output an incorrect answer.

[0309] Next, an example of the operation of the information processing system 1F in a case where a request to create shape data is input to the request receiving unit 11 will be described.

[0310] One thing to be careful of when using machine tools is mechanical interference. Mechanical interference occurs when moving mechanical structures in a machine tool come into contact with each other, or when a mechanical structure of a machine tool comes into contact with a mechanical structure of another device. For example, the spindle that holds a tool may come into contact with the table of the machine tool or an attached measuring device. In this case, the contacting parts may be broken, or damage may be caused to the drive system of the machine tool, such as the motor, feed mechanism, or guide mechanism.

[0311] Many numerical control devices have an interference check function to prevent machine interference. Various methods are used for the interference check function. One known interference check function, for example, holds a 3D model of the machine tool, tool, or workpiece, and checks for interference by simulating the movement of the 3D model. The 3D model of the workpiece, jig, or tool is a necessary element when performing such a simulation.

[0312] For example, suppose that a request such as "generate shape data of a jig," "create shape data of a workpiece," or "create shape data of a tool" is input to the request receiving unit 11. Since the processing in the request receiving unit 11 is simple, a description of the processing in the request receiving unit 11 will be omitted here.

[0313] Here, we will explain the case where the shape data to be created is shape data of a jig. For example, the jig may be designed in advance by the user using CAD or the like. In this case, it is assumed that drawing data of the jig for which shape data is to be created exists in the retained data. Note that the drawing data existing in the retained data may be either two-dimensional data (hereinafter referred to as "2D data") or 3D data.

[0314] For example, if a request input to the request receiving unit 11 includes information identifying drawing data, such as "jig with drawing No. xxx," the input generating unit 12 refers to the stored data and includes the drawing data identified by that information in the input data set.

[0315] The input generator 12 also selects a destination for the input data set depending on the format of the drawing data. For example, if the drawing data is CAD data, software that converts CAD data to STL (Stereolithography) data may be selected as the destination. Alternatively, if the drawing data is 2D data, software with a 3D conversion function, such as CAD software, may be selected as the destination.

[0316] The answer generation unit 14 converts the shape data obtained as the first answer into data in a format usable by the information processing system 1F. For example, if the first answer is STL data, the answer generation unit 14 analyzes the structure indicated in the STL data and converts the STL data into data in a format defined as shape data handled by the information processing system 1F, such as wireframe, surface, or solid. In this way, the answer generation unit 14 obtains a second answer by converting the shape data that is the first answer.

[0317] Next, we will explain the case where the shape data to be created is shape data of a workpiece. Since workpieces often have simple shapes, it is conceivable that an image of the workpiece taken by a camera is used as the shape data. Here, we assume that the workpiece has been photographed in advance by the user, and the image data of the workpiece is saved as retained data.

[0318] For example, suppose a request such as "Create workpiece shape data from image xxx and image yyy" is input to the request receiving unit 11. Here, "image xxx" and "image yyy" are names given to the images. In this case, since information specifying the image is included in the request, the input generating unit 12 refers to the stored data and includes the image data specified by that information in the input data set.

[0319] Alternatively, when the request determination unit 31 determines that the request category is "generation of workpiece shape data," the information processing system 1F may request the user to specify an image of the workpiece for which shape data is to be created. For example, assuming that the information processing system 1F is provided with the information notification unit 22 described in the second embodiment, the request management unit 36 ​​or the input generation unit 12 requests the user to specify an image via the information notification unit 22. In this case, the input generation unit 12 includes image data of the image specified by the user in the input data set.

[0320] The input data set may be sent to various web services or software that generate three-dimensional shape data from images. The format of the shape data obtained as the first answer varies depending on the destination. Therefore, as in the case of the jig, the answer generation unit 14 converts the shape data obtained as the first answer into data in a format that can be used by the information processing system 1F. In this way, the answer generation unit 14 obtains the second answer by converting the shape data that is the first answer.

[0321] Next, a case will be described in which the shape data to be created is tool shape data. Data indicating the three-dimensional shape of a tool is often provided by the tool manufacturer. Therefore, the information processing system 1F may add, for example, a partial request for downloading tool shape data from the tool manufacturer's website via the information supplementation unit 35. In this case, information indicating the tool manufacturer and the tool model number are required. Furthermore, the retained data may contain a certain amount of tool data, which is data about the tool. Even if the request input to the request receiving unit 11 does not include the tool manufacturer and tool model number, it is sufficient if the request includes information that can be used to specify a tool from the retained data in which tool data is set.

[0322] The input generation unit 12 references the stored data to obtain information about the specified tool. This allows the input generation unit 12 to include information about the specified tool in the input data set corresponding to the partial request added by the information complementation unit 35. The destination of this partial request is the website or web server of the tool manufacturer. The first partial response to this partial request is 3D data of the tool provided by the tool manufacturer. The subsequent operations in this case are the same as those in the case of a jig or workpiece, and therefore will not be described here.

[0323] In this manner, the user can obtain shape data for a desired object by inputting a request to the information processing system 1F.

[0324] <Effects> According to the third embodiment, the information processing system performs one or more of generating a plurality of partial requests based on a request, determining the order of the plurality of partial requests, and completing information missing from the request. By generating a plurality of partial requests based on a request, the information processing system can avoid problems such as the artificial intelligence 4 ignoring some of the content when a complex request or a request containing a large amount of information is input. By determining the order of the plurality of partial requests, the information processing system can input the plurality of partial requests to the artificial intelligence 4 in the correct order. By completing information missing from the request, the information processing system can prevent a decrease in the accuracy of the answer due to a lack of information. This allows the information processing system to present an answer with even higher accuracy in response to a request from a user.

[0325] The input dataset also includes one or more of the following: one or more pieces of command information indicating the content of the instructions included in the request; context information including premise information for the command or restrictions on the command; and example information indicating examples of pairs of input to the artificial intelligence 4 and output from the artificial intelligence 4. By inputting such an input dataset to the artificial intelligence 4, it is possible to allow the artificial intelligence 4 to accurately grasp the content indicated by the request when creating a response to the request. This allows the information processing system to further improve the accuracy of the output from the artificial intelligence 4.

[0326] The information processing system also includes a request determination unit 31. The input generation unit 12 changes the input data generation process by switching the contents of one or more of the referenced specification DB, setting parameters, and stored data based on the request classification information. The answer generation unit 14 changes the process of one or more of the basic rule verification, functional specification verification, operation verification, and answer format conversion to a second answer based on the request category information. This allows the information processing system to present more accurate answers to user requests by changing the process by the input generation unit 12 and the process by the answer generation unit 14 depending on the content of the request.

[0327] The information processing system also includes a request management unit 36. By including the request management unit 36, the information processing system can transmit and receive partial requests while taking into account order constraints between partial requests, and can also transmit and receive partial requests in parallel when there are no order constraints. This enables the information processing system to generate more advanced answers and shorten the time required to generate answers.

[0328] Furthermore, the input generation unit 12 sets destination information for each input data set corresponding to each of the plurality of partial requests in accordance with the characteristics of the plurality of partial requests. The destination information can include, in addition to the artificial intelligence 4, an artificial intelligence 4 different from the artificial intelligence 4, or a software function providing unit other than the artificial intelligence 4. This allows the information processing system to reduce the risk of information leakage, etc.

[0329] The effects described in the third embodiment are each exhibited independently. Even when all of the configurations for obtaining each effect are present, it goes without saying that each individual effect is exhibited. Furthermore, additional effects can be obtained by combining the configurations.

[0330] The information processing system according to the third embodiment may include the notification information creation unit 21 and the information notification unit 22 described in the second embodiment. In this case, the information processing system can refine the notification information about the content of the processing by the input generation unit 12, which is the notification information shown in FIGS. 23 to 28 , into the processing content described in the third embodiment. The information processing system can also refine the content of the input data set as described in the third embodiment. The information processing system can notify the user of at least one of command information, example information, and context information included in the input data set by the information notification unit 22. The information processing system may link the generated partial request to the original request, thereby structuring and displaying the partial request and the request.

[0331] The information processing system may further link the notification information creation unit 21 and the information notification unit 22 with the request management unit 36, and control the operation of the request management unit 36 ​​in accordance with the information displayed by the information notification unit 22 and the user's operation in response to that display. With this configuration, the user can check the first answer and determine whether the answer is appropriate or inappropriate at the time the first answer is obtained. Furthermore, the information processing system can reflect the user's determination in the request management unit 36.

[0332] Furthermore, the operation of the request management unit 36 ​​may be controlled for each partial request or each first partial answer in accordance with the information displayed by the information notification unit 22 and the user's operation in response to that display. This configuration allows the user to individually determine whether each subdivided partial request or each subdivided first partial answer is appropriate or inappropriate. Furthermore, the information processing system can reflect the user's determination in the request management unit 36.

[0333] If the first answer is determined to be inappropriate at the time the answer is obtained, no further processing is required. Therefore, the information processing system may output the first answer at that time, or may suspend or terminate processing at that time. Furthermore, if only a specific first partial answer is determined to be inappropriate, the information processing system can remove unnecessary information in advance by having the answer generation unit 14 process only the first partial answer other than the relevant part.

[0334] If the first answer is determined to be appropriate at the time the first answer is obtained, the user can check the content of the answer at the time the first answer is obtained. This improves the transparency of the process leading up to obtaining the answer, and improves the user's trust in the information processing system utilizing the artificial intelligence 4.

[0335] The information processing system can further transmit additional content desired to be included in the input data set as an additional request to the input generation unit 12 via the request receiving unit 11. When an additional request is input, the information processing system may change or regenerate the input data set.

[0336] Embodiment 4. Fig. 34 is a diagram showing a first configuration example of an information processing system according to embodiment 4. Fig. 34 shows an information processing system 1I, which is a first configuration example of an information processing system according to embodiment 4. The information processing system 1I is configured in a terminal device 2I. Fig. 34 shows the terminal device 2I and an artificial intelligence 4 external to the information processing system 1I. In embodiment 4, the same components as those in embodiments 1 to 3 above are assigned the same reference numerals, and the following mainly describes configurations that differ from embodiments 1 to 3.

[0337] The terminal device 2I has a configuration similar to that of the terminal device 2A shown in Fig. 1. Furthermore, the terminal device 2I has a request management unit 36, a correction acceptance unit 41, and a correction detection unit 42. The terminal device 2I can be realized by a hardware configuration similar to that shown in Fig. 2.

[0338] 34, the components of the information processing system 1I are provided in a single device, that is, a terminal device 2I. The information processing system according to the fourth embodiment is not limited to one in which the components of the information processing system are provided in a single device. The components of the information processing system may be distributed across two or more devices that can function together.

[0339] Fig. 35 is a diagram showing a second configuration example of an information processing system according to embodiment 4. Fig. 35 shows an information processing system 1J, which is the second configuration example of the information processing system. The information processing system 1J is configured with a terminal device 2J and a back-end device 3J. That is, in the information processing system 1J, the components of the information processing system 1J are distributed between the terminal device 2J and the back-end device 3J.

[0340] The terminal device 2J has a configuration similar to that of the terminal device 2B shown in Fig. 3 and a correction detection unit 42. The back-end device 3J has a configuration similar to that of the back-end device 3B shown in Fig. 3 and includes a request management unit 36 ​​and a correction acceptance unit 41. The terminal device 2J can be realized by a hardware configuration similar to that shown in Fig. 2. The back-end device 3J can be realized by a configuration similar to that shown in Fig. 2 except for the input device 54 and the display device 55. Note that the manner in which the components of the information processing system are distributed is not limited to the manner shown in Fig. 35 and is arbitrary.

[0341] Next, an overview of each component of the information processing system according to embodiment 4 will be described. Below, each component of the information processing system 1I will be described using the information processing system 1I shown in Fig. 34 as an example. The description of each component of the information processing system 1I also applies to the information processing system 1J shown in Fig. 35. Here, the description of the same content as in embodiments 1 to 3 will be omitted.

[0342] In a configuration including the answer generation unit 14, various verification processes are performed in the answer generation unit 14 as described above, and the verification results are input to the correction receiving unit 41 and the request management unit 36. When the answer generation unit 14 obtains a verification result indicating inappropriateness, the information processing system 1I determines whether to redo the processing in the input generation unit 12. This determination may be made automatically by the request management unit 36. The information processing system 1I may inquire of the user via the answer presentation unit 15 about whether to redo the processing in accordance with the determination result, and the request management unit 36 ​​may make this determination based on the response. Note that in an information processing system including the information notification unit 22 shown in FIG. 21 or 22, the information processing system may inquire of the user via the information notification unit 22 about whether to redo the processing in accordance with the determination result, and the request management unit 36 ​​may make this determination based on the response.

[0343] The correction detection unit 42 functions after an answer (which may be either the first answer or the second answer) is presented, regardless of the presence or absence of the answer generation unit 14. As will be described in detail later, the correction detection unit 42 detects whether or not the user has changed the content of the answer presented by the answer presentation unit 15, that is, whether or not a correction has been made and the content of the change. The correction detection unit 42 outputs detection information, which is information indicating whether or not a correction has been made and the content of the change, to the correction acceptance unit 41.

[0344] The correction receiving unit 41 makes changes to the generation process of the input data set in the input generation unit 12 and to various information or data referenced during the generation process, based on the verification results from the answer generation unit 14 or the detection information from the correction detection unit 42. If the request management unit 36 ​​determines that the processing in the input generation unit 12 should be redone, the correction receiving unit 41 makes changes based on the detection information, and then the processing in the input generation unit 12 is restarted.

[0345] The above is a general flow of the operation of the information processing system 1I in the embodiment 4. Next, the details of each of the above components will be described. Note that the description of the contents that overlap with the first to third embodiments will be omitted.

[0346] <Correction Detector 42 > The correction detector 42 detects a correction operation for the second answer presented by the answer presenter 15 .

[0347] Ideally, all operations performed by a user are logged, and the details of the operations are analyzed to determine which parts of the text have been modified and how. In this case, the information processing system 1I can perform advanced and detailed modification detection. However, it is possible to determine the details of modifications even without such advanced and detailed modification detection. For example, the information processing system 1I can detect a triggering operation that is presumed to be the user's final use of the answer result, record the details of the operation at that time as the final answer state, and compare the originally presented second answer with the recorded details of the operation to determine the details of the modification.

[0348] The correction detection unit 42 determines whether or not a correction has been made based on a comparison result between the final answer state and the originally presented second answer. The correction detection unit 42 transmits the determination result of whether or not a correction has been made and information indicating the difference between the final answer state and the originally presented second answer to the correction acceptance unit 41 as correction detection information.

[0349] For example, if the request input to the information processing system 1I is a request for the creation of a machining program, the operation of executing the machining program or the operation of ending editing can be the above-mentioned trigger, and the content of the machining program at that time becomes the final answer state.The correction detection unit 42 then compares the final answer state with the machining program created in the original second answer, and determines that no corrections have been made if there are no differences.The correction detection unit 42 determines that there have been corrections if there are differences.The correction detection unit 42 also determines that corrections have been made to the parts that differ.

[0350] <Answer Generation Unit 14> If the answer generation unit 14 determines that correction is necessary based on the results of the verification, it notifies the correction receiving unit 41 of the content of the correction, and if it determines that correction is not necessary, it creates a second answer.

[0351] When the answer generation unit 14 obtains a verification result indicating inappropriateness in various verification processes, the answer generation unit 14 transmits the verification result indicating inappropriateness to the correction receiving unit 41. The answer generation unit 14 verifies basic rules, functional specifications, or operations, and therefore acquires information indicating which specific parts were determined to be inappropriate and for what reasons. Such verification result information corresponds to the information on the processing results of the answer generation unit 14 contained in the notification information described in the second embodiment. Therefore, the answer generation unit 14 transmits the verification result information to the correction receiving unit 41. The verification result information transmitted to the correction receiving unit 41 includes information indicating the parts determined to be inappropriate and information on the content of the determination or the reason for the determination. The answer generation unit 14 also transmits the verification result information to the request management unit 36.

[0352] <Request management unit 36> When the answer generation unit 14 obtains a verification result indicating inappropriateness in various verification processes, the request management unit 36 ​​determines whether or not to redo the processing in the input generation unit 12. Such a determination may be made automatically by the request management unit 36. The information processing system 1I may inquire of the user via the answer presentation unit 15 or the information notification unit 22 shown in FIG. 21 or 22 about the determination result and whether or not to redo the processing, and the request management unit 36 ​​may make such a determination based on the response result.

[0353] If it is determined that the processing by the input generation unit 12 should be redone, the request management unit 36 ​​notifies the input generation unit 12 of the redo of the processing. The input generation unit 12 waits for the change processing by the correction receiving unit 41 (described later) to be completed, and starts regeneration of the input dataset, thereby creating an input dataset that reflects the verification results.

[0354] The above-mentioned inquiry to the user and the redoing of processing by the input generation unit 12 based on the answer result may be performed for each partial request or each first partial answer. In this case, the information processing system 1I can regenerate the input data set for each problematic part of the request, rather than for the entire request, and can efficiently generate an answer that satisfies the request.

[0355] <Correction Accepting Unit 41> The correction accepting unit 41 accepts corrections based on the content of the correction operation detected by the correction detecting unit 42. The correction accepting unit 41 changes the processing content in the input generating unit 12 based on at least one of the correction detection information and the verification result information. That is, the correction accepting unit 41 may make the change based on either the correction detection information or the verification result information, or may make the change based on both the correction detection information and the verification result information. The correction detection information and the verification result information may each include multiple pieces of information.

[0356] When the correction receiving unit 41 changes the processing content of the input generating unit 12 based on the correction detection information, the correction receiving unit 41 updates the example information or context information used to generate the input dataset for the original request using information on the portion of the final answer state where a difference occurs from the original second answer. For example, if example information or context information is searched for and output from a table or database using a keyword included in the request, the correction receiving unit 41 replaces the information in the table or database corresponding to the keyword with information on the portion where a difference occurs as an update process. Alternatively, the correction receiving unit 41 may add information in the table or database corresponding to the keyword as an update process. In other words, by including the final answer state in the input dataset as an example of the correct answer to be sought, it is expected that the accuracy of the output by the artificial intelligence 4 will improve when a request similar to or identical to this request is input next time or thereafter.

[0357] Here, we will explain how to change the processing content in the input generation unit 12, taking as an example a tool center point control function required for machine tools with rotary axes, such as five-axis machining centers. Normally, the offset from the base of the tool to the tool center point that actually performs machining is expressed as a one-dimensional vector. In this case, the offset vector is a one-dimensional vector. In machine tools with rotary axes, the tool attitude changes depending on the angle of the rotary axis, and the offset vector changes three-dimensionally. This tool center point control function controls this offset vector so that the user of the machine tool is not aware of it; the user simply needs to specify the coordinates of the tool center point.

[0358] This tool center point control function has two formats depending on the tool orientation command method. The two formats are the G43.4 command, which commands the angle of the rotation axis, and the G43.5 command, which commands the tool orientation using a vector. For example, suppose a request is made to "create a machining program with tool center point commands" and a response created using the G43.4 command is presented. If the final response state is subsequently changed from the G43.4 command to the G43.5 command, it is considered more appropriate in the user's environment to prioritize the G43.5 command over the G43.4 command for tool center point control. Therefore, it is conceivable to change the information in the specification database, for example, regarding which of the G43.4 and G43.5 commands should be prioritized for tool center point control.

[0359] When the correction receiving unit 41 changes the processing content in the input generating unit 12 based on the verification result information, it updates the example information or context information used to generate the input data set for the original request, using the parts determined to be inappropriate and information on the determination content or reason. In particular, when the change is made based on the verification result information, by updating the restrictions in the context information, it is expected that the answer generating unit 14 will generate an answer that is not determined to be inappropriate when a request similar to this one is input next time or later.

[0360] More specifically, as an update process, the correction receiving unit 41 adds, as data in a table or database, context information, particularly the inverted content of the reason for the determination in the verification result information as a restriction, corresponding to, for example, a keyword included in the request. For example, if the reason for the determination in the verification result information is information that "double-byte characters were found outside the comment area (the area enclosed by "(" and ")") of the processing program," the inverted content of that information would be "double-byte characters must be enclosed in the comment area" or "double-byte characters must not be placed outside the comment area." These items become restrictions.

[0361] <Effects> According to the fourth embodiment, the information processing system includes a correction detection unit 42 and a correction acceptance unit 41. If the answer generation unit 14 determines that a correction is necessary based on the results of the verification, it notifies the correction acceptance unit 41 of the content of the correction. If the answer generation unit 14 determines that a correction is not necessary, it creates a second answer. The information processing system can improve the accuracy of the output by the artificial intelligence 4 by updating the processing in the input generation unit 12 when the verification results determine that a correction is necessary. By updating the processing based on the detection result of the correction content by the correction detection unit 42, the information processing system can update the processing with reference to the content of the answer ultimately desired by the user. By updating the processing based on the verification results by the answer generation unit 14, the information processing system can update the processing with reference to content determined to be inappropriate in the previous answer. Furthermore, the information processing system can generate input with updated processing content without having to redo the request input by using the request management unit 36 ​​to determine whether to redo the input generation processing.

[0362] Here, a supplementary explanation will be given regarding the relationship between the information processing systems according to the first to fourth embodiments and the machine tools that are the targets of information processing by the information processing systems. Here, the general functions for controlling machine tools are referred to as numerical control functions. These numerical control functions include functions for controlling machines such as cutting machines, grinding machines, electric discharge machines, laser machines, and AM machines that are targets of general numerical control devices, as well as functions for controlling press machines, injection molding machines, industrial robots, etc.

[0363] The information processing systems according to the first to fourth embodiments may be provided with a numerical control function known to those skilled in the art. Here, the numerical control function known to those skilled in the art is, for example, a function for analyzing a machining program and generating time-series command information for various motors or amplifiers to realize the relative rotation or movement between a tool and a workpiece described in the machining program. In other words, this case is an example of a configuration in which the terminal device of the information processing system is configured by a numerical control device.

[0364] Furthermore, the information processing system may not be provided with a numerical control function that is well known to those skilled in the art, but may be configured to be able to cooperate and communicate with an external device having a numerical control function, such as a numerical control device. Being configured to be able to cooperate and communicate does not only mean being directly connected to an external device having a numerical control function, but also includes being able to connect to a storage area that stores the information and data handled by the external device.

[0365] If the information processing system does not have a numerical control function and is not capable of linking and communicating with an external device having the numerical control function, the information processing system will be unable to perform tasks using data (such as coordinate values ​​or variable values) created or updated by the numerical control function among the stored data. However, the information processing system can achieve all of the effects described in the first to fourth embodiments except for the tasks that it cannot perform.

[0366] The configurations shown in the above embodiments are examples of the contents of the present disclosure. The configurations of each embodiment can be combined with other known technologies. The configurations of each embodiment can also be combined as appropriate. Part of the configuration of each embodiment can be omitted or modified without departing from the gist of the present disclosure.

[0367] 1A, 1B, 1C, 1D, 1E, 1F, 1G, 1H, 1I, 1J Information processing system, 2A, 2B, 2C1, 2C2, 2D, 2E, 2F, 2G, 2H, 2I, 2J Terminal device, 3B, 3C, 3E, 3G, 3H, 3J Back-end device, 4, 41, 42, 43 Artificial intelligence, 51, 52 Software, 6 Web application, 11 Request acceptance unit, 12 Input generation unit, 13 Transmitting / receiving unit, 14 Answer generation unit, 15 Answer presentation unit, 21 Notification information creation unit, 22 Information notification unit, 31 Request determination unit, 32 Partial request generation unit, 33 DB selection unit, 34 Order determination unit, 35 Information complementation unit, 36 Request management unit, 41 Correction acceptance unit, 42 Correction detection unit, 50 Processing circuit, 51 Communication device, 52 Processor, 53 Memory, 54 input device, 55 display device.

Claims

1. An information processing system comprising: a request receiving unit that receives requests for work to be performed when using a machine tool; a specification database that stores information representing the specifications of the information processing system or the specifications of the machine tool or a device other than the machine tool that is an external device to the information processing system; an input generating unit that generates an input data set in response to the request by referencing at least one of setting parameters related to the settings of the information processing system or the settings of a device external to the information processing system, and stored data stored in a memory area accessible by the information processing system; and an answer presenting unit that presents an answer to the request, created based on the output of an artificial intelligence to which the transmitted input data set has been input.

2. The information processing system of claim 1, further comprising an answer generation unit that generates a second answer that is an answer to the request using a first answer that is an output by the artificial intelligence to which the input data set corresponding to the request has been input, wherein the answer presentation unit presents the second answer generated by the answer generation unit, and the requests accepted by the request acceptance unit include at least one of creation of a machining program, creation of a ladder program, creation or output of a data file related to the machine tool or work using the machine tool, and response to a question related to the machine tool or work using the machine tool, and wherein the answer generation unit performs at least one of verification of basic rules, verification of functional specifications, verification of operation, and conversion of the answer format to the second answer on the first answer.

3. The information processing system according to claim 2, further comprising: a notification information creation unit that creates notification information including one or more pieces of information from the content of processing by the input generation unit, the content of processing by the answer generation unit, the input data set, and the first answer; and an information notification unit that notifies the notification information.

4. An information processing system as described in any one of claims 1 to 3, characterized in that it performs one or more of the following: generating multiple partial requests based on the request; determining the order of the multiple partial requests; and supplementing information that is missing in the request.

5. The information processing system described in claim 4, characterized in that the multiple partial requests generated based on the request include a partial request for searching information contained in the specification database to be referenced by the input generation unit based on the request, and a partial request for including information obtained by the search based on the request in the input data set.

6. An information processing system as described in claim 4 or 5, characterized in that the input data set includes one or more of the following: one or more instruction information indicating the content of the instructions included in the request; context information including premise information of the instruction or restrictions on the instruction; and example information indicating examples of pairs of input to the artificial intelligence and output from the artificial intelligence.

7. The information processing system described in claim 6, characterized in that the setting parameters include one or more of the type of machine tool, information indicating the manufacturer of the machine tool, and information indicating the manufacturer of the machine tool's control device, and the specification database to be referenced is switched depending on the setting value of the setting parameters.

8. An information processing system as described in claim 2 or 3, characterized in that the data file includes one or more of the machining program, the ladder program, the setting parameters, the work procedure manual for the machine tool, shape data for each of the tools and jigs attached to the machine tool, and shape data for the object to be machined by the machine tool.

9. An information processing system as described in claim 3, further comprising a request determination unit that determines the category of the request and creates request category information, wherein the input generation unit changes the generation process of the input data set by switching the contents of one or more of the specification database, the setting parameters, and the retained data that are referenced based on the request category information, and the answer generation unit changes the process of one or more of the verification of the basic rules, the verification of the functional specifications, the verification of the operation, and the conversion of the answer format to the second answer based on the request category information.

10. An information processing system as described in claim 3, further comprising: a correction detection unit that detects correction operations on the presented second answer; and a correction acceptance unit that accepts corrections based on the content of the detected correction operations, wherein the answer generation unit notifies the correction acceptance unit of the content of the correction if it determines that a correction is necessary based on the results of verification, and creates the second answer if it determines that no correction is necessary.

11. The information processing system of claim 4 or 5, wherein the input generation unit sets destination information for each of the input data sets corresponding to each of the plurality of partial requests in accordance with the characteristics of the plurality of partial requests, and the destination information can include, in addition to the artificial intelligence, an artificial intelligence different from the artificial intelligence, or a software function providing unit other than the artificial intelligence.

12. The information processing system of claim 11, further comprising a request management unit that controls, for each of the input data sets corresponding to each of the plurality of partial requests, the transmission of the input data set to the artificial intelligence.

13. An information processing system as described in claim 2, characterized in that the conversion into an answer format indicating the second answer includes a conversion that makes the portion created by the artificial intelligence distinguishable.

14. The information processing system described in claim 13, characterized in that when the part created by the artificial intelligence contains a numerical value, converting it into an answer format indicating the second answer includes replacing the numerical value with a specific symbol or specific character, or deleting the numerical value.

15. An information processing system comprising: a request receiving unit that receives requests regarding work to be performed when using a machine tool; an answer generating unit that generates a second answer that is an answer to the request using a first answer that is an output by artificial intelligence to which an input data set corresponding to the request has been input; and an answer presenting unit that presents the generated second answer, wherein the requests received by the request receiving unit include at least one of creating a machining program, creating a ladder program, creating or outputting a data file related to the machine tool or work using the machine tool, and responding to questions related to the machine tool or work using the machine tool, and wherein the answer generating unit performs at least one of verifying basic rules, verifying functional specifications, verifying operation, and converting the answer format to the second answer for the first answer.

16. The information processing system according to claim 15, further comprising: an input generation unit that generates the input data set; a notification information creation unit that creates notification information including one or more pieces of information selected from the content of processing by the input generation unit, the content of processing by the answer generation unit, the input data set, and the first answer; and an information notification unit that notifies the notification information.

17. The information processing system of claim 15, wherein the data file includes one or more of the machining program, the ladder program, setting parameters related to the settings of the information processing system or the settings of devices external to the information processing system, a work procedure manual for the machine tool, shape data of tools attached to the machine tool, and shape data of objects to be machined by the machine tool.

18. An information processing system as described in claim 15, characterized in that the conversion into an answer format indicating the second answer includes a conversion that makes the portion created by the artificial intelligence distinguishable.

19. An information processing program that causes a computer to execute the following steps: receiving a request for work to be performed when using a machine tool; generating an input data set in response to the request by referencing at least one of a specification database that stores information representing the specifications of an information processing system or the specifications of the machine tool or a device other than the machine tool that is an external device to the information processing system, setting parameters related to the settings of the information processing system or the settings of a device external to the information processing system, and stored data stored in a memory area accessible by the information processing system; and presenting an answer to the request that was created based on the output of an artificial intelligence that received the transmitted input data set.

Images