Question answering device, question answering method, question answering program, question answering system, and industrial machine

Abstract

The present invention is a question answering device (10) that answers a question relating to an industrial machine, and is provided with: an information acquisition unit (101) that acquires question information received by a question reception unit; a determination processing unit (102) that determines whether or not the acquired question information includes unique information pertaining to the industrial machine; and an answer generation unit (103) that, when it is determined that the question information contains the unique information, generates an answer to the question information on the basis of a closed learning model that can be provided only to the user of the industrial machine. The closed learning model learns information related to the industrial machine and information related to unique information.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to a technology for answering a user's question, particularly a user who operates an industrial machine. BACKGROUND

[0002] User support in which an operator answers a user's question by mail or telephone is widely performed. In recent years, a response system that flexibly uses AI (Artificial Intelligence) instead of an operator has rapidly spread in relation to such user support. As the response system, there is an AI-type chat robot such as Chat GPT (Chat Generative Pre-trained Transformer) provided by OpenAI, which is an AI that has learned to answer in a dialogue form in response to a question. Such a chat robot can use publicly known information comprehensively, and thus can generate almost the same answer in response to a general question of a user.

[0003] The related technology is at least one communication and open / close loop control system for a filling system that includes a machine having a software communication robot (particularly a chat robot) configured to recognize a voice input and / or a text input of an operator and / or output or display information related to an operation state of the machine, and an open / close loop control device connected to the software communication robot for data communication and controlling the machine of the filling system in an open loop and / or a closed loop based on the voice input and / or the text input recognized by the software communication robot. (Patent Document 1 below).

[0004] PRIOR ART DOCUMENTS

[0005] PATENT DOCUMENTS

[0006] Patent Document 1: Japanese Patent Application Laid-Open No. 2021-532467 SUMMARY

[0007] PROBLEMS TO BE SOLVED BY THE INVENTION

[0008] The problem to be solved by the embodiments of the present application is to provide a technology capable of generating a good answer in response to various questions including a professional question of a user. Other problems and novel features can be understood from the description of the present specification and the drawings.

[0009] MEANS FOR SOLVING THE PROBLEMS

[0010] The question answering device of one embodiment generates an answer to question information of a user of an industrial machine based on a closed learning model that has learned about information related to the industrial machine and information related to inherent information of the industrial machine, which can be provided only to the user of the industrial machine, in a case where the question information of the user of the industrial machine includes the inherent information related to the industrial machine. BRIEF DESCRIPTION OF DRAWINGS

[0011] Figure 1 is a block diagram schematically showing a configuration of a question answering system of the first embodiment.

[0012] Figure 2 is a block diagram showing a hardware configuration of a question answering device of the first embodiment.

[0013] Figure 3 is a block diagram showing a hardware configuration of a control device provided in an injection molding machine of the first embodiment.

[0014] Figure 4 is a block diagram showing a functional configuration of a question answering device of the first embodiment.

[0015] Figure 5 is a block diagram showing a functional configuration of a control device provided in an injection molding machine of the first embodiment.

[0016] Figure 6 is a flowchart showing a question answering process of the first embodiment.

[0017] Figure 7 is a flowchart showing a question answering process of the first embodiment.

[0018] Figure 8 is a flowchart showing a question answering process of the first embodiment.

[0019] Figure 9 is a diagram for explaining an answer article to question information of the first embodiment.

[0020] Figure 10 is a flowchart showing a question correction process of the first embodiment.

[0021] Figure 11 is a diagram for explaining a correction process of question information of the first embodiment.

[0022] Figure 12 is a block diagram showing an application example of a question answering device of the first embodiment.

[0023] Figure 13 is a block diagram showing a functional configuration of a question answering device of the second embodiment.

[0024] Figure 14is a block diagram showing a functional configuration of a control device provided in the injection molding machine of the second embodiment.

[0025] Figure 15 is a flowchart showing a question and answer process of the second embodiment. DETAILED DESCRIPTION

[0026] Embodiments of the present application will be described below with reference to the drawings. The embodiments are not limited to the following. The following description and drawings are appropriately simplified in order to make the description clear. In each drawing, the same elements are denoted by the same reference numerals, and repeated description is omitted as necessary. In addition, there are parts in which hatching is omitted in order to avoid the drawings from being complicated.

[0027] <First Embodiment>

[0028] (Configuration of Question and Answer System)

[0029] The configuration of the question and answer system of the present embodiment will be described. Figure 1 is a block diagram schematically showing the configuration of the question and answer system of the present embodiment.

[0030] As shown in Figure 1 , the question and answer system 1 of the industrial machine of the present embodiment includes a question and answer device 10 and a plurality of injection molding machines 20A to 20N. In the case where the injection molding machines 20A to 20N are not distinguished from each other, it is described as an injection molding machine 20.

[0031] The question and answer device 10 of the industrial machine is an information processing device such as a server that can be communicatively connected to the plurality of injection molding machines 20 via a network NW such as the Internet. The question and answer device 10 acquires a question of the injection molding machine 20, specifically, a question input by a user of the injection molding machine 20 using a hardware I / F 240 (see Figure 3 ) described later, generates an answer to the question, and provides it to the user. The question and answer device 10 is installed with an automatic conversation program such as a chat robot that applies a dialogue-type AI, and an answer is generated by the program. The answer of the question and answer device 10 is generated on the basis of a closed-type learning model 11 and / or an open-type learning model 12. The closed-type learning model 11 is a machine learning model that learns information stored in a closed-type database 160 as a data set. In addition, the open-type learning model 12 is a machine learning model that learns information stored in an open-type database 170 as a data set.

[0032] The injection molding machine 20 has a mold clamping device, etc. provided with a heated cylinder, a screw, a mold, and is an industrial machine that obtains a desired resin molded product by injecting resin in a molten state into a mold after resin pellets in a hopper connected to the heated cylinder are molten and mixed. The injection molding machine 20 is provided with a control device 200 that controls the operation of the aforementioned series of devices. The control device 200 of the present embodiment also receives a question from a user, appropriately performs question correction, etc., and transmits to the question answering device 10, and in the case where a response to the question from the question answering device 10 is received, reports the response to the user.

[0033] (Hardware configuration)

[0034] The hardware configuration of the question answering device 10 and the control device 200 described above will be described in detail. Figure 2 is a block diagram showing the hardware configuration of the question answering device of the present embodiment. Figure 3 is a block diagram showing the hardware configuration of the control device provided in the injection molding machine of the present embodiment.

[0035] As shown in Figure 2 , the question answering device 10 includes a CPU (Central Processing Unit) 110, a ROM (Read Only Memory) 120, a RAM (Random Access Memory) 130, a communication I / F (Inter Face) 140, a storage device 150, a closed-type database 160, and an open-type database 170.

[0036] The CPU 110 executes by spreading the BIOS (Basic Input / Output System) read from the ROM 120 as a non-volatile storage area, the OS (Operating System) read from the storage device 150, general application programs, various programs on the RAM 130 as a volatile storage area. As the programs, in addition to the automatic conversation program described above, for example, a question answering program that assumes a part of the question answering processing (industrial machine question answering method) described later, etc. The communication I / F 140 has a not-shown gateway, etc., and is used for communication with the injection molding machine 20 via a network NW.

[0037] The storage device 150 is, for example, an HDD (Hard Disk Drive), an SSD (Solid State Drive), or the like, and stores various information used in the question-and-answer processing. As the stored information, there is, for example, a list indicating the correspondence between the identification code and the closed-type learning model 11. The identification code will be described later. The closed-type database 160 and the open-type database 170 are storage areas that contain each different information (part of the information can be duplicated) used in the answer to the question. These databases are constructed by, for example, an HDD, an SSD, or the like.

[0038] The closed-type database 160 is a storage area that contains information related to the industrial machine (in the present embodiment, the injection molding machine 20) and information related to the inherent information, and is used as the dataset of the closed-type learning model 11. The information related to the injection molding machine stored in the closed-type database 160 is, for example, the content of the instruction manual of the injection molding machine 20, information that can be provided only to the user of the injection molding machine 20 (including the enterprise to which the user belongs) (in other words, information with high confidentiality). Therefore, the closed-type learning model 11 learned on the basis of the closed-type database 160 is associated with the identification code (identification information) that inherently indicates the user and / or the injection molding machine 20 to which the question is asked.

[0039] The identification code is information that inherently identifies the injection molding machine 20A to 20N, in other words, information for identifying the user (an individual or a group such as an enterprise to which the individual belongs) who owns the injection molding machine 20. Therefore, in the case where the same user owns a plurality of injection molding machines 20 of different kinds, the identification codes of the plurality of injection molding machines 20 can be the same, or the kind of the injection molding machine 20 and the identification code can be associated and stored in the storage device 150. In addition, the identification code becomes information that indicates the closed-type learning model 11 inherent to the user.

[0040] In the present embodiment, the closed-type learning model 11 inherent to each user is prepared, and the closed-type database 160 inherent to the user is associated with the closed-type learning model 11. Thereby, the user indicated by the identification code is given an answer based on the closed-type learning model 11 associated with the user. Therefore, the user is given an answer in a range that can be disclosed, and is not given an answer based on another closed-type learning model 11, in other words, is not given an answer in a range that cannot be disclosed.

[0041] In particular, as highly confidential information, it can provide solutions for each error code, information on the molding conditions (operational conditions) of the injection molding machine 20 for obtaining the desired molded product, the specifications of the injection molding machine 20, and measured values ​​during molding in the injection molding machine 20. Multiple solutions corresponding to the molding conditions of the injection molding machine 20 are prepared for each error code.

[0042] Specifically, the closed-loop learning model 11 learns the relationship between molding condition information, specification information, error codes, information indicating abnormalities, information related to the molded product and / or measured value information, and the molding condition information and / or the answer in the instruction manual, which are presented as answers. Information indicating abnormalities mainly refers to abnormalities not shown in the error code, such as abnormal sounds emanating from the user's question regarding whether an abnormality exists. Furthermore, information related to the molded product mainly refers to questions about situations where the user cannot obtain the desired molded product. Additionally, it can envision situations where the user prioritizes production, energy conservation, or mechanical lifespan. Therefore, if the question includes what the user prioritizes, it is preferable to establish a correlation between the answer pattern (e.g., changes in molding conditions) and the above relationships so that molding condition information conforming to that priority can be included in the answer document.

[0043] Specifications for the injection molding machine include the heating cylinder inner diameter, screw type (L / D), servo motor performance (rated output, rated torque, rated speed, etc.), mold clamping mechanism performance, and the machine's usage history. Molding conditions for the injection molding machine 20 include the heating cylinder temperature control, injection speed, screw speed during metering, back pressure, clamping force of the mold clamping device, mold opening and closing speed, material type, and material supply. Measured values ​​include the detection results from various sensors in the injection molding machine 20, such as temperature sensors and timers. Examples include heating cylinder temperature measurement, molding action information, injection speed, metering completion position, screw rotation speed during metering, clamping force, and mold opening and closing speed. For example, when obtaining specification information, molding condition information, and measured value information (action information) along with error codes as questions, the question-and-answer device 10 can accurately answer the response method for the error code corresponding to the specification information, molding condition information, and measured value information by referring to the closed-loop learning model 11. Preferably, the response method includes information such as how to change the value of the molding condition. By providing such a response method, an accurate answer corresponding to the user's current situation can be provided.

[0044] The inherent information is used to determine how to answer a question using the closed-loop learning model 11. In this embodiment, error codes, specifications of the injection molding machine 20, molding condition information, and measured value information can be cited. Since there are multiple molding condition information items, it can be reduced to specific molding conditions. It should be noted that it is not limited to error codes, specifications, molding condition information, and measured value information; any information that corresponds to generating an answer using the closed-loop learning model 11 is acceptable. For example, the type (model) of the injection molding machine 20 and statements related to the instruction manual, such as words in the instruction manual and symbols recorded in the instruction manual, can also be used as inherent information.

[0045] The closed-loop learning model 11 learns after receiving user feedback following an answer, improving the accuracy of the answer through machine learning such as reinforcement learning, supervised learning, deep learning, and supplementary learning based on that feedback. Preferably, the closed-loop database 160 is also updated each time through this learning process. During learning, not only feedback on the feedback is provided, but also feedback on the process leading to the question and answer. Therefore, the closed-loop learning model 11 and the closed-loop database 160 are customized individually for each user. It should be noted that by establishing a correspondence between the closed-loop learning model 11 and the closed-loop database 160 for each user individually, the results of different learning performed by each user are not shared among users. Therefore, information leakage from the closed-loop database 160 can be prevented.

[0046] On the other hand, the open database 170 is a storage area containing low-secrecy information related to the injection molding machine in this embodiment; in other words, it is publicly known information that can be provided not only to the user of the injection molding machine 20 but also to various other users. The open database 170 is used as the dataset for the open learning model 12. Therefore, the open learning model 12 is a model that has learned the relationships of publicly known information and is used when asking publicly known questions. The open database 170 can use databases such as Chat GPT provided by OpenAI.

[0047] Information is appropriately added to both the closed database 160 and the open database 170. The information added to the open database 170 is publicly known, while information related to industrial machinery (here, injection molding machine 20) is added to the closed database 160. For example, the closed database 160 is automatically updated by the question-and-answer device 10 periodically accessing external information and collecting equipment information from industrial machinery manufacturers via the network NW. This is not limited to equipment manufacturers; it is preferable to do so when new insights related to component manufacturers or industrial machinery are discovered. For example, when new insights are discovered, the maintenance and operation company of the question-and-answer device 10 can appropriately add these new insights to the closed database 160. It should be noted that the closed database 160 can also be automatically updated by accessing a dedicated server that collects new insights from maintenance and operation companies. When information is added to both the closed database 160 and the open database 170 in this way, the closed learning model 11 and the open learning model 12 undergo relearning.

[0048] like Figure 3 As shown, the control device 200 includes a CPU 210, ROM 220, RAM 230, hardware I / F 240, a user interface consisting of a display 241, an input device 242, a microphone 243 and a speaker 244, a communication I / F 250, and a storage device 260.

[0049] CPU 210 executes the BIOS read from ROM 220 (a non-volatile memory area), the OS read from storage device 260, general-purpose applications, and various programs, which are then loaded onto RAM 230 (a volatile memory area). The programs include parts responsible for query-and-answer processing, as described later.

[0050] Hardware I / F 240 is used for communication between the display 241, input device 242, microphone 243, and speaker 244 and the user interface. Specifically, hardware I / F 240 is used to receive user inquiries and report responses to users. The display 241 displays the inquiry and the response as a string. Input device 242 includes a keyboard, mouse, etc., and receives user input. Microphone 243 receives user voice inquiries. Speaker 244 reports responses to inquiries via voice.

[0051] The user interface of these injection molding machines 20 is the question receiving unit of the present invention, which accepts questions from users via voice or keypad input. Additionally, the user interface is also the question answering unit of the present invention, conveying the answer generated by the answer generation unit 103 of the question answering device 10 (described later) to the user via voice from the display 241 or speaker 244. Furthermore, the user interface of the injection molding machine 20 helps solve the problems of the invention. In this embodiment, the user interface, including the display 241, input device 242, microphone 243, and speaker 244, which constitute the question receiving unit and question answering unit of the question answering device of the present invention, is provided in each injection molding machine 20. However, the user interface can also be shared by multiple injection molding machines 20 as a portable device. A user interface dedicated to the injection molding machine 20 and equipped with a portable question receiving unit and question answering unit can be considered part of the injection molding machine 20. Examples of such portable devices (i.e., user interfaces) include tablet computers such as smartphones and laptop PCs (personal computers). Furthermore, desktop PCs installed in user factories can also have the functions of a question receiving unit and a question answering unit.

[0052] The communication I / F250, including a gateway (not shown), is used to communicate with the question-and-answer device 10 via the network NW. The storage device 260, such as an HDD (Hard Disk Drive) or SSD (Solid State Drive), stores various information used in the question-and-answer processing. This stored information includes identification codes, appended information, and answer documents sent from the question-and-answer device 10. The appended information is added during the question correction and determination process described later; details of this will be provided later.

[0053] (Functional Composition)

[0054] The functional configuration of the question-and-answer device 10 and the control device 200 is described in detail. Figure 4 This is a block diagram illustrating the functional configuration of the question-and-answer device according to this embodiment. Figure 5 This is a block diagram illustrating the functional configuration of the control device included in the injection molding machine of this embodiment.

[0055] like Figure 4 As shown, the question-and-answer device 10 includes an information acquisition unit 101, a judgment processing unit 102, an answer generation unit 103, a verification processing unit 104, an answer sending unit 105, and a learning and updating unit 106. These functions are implemented through the coordinated operation of the hardware of the question-and-answer device 10.

[0056] The information acquisition unit 101 acquires various information sent from the injection molding machine 20. The acquired information includes question information representing user questions and evaluation information representing user feedback on the answers to the questions. The decision processing unit 102 performs various decisions during the answer generation process. These decisions may include, for example, whether the question information contains inherent information or whether question information correction is necessary.

[0057] The answer generation unit 103 is a so-called AI-type chatbot that generates answers to questions based on a closed-type learning model 11 and / or an open-type learning model 12. It should be noted that this is not limited to an AI-type chatbot; any automated response (automatic conversation) program capable of using various learning models to generate answers to questions is acceptable. The verification processing unit 104 verifies the answers to determine whether corrections should be made to the generated answers.

[0058] The response sending unit 105 sends the generated and adjusted response to the user interface of the injection molding machine 20, which is equipped with the question information sending source. The learning update unit 106, upon acquiring information for machine learning, performs machine learning based on that information and updates the closed-loop learning model 11 and the closed-loop database 160. The information used for machine learning includes evaluation information and condition change information. The condition change information includes measured values ​​obtained when settings are changed from those contained in the response in the injection molding machine 20, and the changed settings.

[0059] like Figure 5 As shown, the control device 200 includes, as a function, an information acquisition unit 201, a conversion processing unit 202, an information transmission unit 203, a judgment processing unit 204, a reporting unit 205, a monitoring processing unit 206, and a query correction unit 207. These functions are realized through the coordinated operation of the hardware of the control device 200.

[0060] The information acquisition unit 201 acquires input from the user and various information sent from the question-and-answer device 10. User input includes questions and comments. Acquired information includes identification codes and answers generated by the question-and-answer device 10. The conversion processing unit 202 converts the voice-input question into an article.

[0061] The information sending unit 203 sends question information and evaluation information to the question-and-answer device 10. The judgment processing unit 204 performs various judgments in the answer generation process. These judgments include, for example, whether evaluation information has been input, and whether monitoring processing is enabled. Monitoring processing monitors the movement of the injection molding machine 20 after receiving an answer to the question, collecting information such as various settings and detection results obtained from various sensors. The reporting unit 205 reports the answer to the question to the user via voice or text. The monitoring processing unit 206 performs monitoring processing. The question correction unit 207 corrects the question information in the question correction judgment processing described later.

[0062] (Question and Answer Processing)

[0063] The question-and-answer processing performed by the question-and-answer system of this embodiment will be described in detail. Figures 6-8 This is a flowchart illustrating the question-and-answer processing of this embodiment. In this process, a question button is displayed on the various setting screens of injection molding on the display 241 (question reception unit) of the injection molding machine 20, and the selection of this button is used as a trigger. It should be noted that this is not a limitation; for example, the control device 200 may always collect sound through the microphone 243 (question reception unit) and execute this process when a specific voice command is received.

[0064] First, such as Figure 6 As shown, the information acquisition unit 201 acquires a voice question from the user as voice information (S101). Specifically, the information acquisition unit 201 uses the microphone 243 to collect the voice spoken by the person. Voice collection begins, and after a certain period of time following a voice interruption, the collected voice is acquired as voice information representing the question. It should be noted that if, after the voice collection begins, the voice is not interrupted within the preset collection time or no voice is collected, the reporting unit 205 may also use the speaker 244 to report an error or use the display 241 to convey the information to the user in text form. In this case, the question-and-answer process ends.

[0065] After acquiring the voice information, the conversion processing unit 202 performs text conversion on the acquired voice information to generate question information (S102). It should be noted that the voice-to-text conversion process can be performed using existing voice-to-text conversion programs such as AI-based voice recognition. Furthermore, at this time, existing natural language processing can be used to pre-correct the question information to form textual content. It should be noted that the question can also be acquired as text input by the user using the input device 242. In this case, the input text can be directly set as the question information. After the question information is generated, the information acquisition unit 201 acquires the identification code stored in the storage device 260 (S103).

[0066] After the identification code is acquired, a determination is made as to whether the generated question information needs to be corrected. If so, a question correction determination process (S104) is performed. Details of the question correction determination process will be described later. After the question correction determination process, the information sending unit 203 sends the question information and the identification code to the question answering device 10 (S105), and the control device 200 enters a standby state during the question answering process.

[0067] The question-and-answer device 10 enters a standby state, capable of continuously acquiring various information. If a question and identification code are sent in this state, the information acquisition unit 101 of the question-and-answer device 10 acquires (receives) the question and identification code (S106). After acquisition, the determination processing unit 102 determines whether inherent information is included in the question (S107).

[0068] If it is determined that inherent information is contained in the question information (S107 is yes), the information acquisition unit 101 acquires the identification code received along with the question information (S108). After acquisition, the answer generation unit 103 reads the list stored in the storage device 150 and selects the closed-loop learning model 11 corresponding to the acquired identification code (S109). After selection, the answer generation unit 103 generates an answer document corresponding to the question information based on the selected closed-loop learning model 11 (S110).

[0069] The response document generated using the closed-loop learning model 11 preferably includes more detailed explanations of countermeasures for the error codes, especially those described in the instruction manual, if the question information contains error codes. Additionally, if the question information includes molding condition information, measured value information, or specification information, a response derived from those conditions and measured values ​​is generated. This response includes appropriate changes to the molding conditions.

[0070] Especially when the error code is included in the query message, it is preferable to suggest multiple modified values ​​for the molding conditions that can avoid the abnormal state indicated by the error code. Furthermore, when the aforementioned priorities are included in the query message, the molding conditions according to those priorities are included in the response document. It should be noted that even if such priorities are not included in the query document, multiple response modes can be suggested separately. Moreover, it is preferable to respond to the candidates for molding conditions in descending order of their likelihood of not hindering the operation of the injection molding machine.

[0071] On the other hand, if it is determined that the question information does not contain inherent information (S107 is no), the answer generation unit 103 selects the open learning model 12 (S111), and in step S110, generates an answer document corresponding to the question information based on the open learning model 12. In the case where inherent information is not contained, since it can be determined that a general question with low concealment has been asked, a sufficient answer can be obtained from the open learning model 12. It should be noted that, for example, even if there are unclear aspects regarding obtaining an answer that is expected to be very satisfactory to the user, the question information is corrected through the question correction determination process described in detail later to obtain a high-precision answer that satisfies the user.

[0072] like Figure 7 As shown, after the question document is generated, the determination processing unit 102 determines whether the generated answer document can be verified (S112). This determination is based on whether the answer document contains molding conditions that allow for verification processing. Verification processing is a so-called simulation processing using molding conditions. If the answer document contains the molding conditions required for this simulation processing, it is determined that the generated answer document can be verified. It should be noted that, in addition to determining whether the answer document contains molding conditions that allow for verification processing, it is also possible to determine whether the question information corresponding to the answer document contains specification information for verification processing. In this case, if both determination results are yes, it is determined that verification processing can be performed, and simulation processing using molding conditions and specification information is performed.

[0073] If the generated response document can be verified (S112 is yes), the verification processing unit 104 performs verification processing (S113) and obtains a verification result. After the verification result, the determination processing unit 102 determines whether the response document needs to be corrected (S114). This determination is based on factors such as whether the verification result has caused problems, such as the molded product not typically being of good quality or errors occurring (e.g., inconsistencies arising from changes in molding conditions, physical damage to the equipment, etc.). Information about the quality of the molded product can be pre-stored in the storage device 150 as data used in the verification process, or it can be obtained through the closed-loop learning model 11 or the open-loop learning model 12. Alternatively, information about the quality of the molded product can be included in the response document. Furthermore, it is preferable that the obtained verification result is used as supervisory data for learning the closed-loop learning model 11 at a predetermined time. At this time, the weights of the neural network of the closed-loop learning model 11 can be adjusted only by methods such as backpropagation.

[0074] If it is determined that the answer document needs to be revised (S114 is yes), the answer generation unit 103 revises the answer document (S115) and performs the verification process of step S113 again. Here, the revision is preferably determined based on the verification result, i.e., the change in molding conditions. The correspondence between the verification result and the change can be pre-stored in the storage device 150. Alternatively, the priority order of the types of pre-changed molding conditions can be determined. This priority order can also be changed according to the type of resin and the type of injection molding machine 20. Furthermore, based on the question information including the verification result, the closed-loop learning model 11 or the open-loop learning model 12 is used again to generate the answer.

[0075] On the other hand, if it is determined that the answer document does not need correction (S114 is no), and if it is determined that the generated answer document cannot be verified (S112 is no), the answer sending unit 105 sends the answer document to the injection molding machine 20, the source of the question information (S116). After sending, the question and answer device 10 enters a standby state. It should be noted that when the sent answer document is generated using the closed-type learning model 11, it is preferable to store it together with the identification code in the storage device 150. The reason for this is that the answer document is used for subsequent learning of the closed-type learning model 11.

[0076] After transmission, the information acquisition unit 201 of the injection molding machine 20 acquires (receives) the response document (S117). After acquisition, the reporting unit 205 reports the response document to the user (S118). As part of this report, the response document is displayed on the display 241, and the response document is transmitted via voice using the speaker 244. Therefore, the user interface, such as the display 241 and the speaker 244, helps to solve the problems of the present invention.

[0077] Following the report, the determination processing unit 204 determines whether the user has asked a follow-up question (S119). For example, it may display a text message indicating whether a follow-up question exists, and display a window with selectable "YES" and "NO" buttons on the display 241 as a GUI (Graphical User Interface), and determine whether the button is selected. Alternatively, it may use the speaker 244 to verbally indicate whether a follow-up question exists, and determine whether the user's voice response (e.g., a voice like "yes" or "no," indicating a follow-up question) is received using the microphone 243 within a specified period.

[0078] If it is determined that a user has asked a question again (S119 indicates yes), proceed... Figure 6 The processing of voice question input in step S101 is shown. On the other hand, if it is determined that there is no further question from the user (S119 is not), as...Figure 8 As shown, the determination processing unit 204 determines whether a user has provided an evaluation (S120). This determination is made after the determination in step S119 by checking whether an evaluation has been obtained by the information acquisition unit 201 within a specified time period. For example, a window with an "evaluate" button is displayed on the display 241. Corresponding to the selection of this button, a window with "resolved" (positive review) and "unresolved" (negative review) buttons is displayed on the display 241 as an evaluation. An evaluation is obtained corresponding to the selection of the "resolved" or "unresolved" button, and it is determined that an evaluation has been provided.

[0079] Alternatively, the evaluation can be obtained from the text or voice input after selecting the "Evaluate" button. Furthermore, multi-level evaluations, such as 5 levels, are possible, in addition to "Resolved" and "Unresolved". Preferably, the time limit for this determination is set to a time sufficient for the injection molding machine 20 to operate corresponding to the answer document and for the user to recognize the evaluation result. Such a time could be, for example, a few minutes to a few hours. The evaluation time varies depending on the content of the question. Therefore, it is preferable to store the obtained answer document in the storage device 260 in advance, so that the evaluation can be performed at any time. In this case, it is preferable that steps S120 and S121, and steps S126 to S129 (described later) executed as a result of these steps, are separated from the question-and-answer processing and can be performed as evaluation processing at any time.

[0080] If it is determined that the user has provided an evaluation (S120 is yes), the information sending unit 203 sends the evaluation information indicating the evaluation result along with the identification code to the question and answer device 10 (S121), and proceeds to step S122. On the other hand, if it is determined that the user has not provided an evaluation (S120 is no), the process in step S121 is skipped and the process proceeds to step S122.

[0081] In step S122, the determination processing unit 204 determines whether the monitoring processing is set to ON (S122). The monitoring processing is the process by which the monitoring processing unit 206 monitors the operation of the injection molding machine 20 and collects molding conditions and measured values ​​during operation (injection molding). The ON / OFF of the monitoring processing can be switched during the operation of the injection molding machine 20 by using the selectable "ON" and "OFF" buttons included in the screen displayed on the display 241.

[0082] If the monitoring process is set to ON (S122 is yes), the determination processing unit 204 determines whether the user has changed the molding conditions (S123). This determination is made after obtaining the response document by checking whether the molding conditions have been changed within a preset change determination time. If the user has changed the molding conditions (S123 is yes), the determination processing unit 204 determines whether the changed molding conditions are OK, i.e., whether the molding conditions have become a problem (S124). This determination is made by checking whether molding has been performed a predetermined number of times under the changed molding conditions. Alternatively, the user can input that the molding conditions are not a problem through button selection or other means, and the determination processing unit 204 will determine this by checking whether such input has been made.

[0083] If the modified molding conditions are determined to be NG (meaning there is a problem with the molding conditions) (S124 is 'No'), since it is anticipated that the user may change the molding conditions again independently, the determination processing unit 204 performs the determination process in step S124 again, repeating this determination until it is determined that the molding conditions are not problematic. It should be noted that since continuous monitoring is performed, the determination process in step S124 can also be triggered by another change in molding conditions. Furthermore, during this stage, the user may ask another question. If a second question is received, this process can be interrupted or terminated, and question information can be generated and sent. In this case, in the question-and-answer device 10, it is preferable to manage the received question as a question associated with the previous question.

[0084] If the modified molding conditions are determined to be OK (S124 is OK), the information sending unit 203 sends condition change information, including the modified molding conditions and the identification code, to the question and answer device 10 (S125), and the question and answer processing on the injection molding machine 20 side ends. Preferably, the condition change information includes the modified conditions and all information that can be collected during monitoring. On the other hand, if the monitoring processing is determined to be OFF (S122 is No), and if the user determines that the molding conditions have not changed (S123 is No), the question and answer processing on the injection molding machine 20 side ends. It should be noted that the case where the user determines that the molding conditions have not changed refers to the case where the elapsed time after the acquisition of the answer document exceeds the aforementioned change determination time.

[0085] When evaluation information and an identification code are sent to the question-and-answer device 10, the information acquisition unit 101 of the question-and-answer device 10 acquires (receives) the evaluation information and acquires the identification code (S126, S127). After acquisition, the determination processing unit 102 determines whether the answer document corresponding to the evaluation information uses the closed-loop learning model 11 (S128). The storage device 150 stores the identification code associated with the answer document. Therefore, it is possible to determine whether the answer document corresponding to the evaluation information is a document generated using the closed-loop learning model 11 based on the identification code acquired along with the evaluation information.

[0086] If the answer document corresponding to the judgment and evaluation information uses the closed-type learning model 11 (S128 is yes), the learning update unit 106 updates the closed-type learning model 11 based on the evaluation information (S129). For example, this update can be supplementary learning of the closed-type learning model 11 based on the evaluation information and the answer document. In this case, it is preferable that the closed-type database 160 is also updated. After the update, the question-and-answer device 10 enters a standby state (S130). On the other hand, if the judgment storage device 150 does not store answer documents or other answer documents corresponding to the evaluation information and does not use the closed-type database 160 (S128 is no), the process proceeds to the standby state in step S130.

[0087] When the question-and-answer device 10 receives condition change information and an identification code, the information acquisition unit 101 of the question-and-answer device 10 acquires (receives) the condition change information and acquires the identification code (S131, S132). After acquisition, the determination processing unit 102 determines whether the answer document corresponding to the condition change information is a document generated using the closed-type learning model 11 (S133). The answer document and the identification code are stored together in the storage device 150. Therefore, it is possible to determine whether the answer document corresponding to the condition change information is a document generated using the closed-type learning model 11 based on the identification code acquired along with the condition change information.

[0088] If it is determined that the answer document corresponding to the condition change information uses the closed-type learning model 11 (S133 is yes), the learning update unit 106 updates the closed-type learning model 11 based on the condition change information (S134). For example, this update could be supplementary learning of the closed-type learning model 11 based on the condition change information and the answer document. At this time, the preferred closed-type database 160 is also updated. After the update, the question-and-answer processing on the question-and-answer device 10 side ends. On the other hand, if it is determined that no answer document is stored in the storage device 150, or that the answer document corresponding to the condition change information does not use the closed-type learning model 11 (S133 is no), the question-and-answer processing on the question-and-answer device 10 side ends.

[0089] This is an example of a document that provides an answer to a question. Figure 9 This is a diagram used to illustrate the answers to questions regarding this embodiment. Figure 9 The attached diagram shows a reference numeral Qu representing the question information, and reference numerals An1 to An3 representing different answer documents for the question information Qu. The question information Qu contains error code A1, specification information, molding condition information, and measured value information as inherent information. Therefore, a closed-form learning model 11 is used to generate the answer documents for such a question information Qu. In this example, the hidden information in the question information Qu is different, thus generating answer documents An1 to An3.

[0090] According to response document An1, an appropriate solution including molding conditions is provided for the servo motor overload abnormality, which is error code A1. According to response document An2, it is indicated that the user cannot eliminate the overload abnormality by forcibly using the injection molding machine 20 that is normally used, given the model of the injection molding machine 20 shown in the question document, thus providing an appropriate solution. In addition, according to response document An2, the injection molding machine 20 owned by the user can be identified based on the user's identification code. Therefore, the response includes information that the problem would not occur if the user owned a different type of injection molding machine 20. In addition, according to response document An3, it is indicated that the injection molding machine 20 needs to be checked. These responses are appropriate and accurate because the question information Qu contains inherent information, especially specification information, molding condition information, and measured value information. However, even if the question information Qu does not contain specification information, molding condition information, or measured value information, the question-and-response device 10 can at least refer to the content described in the instruction manual to provide an accurate response, and can obtain a response faster than the time required for the user to refer to the instruction manual themselves.

[0091] The question correction and judgment process is explained in detail. User questions can be subjective, and the subject-verb relationship can be unclear. In these cases, even directly inputting the user's question into the closed-loop learning model 11 of the question-and-answer device 10 may not yield a suitable answer. In such situations, an open-loop learning model with high language model processing capabilities, such as Chat GPT, is used to organize the question, or it is converted and processed by a dedicated conversion processing unit 202 before being input into the closed-loop learning model 11. This process is also referred to as prompt engineering. Figure 10 This is a flowchart illustrating the question correction determination process of this embodiment.

[0092] like Figure 10As shown, firstly, the determination processing unit 204 of the injection molding machine 20 determines whether additional information needs to be saved (S201). The additional information to be saved here can include explanatory text detailing various error codes, specification information of the injection molding machine 20 from which the query originated, such as its type, model, and product name, explanatory text of specialized terminology related to the injection molding machine 20, and explanatory text of the instruction manual for the error codes. Therefore, if the query information contains error codes, specialized terminology related to the injection molding machine 20, specialized terminology or symbols recorded in the instruction manual, or basic information about the injection molding machine 20, it is determined that additional information needs to be saved.

[0093] If it is determined that additional information needs to be saved (S201 is yes), the query correction unit 207 retrieves the required information from the storage device 260 (S202) and corrects the query information (S203). For example, Figure 9 The section "Model number of the injection molding machine 〇〇…" in the third paragraph from the top of the question information Qu shown is specification information that has been corrected as additional saved information. In other words, this section is an automatically added text in response to the user's verbal question, intended to be replaced with linguistic information and input into a conversational closed-loop learning model. After correction, the determination processing unit 204 determines whether the question information requires molding conditions and / or measured values ​​(S204). This determination is made by judging whether the question information contains specific keywords. Error codes can be identified as keywords. It should be noted that all error codes are preferred, but only some error codes are also possible. Some error codes can be identified, such as those that provide a wide range of options for answering without molding conditions or measured values, making it difficult to obtain a highly accurate answer.

[0094] If it is determined that the query information requires molding conditions and / or measured values ​​(S204 is yes), the determination processing unit 204 determines whether the molding conditions and / or measured values ​​are settings that can be sent to the query response device 10 (S205). Sending molding conditions and / or measured values ​​to external devices is often confidential information for the company. Therefore, even when sending information to the query response device 10 for information security management, it is preferable to be able to switch the sending setting to ON / OFF. For example, to change the settings of the injection molding machine 20, a switching button can be selectively displayed on the setting screen shown on the display 241. The switching button can also be configured to switch independently according to molding conditions and measured values. The ON / OFF switching of the sending setting can be performed corresponding to the selection of these buttons. Therefore, the determination in step S205 is performed by determining which of the sending settings is set to ON / OFF.

[0095] If the molding conditions and / or measured values ​​are determined to be settings that can be sent to the question-and-answer device 10 (S205 is yes), the question correction unit 207 acquires the molding conditions and measured values ​​at the current moment when the settings can be sent (S206). After acquisition, the question correction unit 207 adds the acquired molding conditions and / or measured values ​​to the question information (S207), and the question correction determination process ends. For example, Figure 9 The section "Operating status of the injection molding machine before the question..." in the fourth paragraph from the top of the query information Qu shown represents the additional calibration molding conditions and measured values. This section is preferably like... Figure 9 As shown, this is transformed into language information and input into a conversational, closed-loop learning model.

[0096] On the other hand, if it is determined that the molding conditions and / or measured values ​​are not set to be sent to the question-and-answer device 10 (S205 is no), and if it is determined that the question information does not require molding conditions and / or measured values ​​(S204 is no), the question correction determination process ends. Additionally, if it is determined that no additional information needs to be saved (S201 is no), the process proceeds to the determination process in step S204.

[0097] Figure 11 This is a diagram illustrating the correction process for the question information used to explain this embodiment. Figure 11 The attached diagrams, labeled Qu1 to Qu3, represent query information, and the process is corrected as the last number increases. Query information Qu1 is the uncorrected query information generated initially. While the error code is known here, the specific error is unknown. Therefore, if no error code is set in the inherent information, an answer using the open database 170 will be obtained, but there is a high probability of not receiving an answer or receiving an incorrect answer.

[0098] On the other hand, question information Qu2, after question correction and determination processing, becomes information with error codes added as supplementary information. In this state, even if the answer from the open database 170 is used, although a corresponding answer is obtained, the likelihood of obtaining a satisfactory answer for the user is low. Therefore, question information Qu3, after question correction and determination processing, becomes information with user manual content corresponding to the error codes added as supplementary information. In this state, even if the answer from the open learning model 12 is used, it can be said that a satisfactory answer for the user can be obtained to some extent. Furthermore, when shaping conditions and measured values ​​are added to the question information through question correction and determination processing, it becomes... Figure 9 The question information Qu is shown. If this state is met, a highly accurate answer that satisfies the user will be obtained.

[0099] It should be noted that if the question information originally contains an error code and the error code is set in the inherent information, then because the closed-loop learning model 11 has already learned (and stored in the closed-loop database 160) the error code information and the information in the instruction manual, even for question information Qu1, it can generate the same answer as in the case of accepting question information Qu3 without correction. Therefore, the feature that can correct the shaping conditions and / or measured values ​​is particularly useful in the question correction determination process.

[0100] According to the embodiment described above, it is possible to provide good answers to users' professional questions. Furthermore, compared to users referring to the instruction manual themselves, answers can be obtained much faster. Moreover, for questions including molding conditions and measured values, since the closed-loop learning model 11, which uses learned professional insights, generates answers, more specific and accurate answers can be obtained. Furthermore, since the closed-loop learning model 11 is prepared specifically for each user, it prevents confidential information such as technical secrets from leaking to other users. Similarly, the questioning, answering, and feedback on the answer evaluation are only performed on the closed-loop learning model 11, thus continuously improving the accuracy of the learning model's answers while preventing confidential information from leaking to external parties.

[0101] Furthermore, the generated answers are validated as much as possible, thus avoiding sending incorrect answers to users. Because the questions undergo appropriate question correction and judgment processing, the content of the questions can be enriched, resulting in highly accurate answers.

[0102] In addition, users can ask questions verbally or receive answers through voice or text, thus easily obtaining responses to their questions, just like having a skilled technician by their side, making it extremely convenient.

[0103] It should be noted that in this embodiment, the question-and-answer system 1, which answers questions from users of the injection molding machine 20, has been described. However, the user is not limited to the user of the injection molding machine 20. For example, it can be the user of all industrial machinery such as mixing equipment, extruders, stamping equipment, and semiconductor manufacturing equipment. In addition, the aforementioned industrial machinery such as the injection molding machine 20 also includes its peripheral equipment. Furthermore, complete sets of equipment that include at least one or more of the mixing equipment, extruders, and stamping equipment are also included in the industrial machinery of the present invention. In this case, the closed-loop database 160 of the question-and-answer device 10 of the question-and-answer system 1 naturally stores information related to various industrial machinery, and the closed-loop learning model 11 learns based on the information related to various industrial machinery.

[0104] Furthermore, in this embodiment, the content of performing question correction determination processing in the control device 200 of the injection molding machine 20 has been described. However, question correction determination processing can also be performed in the question answering device 10. In this case, a question correction unit 207 is provided for functionality in the question answering device 10, and additional saved information is pre-stored in the storage device 150. In addition, in step S205, the question answering device 10 confirms whether the injection molding machine 20, the source of the question, can send molding conditions and / or measured values. Then, if the molding conditions and / or measured values ​​can be sent, the molding conditions and / or measured values ​​are sent to the question answering device 10 as a response from the injection molding machine 20, and the question answering device 10 receives the response. The question correction unit 207 provided in such a question answering device 10 constitutes part of the question answering device of the present invention. In addition, the question correction unit 207 can also be provided in Figure 5 The middle part of the question-and-answer device 10 that connects the information acquisition unit 101 and the determination processing unit 102.

[0105] In this embodiment, the content sent to the question-and-answer device 10 after generating and correcting the question information is described in the case of a single inquiry. However, it is also possible to return a question to the user to enrich the question information based on the content of the question. For example, when receiving the question "What should I do if an abnormality occurs during measurement?", the determination processing unit 204 determines whether an error code is included based on the word "abnormality". If it is determined that no error code is included, the reporting unit 205 can report to the user, "Is an abnormality number displayed on the screen? If an abnormality number is displayed, please provide the number (0 digit)." Thus, it is possible to include an error code in the question. This can also prompt the user to answer other questions about the status of the injection molding machine 20. Therefore, the operation part in which the question correction unit 207 of the question-and-answer device returns a question to the user to enrich the information and sends it to the answer generation unit of the question-and-answer device is equivalent to the prompt engineering unit.

[0106] Furthermore, in this embodiment, the question-and-answer device 10 is described as having an open learning model 12 and an open database 170. However, it may also have only a closed learning model 11 and a closed database 160 instead of an open one. That is, the closed database 160 may contain various publicly known information so that general questions can also be answered by the closed learning model 11. In other words, the open database 170 may be embedded in the closed database 160.

[0107] It should be noted that in the question correction determination process of this embodiment, it is explained that the molding conditions and measured values ​​need to be set to be able to be sent to the question answering device 10 in order to obtain them. However, the molding conditions and measured values ​​can also be automatically obtained when the operator asks a question verbally. In addition, as described above, when the question answering device 10 has a question correction unit 207 to perform the question correction determination process on the question answering device 10 side, the molding conditions and measured values ​​of the injection molding machine 20 can be confirmed after the question information is obtained. For example, when the question correction unit 207 of the question answering device 10 requests the molding conditions and measured values ​​from the injection molding machine 20, it sends the molding conditions and measured values ​​set by the injection molding machine 20 at that moment to the question answering device 10.

[0108] In addition, such as Figure 12 As shown, the question-and-answer device 10 can also have a first question-and-answer device 10-1, which consists of hardware such as a closed database 160 and a CPU 110, and a second question-and-answer device 10-2, which consists of hardware such as an open database 170 and a CPU 110, located in different locations and interconnected via communication I / F through the Internet or the like. For example, by storing a conversational AI such as a chatbot like Chat GPT, which is the open database 170, on a server of the operating company, that server can also function as the second question-and-answer device 10-2. On the other hand, the first question-and-answer device 10-1, which consists of hardware such as a server such as a closed database 160 and a CPU 110, can also be configured in a manufacturer that manufactures the injection molding machine 20. In addition, the first question-and-answer device 10-1, which consists of hardware such as a closed database 160 and a CPU 110, can also be configured in a factory of a user that operates the injection molding machine 20.

[0109] Alternatively, the functions of the information acquisition unit 101, judgment processing unit 102, answer generation unit 103, verification processing unit 104, answer sending unit 105, and learning update unit 106 of the question-and-answer device 10, as well as each learning model, can be installed in the injection molding machine 20. In this case, through the coordinated operation of the hardware resources of the injection molding machine 20, the various functions of the question-and-answer device 10 described above can be realized, and the injection molding machine 20 is constructed as the question-and-answer device of the present invention. Of course, only a portion of the functions of the first question-and-answer device 10-1, such as the server constituting the closed database 160 CPU 110, etc., which constitutes the question-and-answer device 10, can be installed in the injection molding machine 20.

[0110] <Second Implementation>

[0111] Figure 13 This is a block diagram illustrating the functional configuration of the question-and-answer device according to this embodiment.Figure 14 This is a block diagram illustrating the functional configuration of the control device included in the injection molding machine of this embodiment.

[0112] In this embodiment, the question-and-answer system 1 has a question-and-answer device 10A instead of a question-and-answer device 10, which differs from the first embodiment. Additionally, the injection molding machine 20 has a control device 200A instead of a control device 200, which also differs from the first embodiment.

[0113] like Figure 13 As shown, the question-and-answer device 10A differs from the question-and-answer device 10 of the first embodiment in that it includes an open-type answer generation unit 103A and a closed-type answer generation unit 103B instead of an answer generation unit 103. For example... Figure 14 As shown, the control device 200A also includes an information extraction unit 208, which is different from the control device 200 of the first embodiment.

[0114] The open-type answer generation unit 103A in the question-and-answer device 10A is an AI-type chatbot that generates answers to known question information (described later) based on the open-type learning model 12. On the other hand, the closed-type answer generation unit 103B is an AI-type chatbot that generates answers to question information based on the closed-type learning model 11. Furthermore, the closed-type answer generation unit 103B merges the two obtained answers and generates an answer document that is sent to the injection molding machine 20.

[0115] The information extraction unit 208 in the control device 200A extracts known information from the question information and generates known question information.

[0116] (Question and Answer Processing)

[0117] The question-and-answer processing of the question-and-answer system in this embodiment will be described in detail. Figure 15 This is a flowchart illustrating the question-and-answer processing of this embodiment. The question-and-answer processing of this embodiment adds and modifies some processes compared to the question-and-answer processing of the first embodiment. Here, only the added and modified processes will be described.

[0118] like Figure 15As shown, after the question correction determination process, the information extraction unit 208 of the control device 200A generates known question information obtained by extracting known information from the question information (S301). Here, the known information is obtained by subtracting inherent information from the question information. That is, information with high confidentiality, such as error codes, molding condition information, measured value information, and specification information, is removed from the question information and extracted as known information. It should be noted that, according to the question correction determination process, detailed explanations are added to the question information for error codes. Preferably, these detailed explanations are also processed as known information. Figure 11 The "(Error Code A1)" and other information contained in the question information Qu2 shown are not processed as publicly known information and are not extracted. After generating publicly known question information, the information sending unit 203 sends the question information, the publicly known question information, and the identification code to the question answering device 10A (S302).

[0119] The question-and-answer device 10A enters a standby state, always capable of acquiring various types of information. When a question, a known question, and an identification code are sent in this state, the information acquisition unit 101 of the question-and-answer device 10 acquires (receives) this content (S303). After acquisition, the open-type answer generation unit 103A generates an open-type answer document (first answer) corresponding to the known question based on the open-type learning model 12 (S304). After selecting the closed-type learning model 11, the closed-type answer generation unit 103B generates a closed-type answer document (second answer) corresponding to the question based on the closed-type learning model 11 (S305). After the closed-type answer document is generated, the closed-type answer generation unit 103B merges the obtained open-type answer document and the closed-type answer document to generate a single answer document (S306). It should be noted that the question information in the closed-type learning model 11 can be not only a closed-type answer document (second answer) corresponding to conversational question information, but also closed-type answer data for the question data. The aforementioned "question data" refers to measured value information (including resume information) and is a non-conversational question. Similarly, "answer data" refers to numerical values ​​such as forming conditions and is a non-conversational answer. In the aforementioned case, the closed-ended answer generation unit 103B also merges the obtained open-ended answer document with the closed-ended answer data to generate a single answer document (S306).

[0120] The merging here can refer to adding a closed-ended answer document (containing closed-ended answer data) to an open-ended answer document, or it can refer to a document after appropriate information organization. Information organization can include correcting the answer document style, deleting duplicate content, and deleting answers from the open-ended answer document side when it contains opposing answers. In particular, open-ended answer documents are more likely to have a user-friendly style. Therefore, it is preferable to maintain the style of the open-ended answer document while supplementing the closed-ended answer document with information organization.

[0121] The subsequent processing is the same as the question-and-answer processing in the first embodiment, so the description here is omitted. Furthermore, in step S107, if it is determined that the inherent information is not included in the question information (S107 is not), the process proceeds to step S306. In this case, since there is no closed-type answer document to be merged, the open-type answer document becomes the answer document sent to the injection molding machine 20.

[0122] According to this embodiment, since both open-ended and closed-ended answer documents can be generated, the accuracy of the answer documents can be further improved.

[0123] Furthermore, it is also conceivable to sign a confidentiality agreement with the operating company of the question-and-answer device 10A to prevent highly confidential information from being leaked to external parties. In this case, publicly known question information can be avoided; for example, even if it contains inherent information, the answer to the question information can be generated based on the open learning model 12. If the open learning model 12 is borrowed from an external company, and a confidentiality agreement is signed with that company, the answer to question information containing inherent information can also be generated based on the open learning model 12 in the same way. In the aforementioned cases, highly confidential question information regarding the injection molding machine 20, etc., does not need to be added to the open database 170, but can be used only in the answer to the question information.

[0124] In the above embodiments, the program (question and answer program related to industrial machinery) that implements various functions of the question and answer device 10 and the control device 200 is described as being pre-installed in the question and answer device 10 and the control device 200. However, it can also be provided as a file or other format capable of installing or executing the program, recorded on a recording medium that can be read by a computer. The storage medium here includes media that can be removably attached to the control device 7, such as magnetic tape, disk (hard disk drive, etc.), optical disc (CD-ROM, DVD, etc.), optical disk (MO, etc.), flash memory, etc., and media that can be further transmitted via a network, all media that can be read and executed by the computer that serves as the question and answer device 10 and the control device 200. It should be noted that as a medium that can be transmitted via a network, for example, a medium stored on a computer such as an external server connected to a network such as the Internet can be provided via the network. It should be noted that at least a portion of the functions of the question-and-answer device 10 and the control device 200 can also be constructed by an external server or the like, and data can be provided from the server to the question-and-answer device 10 and the control device 200 via network communication.

[0125] Several embodiments of the invention have been described, but these embodiments are provided as examples and are not intended to limit the scope of the invention. These new embodiments can be implemented in many other ways and can be omitted, substituted, or modified in various ways without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention and in the scope of the claims and their equivalents. Explanation of reference numerals in the attached figures

[0126] 1. Question and Answer System

[0127] 10 Question and Answer Device

[0128] 11 Closed-loop learning model

[0129] 12 Open Learning Models

[0130] 16 Closed-type databases

[0131] 17 Open Databases

[0132] 101 Information Acquisition Department (Question Information Acquisition Department, Recognition Information Acquisition Department, Evaluation Information Acquisition Department, Action Information Acquisition Department)

[0133] 102 Judgment Processing Department (Inherent Information Judgment Department, Correction Judgment Department, Verification Judgment Department)

[0134] 103 Answers Generation Department

[0135] 103A Open-ended Response Generation Department

[0136] 103B Closed-Type Response Generation Department

[0137] 104 Verification Processing Department (Verification Department)

[0138] 105 Learning and Updating Department (Updating Department)

[0139] 20 Injection Molding Machine

[0140] 200 control device

[0141] Question 207: Correction Department

[0142] 241 Monitor (Question and Answer Section)

[0143] 242 Input Device (Question Handling Department)

[0144] 243 Microphone (Question Handling Department)

[0145] 244 speakers (question and answer section).

Claims

1. A question-and-answer device, characterized in that, Answering questions related to industrial machinery, The question-and-answer device includes: The Question Information Acquisition Department acquires question information received by the Question Handling Department. The inherent information determination unit determines whether the acquired query information contains inherent information related to the industrial machinery; and The answer generation unit, upon determining that the question information contains the inherent information, generates an answer to the question information based on a closed-loop learning model that can be provided only to the user of the industrial machinery, wherein the closed-loop learning model has learned information related to the industrial machinery and information related to the inherent information.

2. The question-and-answer device according to claim 1, characterized in that, It also includes an identification information acquisition unit that acquires identification information related to the user of the industrial machinery. The answer generation unit generates an answer within the scope that can be disclosed to the user as indicated by the acquired identification information.

3. The question-and-answer device according to claim 2, characterized in that, It has multiple closed-type learning models, Each of the aforementioned closed-loop learning models is individually associated with the identification information of each user. The answer generation unit selects a closed-loop learning model corresponding to the acquired recognition information and generates an answer based on the selected closed-loop learning model.

4. The question-and-answer device according to claim 1, characterized in that, The answer generation unit generates answers to the acquired question information based on an open learning model and the closed learning model, wherein the open learning model learns about publicly known information that can be provided not only to users of the industrial machinery but also to other users.

5. The question-and-answer device according to claim 4, characterized in that, The question information acquisition unit, together with the question information, acquires publicly known question information containing publicly known information extracted from the question. The answer generation unit generates a first answer based on the open learning model for the known question information, and generates a second answer based on the closed learning model for the question information. The first and second answers are combined to generate an answer for the question information.

6. The question-and-answer device according to claim 1, characterized in that, Also includes: The correction determination unit determines whether to perform correction on the acquired query information; as well as The question correction unit corrects the question information when it determines that correction is needed for the acquired question information.

7. The question-and-answer device according to claim 1, characterized in that, Also includes: The verification and determination unit determines whether the answer generated by the answer generation unit can be verified. as well as The verification department verifies the answer if it determines that the answer can be verified.

8. The question-and-answer device according to claim 1, characterized in that, Also includes: The evaluation acquisition department acquires evaluations of the answer. as well as The updating unit updates the closed-loop learning model based on the obtained evaluation.

9. The question-and-answer device according to claim 1, characterized in that, It also includes a motion information acquisition unit, which acquires motion information including the motion conditions of the industrial machinery that has been questioned. The answer generation unit uses the closed-loop learning model and the action information to generate the answer.

10. A question-and-answer method, characterized in that, Answer questions related to industrial machinery. The question-and-answer device acquires question information representing the question. Determine whether the obtained query information contains inherent information related to the industrial machinery. If it is determined that the question information contains the inherent information, an answer to the question information is generated based on a closed-loop learning model that can be provided only to the user of the industrial machinery, wherein the closed-loop learning model has learned information related to the industrial machinery and information related to the inherent information.

11. A question-and-answer program, characterized in that, This section answers questions related to industrial machinery. The question-and-answer program enables the computer to function as the following components: The Question Information Acquisition Department acquires question information received by the Question Handling Department. The inherent information determination unit determines whether the acquired query information contains inherent information related to the industrial machinery. as well as The answer generation unit, upon determining that the question information contains the inherent information, generates an answer to the question information based on a closed-loop learning model that can be provided only to the user of the industrial machinery, wherein the closed-loop learning model has learned information related to the industrial machinery and information related to the inherent information.

12. A question-and-answer system, characterized in that, The question-and-answer system includes industrial machinery and a question-and-answer device for answering questions related to the industrial machinery, wherein the industrial machinery is at least one of an injection molding machine, an extruder, a stamping device, and a semiconductor manufacturing device. The industrial machinery includes: The Question Handling Department handles user questions; and The sending unit sends the question as question information to the question-and-answer device. The question-and-answer device includes: The question information acquisition unit acquires the question information sent by the sending unit; The inherent information determination unit determines whether the acquired query information contains inherent information related to the industrial machinery; and The answer generation unit, upon determining that the question information contains the inherent information, generates an answer to the question information based on a closed-loop learning model that can be provided only to the user of the industrial machinery, wherein the closed-loop learning model has learned information related to the industrial machinery and information related to the inherent information.

13. An industrial machine, characterized in that, This is an answer to questions related to industrial machinery. The industrial machinery is at least one of injection molding machines, extruders, stamping devices, and semiconductor manufacturing equipment, including: The Question Handling Department is responsible for handling user questions. The question information acquisition unit acquires questions received by the question handling unit as question information; and The question and answer department is responsible for answering the questions asked by users. The industrial machinery determines whether the question information contains inherent information related to the industrial machinery. If it is determined that the question information contains the inherent information, an answer to the question information is generated based on a closed-loop learning model that can only be provided to the user of the industrial machinery, and an answer to the question information is provided from the question-and-answer unit. The closed-loop learning model has learned information related to the industrial machinery and information related to the inherent information.

Images