Information processing device, information processing method, and information processing program

Abstract

Provide appropriate answers to questions from machine tool operators. [Solution] The information processing device includes: an acquisition unit that acquires question data from an operator indicating a question or instruction regarding a machine tool in natural language; an extraction unit that extracts keywords from the question data using a generation AI model; a selection unit that selects a related information source related to the question data from among a plurality of information sources that store information about a machine tool by comparing the extracted keywords extracted by the extraction unit with a keyword table that stores related keywords related to the question data in association with the plurality of information sources; an answer data creation unit that creates answer data indicating a response to the question or instruction based on the related information source and the extracted keywords; and an output unit that outputs the answer data, wherein at least one of the plurality of information sources includes a memory that stores real-time information about the machine tool.

Description

Technical Field

[0001] The present invention relates to a technique for responding to questions from an operator in a machine tool.

Background Art

[0002] At the work site of a machine tool, the machine tool may suddenly malfunction or the operation method may become unclear. In this case, it is very time-consuming for the operator to check the manual of the machine tool to find a countermeasure. On the other hand, in recent years, the performance of large language models such as ChatGPT has been rapidly improving. Therefore, technical development is underway to use large language models to create appropriate responses to questions from operators regarding machines.

[0003] For example, Patent Document 1 discloses an information processing apparatus that refers to a knowledge source in which various information related to a copying machine is aggregated, creates a response sentence or an additional question corresponding to the content of a given question, and transmits additional content request information for requesting an input of additional content necessary for narrowing down the response sentence to the question or the response to the question as a question response to an input device.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] Incidentally, with respect to machine tools, it's not enough to simply summarize and present the information that should be provided from manuals; depending on the question, there may be cases where real-time operating data of the machine tool should be provided as a response. Furthermore, information about machine tools is managed by various sources, such as cloud servers managed by the machine tool manufacturer and on-premise servers managed by the machine tool user. Thus, with respect to machine tools, there are various sources of information, so in order to create an appropriate response, it is necessary to select the most suitable source from among the various sources and create the response from that source.

[0006] In the technology described in Patent Document 1, the response is generated from a single information source. Therefore, if the single information source contains inappropriate information, it may not be possible to generate an appropriate response to the question. Furthermore, since the technology in Patent Document 1 does not include equipment as an information source, it is not possible to provide a response based on real-time operational data. In this respect, it is insufficient for creating an appropriate response.

[0007] The present invention aims to provide a technology for providing appropriate responses to questions from operators regarding machine tools. [Means for solving the problem]

[0008] An information processing device in one aspect of the present disclosure includes: an acquisition unit that acquires question data from an operator indicating a question or instruction regarding a machine tool in natural language; an extraction unit that extracts keywords from the question data using a generation AI model; a selection unit that selects a related information source related to the question data from a plurality of information sources that store information about the machine tool by comparing the extracted keywords extracted by the extraction unit with a keyword table that stores related keywords related to the question data in association with the plurality of information sources; an answer data creation unit that creates answer data indicating a response to the question or instruction based on the related information source and the extracted keywords; and an output unit that outputs the answer data, wherein at least one of the plurality of information sources includes a memory that stores real-time information about the machine tool. [Effects of the Invention]

[0009] This invention can provide appropriate responses to questions from operators regarding machine tools. [Brief explanation of the drawing]

[0010] [Figure 1] This figure shows an example of the overall configuration of the answer creation system in this embodiment. [Figure 2] This figure shows an example of data stored by multiple information sources. [Figure 3] This is a block diagram showing an example of the configuration of the answer generation device in this embodiment. [Figure 4] This figure shows an example of the process for extracting keywords. [Figure 5] This figure shows an example of the process for determining the answer type. [Figure 6] This figure shows an example of the data structure of a keyword table. [Figure 7] This figure shows an example of the data structure of a parameter table. [Figure 8] This figure shows an example of the process for creating response data. [Figure 9] This is a block diagram showing an example of the configuration of a terminal device. [Figure 10] This is a flowchart showing the processing of the answer generation system in this embodiment. [Figure 11] This is a diagram illustrating the process of obtaining datasets based on the response type. [Figure 12] This is a flowchart that continues from Figure 10. [Figure 13] This figure shows an example of a process for identifying information sources. [Figure 14] This is a flowchart that continues from Figure 12. [Figure 15] This is a flowchart that continues from Figure 14. [Modes for carrying out the invention]

[0011] (Embodiment) Note that each of the embodiments described below shows a specific example of the present disclosure. The numerical values, shapes, components, steps, order of steps, etc. shown in the following embodiments are merely examples and are not intended to limit the present disclosure. Also, among the components in the following embodiments, components not described in the independent claims indicating the most general concept are described as optional components. Also, in all embodiments, the respective contents can be combined with each other.

[0012] FIG. 1 is a diagram showing an example of the overall configuration of the answer creation system 1 in the present embodiment. The answer creation system 1 is a system that receives a question or instruction regarding the machine tool 60 from an operator, creates answer data indicating a response to the question or instruction, and presents the created answer data to the user.

[0013] The answer creation system 1 includes a terminal device 10, an answer creation device 20 (an example of an information processing device), a cloud server 30, an on-premises server 40, an external server 50, and one or more machine tools 60. The terminal device 10 to the machine tool 60 are communicably connected to each other via a network 70.

[0014] The network 70 is a wide area communication network including, for example, a mobile phone communication network and an Internet communication network.

[0015] The terminal device 10 consists of, for example, a mobile terminal such as a smartphone or tablet computer, or a stationary computer such as a desktop computer. Alternatively, the terminal device 10 may be implemented in the answer generation device 20 to constitute an information processing device. Alternatively, the terminal device 10 may be a console terminal that receives operations from an operator regarding the machine tool 60. The terminal device 10 acquires question data from the operator, indicating questions or instructions regarding the machine tool in natural language. The terminal device 10 transmits the acquired question data to the answer generation device 20, and the answer generation device 20 acquires answer data indicating a response to the question or instruction and presents it to the operator. The operator may be, for example, the manager of the machine tool 60 or the operator who actually operates the machine tool 60.

[0016] The response generation device 20 is composed of a computer, such as a cloud server. The response generation device 20 acquires question data from the terminal device 10, creates response data indicating a response to the question or instruction, and outputs it to the terminal device 10. The response generation device 20 may be implemented in the terminal device 10 or in the machine tool 60.

[0017] The cloud server 30 is a computer managed by, for example, the manufacturer of the machine tool 60. The cloud server 30 is a computer that stores data related to the machine tool 60 managed by the manufacturer (hereinafter referred to as cloud data).

[0018] The on-premise server 40 is a computer managed by the user of the machine tool 60. The on-premise server 40 stores data related to the machine tool 60 that is locally managed by the user of the machine tool 60 (hereinafter referred to as on-premise data).

[0019] The external server 50 is, for example, a computer that provides a website that publishes data (referred to as external data) about the machine tool 60 provided on the internet. The external server 50 may also include a computer that provides Generative Artificial Intelligence (AI) models on the internet.

[0020] The machine tool 60 is a machine tool installed at the material processing site. The machine tool 60 includes memory for storing machine tool data. The machine tool 60 is a machine used to process materials. The machine tool 60 may be a fully automatic or semi-automatic NC machine tool that processes materials, or a computer-controlled CNC machine tool.

[0021] Cloud servers 30, on-premise servers 40, external servers 50, and machine tools 60 are examples of multiple information sources.

[0022] Figure 2 shows an example of data stored by multiple information sources. The data stored by the information sources includes the machine tool data, on-premise data, cloud data, and external data mentioned above.

[0023] The machine tool data includes tool data, programs, and operational information. The tool data is data related to the type and shape of the tool attached to the machine tool 60. The program is the material processing program for the machine tool 60. The operational information is data including the current operational information of the machine tool 60 (an example of real-time information) and the history of alarms that have occurred in the past. The alarm history includes data such as the date and time the alarm occurred, the content of the alarm, and the actions taken in response to the alarm.

[0024] On-premise data includes processing performance data, machine status data, and tool status data. Processing performance data shows what materials and shapes the machine tool 60 has processed in the past. Machine status data shows what types of machine tools 60 are installed and where they are located throughout the factory. Tool status data shows what types of tools are installed on each machine tool 60. Tool status data also includes data showing what tools are stored in tool storage areas such as the factory's warehouse or shelves.

[0025] Cloud data includes service history, manual data, parts order history, and inquiry data. Service history is data showing what maintenance services were provided to which machine tool 60. Maintenance services include, for example, installation work for machine tool 60 and repairs for machine tool 60. The service history includes the date and time the maintenance service was provided and the content of the maintenance service. Manual data is electronic data of the operation manual for machine tool 60. Parts order history is data showing what parts have been ordered for which machine tool 60. The parts order history includes the date and time the parts were ordered and the type of part. Inquiry history shows the history of inquiries made by users regarding machine tool 60. The inquiry history includes the date and time the inquiry was made and the content of the inquiry.

[0026] External data includes internet data, generative AI models, machining data from tool manufacturers, and cloud data from affiliated companies. Internet data refers to web data publicly available on the internet. Generative AI models refer to generative AI models publicly available on the internet.

[0027] Generative AI models are artificial intelligence systems configured to generate text and other data in response to given prompts by training a neural network with a large number of parameters (e.g., billions to hundreds of billions) on a large amount of data. Examples of generative AI models include OpenAI's chat-gpt and Google's Gemini.

[0028] Tool manufacturer machining data refers to tool-related data that tool manufacturers publish on the internet. Related company cloud data refers to data published on the internet by the user's related companies or partner companies.

[0029] Figure 3 is a block diagram showing an example of the configuration of the answer generation device 20 in this embodiment. The answer generation device 20 includes a communication unit 21, a processor 22, and a memory 23. The communication unit 21 is a communication interface that connects the answer generation device 20 to the network 70. The communication unit 21 receives question data from the terminal device 10 and transmits answer data to the terminal device 10.

[0030] The processor 22 consists of a central processing unit (CPU). The processor 22 includes an acquisition unit 220, an extraction unit 221, a type determination unit 222, a selection unit 223, a response data creation unit 224, and an output unit 225. The acquisition unit 220 to the output unit 225 are realized by the CPU executing an information processing program stored in memory 23. However, this is just an example, and the acquisition unit 220 to the output unit 225 may be configured as dedicated hardware circuits.

[0031] The acquisition unit 220 acquires question data from the operator, which consists of questions or instructions regarding the machine tool in natural language. Specifically, the acquisition unit 220 acquires the question data transmitted from the terminal device 10 via the communication unit 21.

[0032] The extraction unit 221 extracts keywords from the question data. Hereinafter, the extracted keywords will be referred to as extracted keywords. For example, the extraction unit 221 can extract the extracted keywords by inputting the question data into a generation AI model. The generation AI model may be the generation AI model provided by the external server 50, or if the answer creation device 20 has a generation AI model, the extracted keywords may be extracted by inputting the question data into this generation AI model.

[0033] Figure 4 shows an example of the process for extracting keywords. In the example in Figure 4, the acquisition unit 220 acquires question data 1401. This question data 1401 includes a message that queries for data necessary to delete old tool data in a certain machine tool 60. Before inputting the question data 1401 to the generation AI, the extraction unit 221 inputs a prompt 1402 to the generation AI model. The prompt 1402 includes an instruction field 1402a, a keyword field 1402b, and a fuzzy keyword field 1402c.

[0034] The instruction field 1402a contains natural language instructions to extract keywords from the question data 1401, including those with the same meaning, as shown in the keyword field 1402b. The keyword field 1402b contains keywords that the generating AI model should extract from the question data 1401, such as tool data, machine, operating status, repair, and processing method. The ambiguous keyword field 1402c contains extraction rules for keywords that become ambiguous when the generating AI model extracts keywords. In this example, the rules specify that a 6-digit number indicates a machine, and tool data or ToolData should be extracted as tool data.

[0035] The extraction unit 221 inputs the question data 1401 to the generating AI following the input of prompt 1402. This allows the extraction unit 221 to extract keywords contained in the question data 1401. The extraction unit 221 can access the generating AI model using the API (Application Programming Interface) function of the generating AI model. This allows the extraction unit 221 to access the generating AI model without using a browser. Alternatively, the extraction unit 221 may extract keywords using a rule-based method without using the generating AI model.

[0036] The type determination unit 222 determines, based on the question data, which of the multiple answer types the question data belongs to. Figure 5 shows an example of the answer type determination process. The answer type is data that indicates the format of the answer from the answer creation device 20. The answer types include "Create", "Read", "Update", and "Delete".

[0037] "Creation" is a response type in the response creation system 1 that creates new data, such as creating a new repair request or creating new tool data for machine tool 60.

[0038] "Readout" is a response type that retrieves information about the machine tool 60, such as repair methods, machining methods, and tool data for the machine tool 60.

[0039] "Update" is a response type that instructs the machine tool 60 to update its tool data and program.

[0040] "Delete" is a response type that instructs the deletion of the program and tool data for the machine tool 60.

[0041] As shown in Figure 5, the type determination unit 222 inputs the question data 1401 to the generation AI. Before inputting keywords, the type determination unit 222 inputs a prompt 1302 to the generation AI. The prompt 1302 includes a natural language instruction that the AI ​​should respond with a number indicating which of the following categories the content of the question or instruction in the question data 1401 belongs to, based on the input extracted keywords. Here, the numbers "1" are assigned to creation, "2" to reading, "3" to updating, and "4" to deleting. Therefore, the generation AI model outputs a numerical value for the category to which the question content belongs from the input question data 1401, and the type determination unit 222 obtains this numerical value from the generation AI. This allows the type determination unit 222 to obtain the answer type. Note that, similar to the extraction unit 221, the type determination unit 222 can access the generation AI model using the API (Application Programming Interface) function of the generation AI model.

[0042] The selection unit 223 compares the extracted keywords extracted by the extraction unit 221 with the keyword table 231 to select relevant information sources related to the question data from among multiple information sources that store information about the machine tool 60.

[0043] Figure 6 shows an example of the data structure of keyword table 231. Keyword table 231 stores relevant keywords related to question data, associating them with multiple information sources. More specifically, keyword table 231 stores relevant keywords categorized according to combinations of multiple response types, multiple information sources, and multiple sub-information sources.

[0044] Keyword table 231 includes multiple datasets 231a. Each dataset 231a has fields for "ID", "Answer Type", "Related Keyword K1" to "Related Keyword KX", "Probability of Occurrence" for each of "Related Keyword K1" to "Related Keyword KX", "Source", and "Sub-Source".

[0045] "ID" is the identifier for dataset 231a. The "Response Type" field stores the response type to which dataset 231a belongs. "Related Keyword K1" to "Related Keyword KX" are keywords related to the question data corresponding to each dataset 231a. The probability of occurrence indicates the probability that each related keyword K1 to KX appears in each dataset 231a.

[0046] The probability of occurrence is obtained by statistically analyzing the operator's past question data. For example, when analyzing a set of past question data where the answer type was "Read," the information source was "Machine Tools," and the sub-information source was "Operating Status," the keywords "Machine" and "Operating" were included as characteristic keywords. In this set of question data, the probability of occurrence of "Machine" was "91%" and the probability of occurrence of "Operating" was "91%." Therefore, in the first row of dataset 231a, "Machine" is stored as related keyword K1 and "Operating" as related keyword K2, and the probability of occurrence of each is stored as "91%." In the subsequent rows of dataset 231a, the related keywords and their probabilities of occurrence obtained by analyzing past question data are stored in the same way as in the first row of dataset 231a.

[0047] The "Information Source" field stores the information source that best matches the content of the question or instruction in the question data corresponding to dataset 231a.

[0048] The "Sub-information Source" field stores the sub-information source that best matches the content of the question or instruction in the question data corresponding to dataset 231a. A sub-information source is an information source that belongs to a lower hierarchy than the main information source. For example, the information source "machine tool" includes "operating status," "program," and "tool" as sub-information sources.

[0049] Referring to Figure 2, the selection unit 223 determines the relevant information source by matching the extracted keywords with the relevant keywords related to the response type determined by the type determination unit 222. More specifically, the selection unit 223 selects a relevant sub-information source from among multiple sub-information sources included in the relevant information source by matching the extracted keywords with the keyword table 231. More specifically, the selection unit 223 calculates the degree of agreement between the extracted keywords and each of the multiple datasets 231a, and selects the relevant information source by comparing the calculated degree of agreement with a threshold.

[0050] Here, if there is more than one dataset 231a with a matching score above a threshold, the selection unit 223 creates additional question data based on related keywords and performs an interaction to obtain additional response data from the operator by presenting the additional question data to the operator. The selection unit 223 recalculates the matching score based on the additional response data and related keywords and repeats the interaction until a related information source is identified.

[0051] The response data creation unit 224 creates response data that indicates a response to a question or instruction based on the relevant information source selected by the selection unit 223 and the extracted keywords extracted by the extraction unit 221.

[0052] The response data creation unit 224 obtains parameter attributes corresponding to relevant information sources by referring to a parameter table that stores the parameter attributes necessary to access the information sources. The response data creation unit 224 extracts parameters with the obtained parameter attributes from the question data and uses the extracted parameters to access the relevant information sources.

[0053] Figure 7 shows an example of the data structure of parameter table 232. Parameter table 232 stores multiple parameter datasets 232a. Each parameter dataset 232a has the fields "ID", "Information Source", "Sub-Information Source", "Parameter Attribute", and "Sub-Parameter Attribute". "ID" is the identifier of parameter dataset 232a. The "Information Source" field stores the information source being accessed. The "Sub-Information Source" field stores the sub-information source being accessed. The "Parameter Attribute" field stores the parameter attributes necessary to access the information source. "Sub-parameters" are parameters at a lower hierarchy than the main parameter. The "Sub-parameter" field stores the sub-parameter attributes necessary to access the sub-information source. Note that sub-parameter attributes are not mandatory. In the example in Figure 7, a machine tool is shown as the information source, but parameter table 232 also stores parameter datasets 232a for other information sources.

[0054] The response data creation unit 224 inputs a prompt to the generating AI model that includes a message to extract parameters from the question data that have parameter attributes identified from the parameter table 232. This allows the response data creation unit 224 to obtain parameters from the question data for accessing information sources.

[0055] For example, if the selection unit 223 selects "machine tools" as the relevant information source and "operational information" as the relevant sub-information source, the response data creation unit 224 obtains the serial number of the machine tool 60 as a parameter attribute. The response data creation unit 224 extracts the serial number of the machine tool 60 from the question data.

[0056] In this case, the response data creation unit 224 can simply have the generating AI model extract the serial number from the question data. Then, the response data creation unit 224 can create the response data by reading the operating status from the machine tool 60 indicated by the acquired serial number. If the question data does not contain a serial number, the response data creation unit 224 can obtain the serial number by creating additional question data to elicit the serial number from the operator and presenting it to the operator.

[0057] The response data creation unit 224 may, as appropriate, have the generating AI model create the response data. For example, when the response type is "readout" and the repair method or processing method of the machine tool 60 is used as a sub-information source, the response data creation unit 224 can have the generating AI model create the response data by providing the generating AI model with extracted keywords and access data for accessing the information source, and by inputting a prompt that includes an instruction to create a summary about the keywords from the accessed information source.

[0058] The output unit 225 outputs the response data created by the response data creation unit 224 to the terminal device 10. The output unit 225 also outputs additional question data to the terminal device 10.

[0059] Figure 8 shows an example of the process for creating response data. In this example, the same question data 1401 as in Figure 5 is used as the question data. The response data creation unit 224 inputs prompt 1502 to the generating AI model before inputting the question data 1401. In this example, prompt 1502 is shown for when the parameter "machine tool serial number" and the sub-parameter "program name" are obtained in the parameter dataset 232a of the second row of the parameter table 232.

[0060] Prompt 1502 includes an instruction field 1502a, a parameter field 1502b, and a sub-parameter field 1502c.

[0061] The instruction field 1502a contains instructions for extracting data from the question data 1401 that possess the characteristics described in the parameter field 1502b and the sub-parameter field 1502c.

[0062] The parameter field 1502b describes the characteristics of the parameter. Here, a message indicating that the data to be extracted is the serial number of the machine tool 60, and a message describing the characteristics of the serial number's data structure are included.

[0063] Subparameter field 1502c describes the characteristics of the subparameter. Here, subparameter field 1502c contains a message indicating that the data to be extracted is the pocket number of the machine tool 60, a message explaining the meaning of the pocket number, and a message indicating that a numerical value representing the pocket number will be output.

[0064] As a result, the generating AI model accesses the machine tool 60 with serial number "1111" according to the instructions in prompt 1502 from the question data 1401, reads all the tool data stored in the accessed machine tool 60, and passes the read tool data to the answer data creation unit 224.

[0065] Memory 23 consists of a non-volatile, rewritable storage device and stores the keyword table 231 and the parameter table 232.

[0066] Figure 9 is a block diagram showing an example of the configuration of the terminal device 10. The terminal device 10 includes a display 11, a processor 12, an operation unit 13, a communication unit 14, and a memory 15. The display 11 is composed of a display device such as a liquid crystal panel and displays the answer data. The display 11 displays the question data entered by the operator.

[0067] The processor 12 consists of a central processing unit (CPU) and performs overall control of the terminal device 10.

[0068] The operation unit 13 consists of input devices such as a keyboard, mouse, and touch panel. The operation unit 13 receives question data from the operator.

[0069] The communication unit 14 is a communication interface that connects the terminal device 10 to the network 70. The communication unit 14 transmits question data to the answer creation device 20 and receives answer data from the answer creation device 20.

[0070] Memory 15 stores data necessary for the terminal device 10 to process input question data and display answer data.

[0071] Figure 10 is a flowchart showing the processing of the answer creation system 1 in this embodiment.

[0072] (Step S101) The terminal device 10 receives question data. In this case, the terminal device 10 acquires the question data entered into the operation unit 13. Alternatively, the terminal device 10 may acquire the question data by voice. In this case, the terminal device 10 can acquire the text data of the question data by picking up the voice of the question data with a microphone and converting the picked-up voice into text data using speech recognition processing.

[0073] (Step S102) The acquisition unit 220 of the answer creation device 20 acquires question data from the terminal device 10 via the communication unit 21.

[0074] (Step S103) The type determination unit 222 inputs a request to extract the answer type to the generating AI model. Specifically, as shown in Figure 5, the type determination unit 222 inputs an extraction request to the generating AI model that includes the prompt 1302 and the question data 1401. As a result, the type determination unit 222 obtains the answer type of the question data 1401 from the generating AI model.

[0075] (Step S104) The selection unit 223 outputs an instruction to the keyword table 231 to acquire a dataset 231a corresponding to the response type obtained in step S103.

[0076] (Step S105) The selection unit 223 retrieves datasets 231a corresponding to the response type from the keyword table 231. Figure 11 is an explanatory diagram of the process of retrieving datasets 231a corresponding to the response type. In this example, the response type obtained in step S103 was "Read". Therefore, the selection unit 223 retrieves datasets 231a with a response type of "Read" from the keyword table 231. In this example, three datasets 231a enclosed in thick borders are extracted. The three extracted datasets 231a are collectively called dataset group 231b.

[0077] Figure 12 is a flowchart that continues from Figure 10.

[0078] (Step S201) The extraction unit 221 inputs a request for extraction of keywords into the generating AI model. Specifically, as shown in Figure 4, the extraction unit 221 inputs an extraction request including question data 1401 and prompt 1402 into the generating AI model. As a result, the extraction unit 221 extracts keywords from the question data.

[0079] (Step S202) The selection unit 223 calculates the degree of agreement for each dataset 231a by comparing the extracted keywords obtained in step S201 with the related keywords included in the dataset group 231b acquired in step S105. The selection unit 223 then determines whether or not a related information source has been identified by comparing the calculated degree of agreement with a threshold.

[0080] Figure 13 shows an example of the process for identifying relevant information sources. In this example, "machine" is extracted as the keyword. The threshold is 0.9. Among the dataset group 231b, the datasets 231a that contain "machine" are the dataset 231a in the first row and the dataset 231a in the second row. The probability of "machine" appearing in both the first and second row datasets 231a is 91%. That is, the degree of agreement between the first row dataset 231a and the second row dataset 231a is 0.91. Therefore, there are two datasets 231a that exceed the threshold. For this reason, the selection unit 223 determines that it could not identify relevant information sources because there is more than one dataset 231a that exceeds the threshold.

[0081] If a relevant information source is identified (YES in step S202), the process proceeds to step S301; if a relevant information source is not identified (NO in step S202), the process proceeds to step S203.

[0082] (Step S203) As shown in Figure 13, the selection unit 223 determines additional keywords from the related keywords of the dataset group 231b. First, the selection unit 223 extracts dataset group 231c from dataset group 231b that contains the extracted keyword "machine". Then, the selection unit 223 determines the related keyword with the highest probability of occurrence and that has not been extracted as an extracted keyword in dataset 231a, which has the lowest "ID", as an additional keyword.

[0083] In the example in Figure 13, the dataset with the lowest "ID" among the dataset group 231c is dataset 231a in the first row. In dataset 231a in the first row, the related keyword with the highest probability of occurrence among related keywords not included in the extracted keyword is "operation". Therefore, the selection unit 223 decides "operation" to be added as an additional keyword.

[0084] (Step S204) The selection unit 223 inputs a request for the creation of additional question data to the generating AI model and retrieves the additional question data from the generating AI model. Specifically, the selection unit 223 creates a prompt indicating a request for the creation of a question sentence related to the additional keyword "operation" as the request for the creation of additional question data. As a result, the generating AI model creates additional question data such as, for example, "Is this a question about the operation of the machine?"

[0085] (Step S205) The selection unit 223 outputs additional question data to the terminal device 10 via the communication unit 21.

[0086] (Step S206) The communication unit 14 of the terminal device 10 acquires additional question data.

[0087] (Step S207) The display 11 of the terminal device 10 displays additional question data.

[0088] (Step S208) The operation unit 13 of the terminal device 10 acquires input from the operator of additional response data, which is response data to additional question data, and the communication unit 14 of the terminal device 10 outputs the acquired additional response data to the response creation device 20. The additional response data includes, for example, a statement agreeing with the additional question data.

[0089] (Step S209) The selection unit 223 acquires additional response data via the communication unit 21.

[0090] (Step S210) The selection unit 223 determines whether or not it was able to identify relevant information sources. Specifically, the selection unit 223 calculates the degree of agreement between the additional keyword "operation" and the relevant keyword for each of the two datasets 231a included in the dataset group 231c.

[0091] In the upper example of Figure 13, the first row of dataset 231a contains the related keyword "operation," and the probability of "operation" appearing is "91%." Therefore, the selection unit 223 calculates the degree of matching for the first row of dataset 231a as 0.91. On the other hand, the second row of dataset 231a does not contain "operation" as a related keyword. Therefore, the selection unit 223 calculates the degree of matching for the second row of dataset 231a as 0. Here, the only dataset 231a that exceeds the threshold of "0.9" is the first row of dataset 231a. Therefore, as shown in the lower part of Figure 13, the selection unit 223 determines that it has been able to uniquely identify dataset 231a and thus identify the related information source.

[0092] In this case, as shown in the lower part of Figure 13, the first row of the dataset 231a contains "machine tool" as the information source and "operating status" as the sub-information source. Therefore, the selection unit 223 identifies the machine tool 60 as the related information source and "operating status" as the related sub-information source.

[0093] If relevant information sources can be identified (YES in step S210), the process proceeds to step S301. On the other hand, if relevant information sources cannot be identified (NO in step S210), the process returns to step S204, and the selection unit 223 determines additional keywords again and narrows down the dataset 231a.

[0094] Figure 14 is a flowchart that continues from Figure 12.

[0095] (Step S301) The response data creation unit 224 refers to the parameter table 232 and obtains parameter attributes using the identified related information source and related sub-information source. In the example above, "machine tools" was identified as the related information source and "operating status" was identified as the related sub-information source. The related information source "machine tools" and the related sub-information source "operating status" correspond to the parameter dataset 232a in the first row of the parameter table 232.

[0096] Therefore, the response data creation unit 224 extracts the parameter dataset 232a from the first row of the parameter table 232. This first row of parameter dataset 232a contains only "machine tool serial number" as a parameter attribute, and no sub-parameter attributes are listed. Therefore, the response data creation unit 224 determines that "machine tool serial number" is the parameter attribute.

[0097] Furthermore, the response data creation unit 224 should retrieve the sub-parameter attributes along with the parameter attributes if they are present in the parameter dataset 232a extracted from the parameter table 232.

[0098] (Step S302) The response data creation unit 224 checks whether the question data obtained in step S102 contains a parameter with the parameter attribute identified in step S301. Specifically, the response data creation unit 224 creates a prompt as a check request that causes the generating AI model to check whether the question data contains a parameter with the identified parameter attribute. If the question data contains the corresponding parameter, the generating AI model outputs that parameter to the response data creation unit 224. On the other hand, if the corresponding parameter is not included, the generating AI model outputs a response to the response data creation unit 224 indicating that the parameter is not included.

[0099] Furthermore, if the response data creation unit 224 has obtained sub-parameter attributes in step S301, it should create a prompt as a check request that causes the generating AI model to check whether the question data contains sub-parameters that have sub-parameter attributes.

[0100] (Step S303) If the check result in step S302 indicates that the question data contains parameters (YES in step S303), the process proceeds to step S401. If the check result indicates that the question data does not contain parameters (NO in step S303), the process proceeds to step S304. If sub-parameter attributes are also obtained in step S301, the response data creation unit 224 only needs to determine YES in step S303 if both the parameter and sub-parameter have been obtained in step S302.

[0101] (Step S304) The response data creation unit 224 inputs a request to the generating AI model to create a parameter acquisition statement for obtaining parameters, and retrieves parameter acquisition statement data from the generating AI model. More specifically, the response data creation unit 224 can create a prompt as a request to create a parameter acquisition statement, which is a prompt to create a statement that causes the operator to answer with parameters that have parameter attributes.

[0102] For example, if the parameter attribute is the serial number of machine tool 60, a statement like "Please tell me the serial number of the work tool" will be created as the parameter retrieval statement.

[0103] Furthermore, if the response data creation unit 224 has obtained sub-parameter attributes in step S301, it should include a prompt in the request to create the parameter acquisition statement that prompts the operator to provide a statement containing the sub-parameters that have sub-parameter attributes.

[0104] For example, if "program name" is obtained as a sub-parameter attribute, a statement like "Please tell me the program name" will be created as the parameter retrieval statement.

[0105] (Step S305) The response data creation unit 224 transmits the parameter acquisition statement data to the terminal device 10 via the communication unit 21.

[0106] (Step S306) The communication unit 14 of the terminal device 10 acquires parameter acquisition statement data, and the display 11 displays the parameter acquisition statement data. The terminal device 10 may also output the parameter acquisition statement data as audio.

[0107] (Step S307) The operation unit 13 of the terminal device 10 receives parameters input by the operator. The communication unit 14 of the terminal device 10 outputs the received parameters to the response creation device 20. If, in step S305, a statement prompting the operator to provide sub-parameters is created as parameter acquisition statement data, the terminal device 10 only needs to acquire the sub-parameters from the operator.

[0108] Figure 15 is a flowchart that continues from Figure 14.

[0109] (Step S401) The response data creation unit 224 determines whether the response type identified in step S103 is anything other than read. If the response type is anything other than read (YES in step S401), the process proceeds to step S402. On the other hand, if the response type is read (NO in step S401), the process proceeds to step S409.

[0110] (Step S402) The response data creation unit 224 creates confirmation message data to confirm with the operator whether they want to perform the action indicated by the response type, and outputs the created confirmation message data to the terminal device 10. For example, if the response type is delete, the related information source is machine tool 60, and the related sub-information source is tool data, then confirmation message data such as "Are you sure you want to delete the tool data?" is created.

[0111] (Step S403) The communication unit 14 of the terminal device 10 acquires the confirmation message data, and the display 11 of the terminal device 10 displays the confirmation message data. The terminal device 10 may also output the confirmation message data as audio.

[0112] (Step S404) The operation unit 13 of the terminal device 10 receives input from the operator indicating consent to the confirmation statement data.

[0113] (Step S405) The response data creation unit 224 obtains a consent instruction via the communication unit 21.

[0114] (Step S406) The response data creation unit 224 outputs an operation instruction to the information source via the communication unit 21 to perform an operation corresponding to the response type. For example, if the response type is deletion, the related information source is the machine tool 60, and the sub-information source is tool data, an operation instruction to delete the tool data is output to the machine tool 60. The related information source and related sub-information source to which the operation instruction is output are the related information source and sub-information source having the parameters and sub-parameters identified in step S303. For example, the response data creation unit 224 outputs an operation instruction that includes the serial number of the target machine tool 60 and information specifying the tool data.

[0115] The relevant information source that has received the operation instruction executes the operation instruction. The relevant information source that has executed the operation instruction outputs a completion notification to the response generation device 20.

[0116] (Step S407) The response data creation unit 224 receives a completion notification via the communication unit 21. The response data creation unit 224 outputs the completion notification to the terminal device 10 via the communication unit 21.

[0117] (Step S408) The communication unit 14 of the terminal device 10 receives the completion notification, and the display 11 displays the completion notification. The terminal device 10 may also output the completion notification by voice.

[0118] (Step S409) The response data creation unit 224 outputs a read instruction to the information source via the communication unit 21. For example, if the information source is a machine tool 60 and the sub-information source is operational information, the response data creation unit 224 creates a read instruction to read the operational information of the machine tool 60. The information source and sub-information source to which the read instruction is output are the related information source and related sub-information source identified in step S303. For example, the response data creation unit 224 creates a read instruction that includes the serial number of the target machine tool 60 and reads the operational information. The information source that has received the read instruction reads the requested information according to the read instruction and outputs the read information (hereinafter referred to as read information) to the response creation device 20. Here, the operational information to be read includes real-time information of the machine tool 60.

[0119] (Step S410) The response data creation unit 224 inputs a request to the generating AI model to create response data from the read information acquired in step S409. For example, if the acquired read information includes an alarm, the response data creation unit 224 inputs a request to the generating AI model to create response data based on the information that it is an alarm. In this case, the generating AI model creates response data that says, "The current state is an alarm," and outputs it to the response data creation unit 224.

[0120] For example, if the information source is a cloud server 30 and the sub-information source is manual data, the response data creation unit 224 inputs a request to the generating AI model to create response data that summarizes the manual data based on extracted keywords.

[0121] (Step S411) The response data creation unit 224 outputs the response data to the terminal device 10 via the communication unit 21.

[0122] (Step S412) The communication unit 14 of the terminal device 10 acquires the response data, and the display 11 displays the response data.

[0123] Thus, the answer generation device 20 in this embodiment extracts keywords from the question data, compares the extracted keywords with a keyword table, selects relevant information sources from among multiple information sources that are related to the question data, and creates answer data from the selected relevant information sources. Therefore, the answer generation device 20 can provide appropriate answers to questions from operators regarding the machine tool 60. Furthermore, the answer generation device 20 stores real-time information of the machine tool in at least one of the multiple information sources. Therefore, this configuration can generate appropriate answers to questions that request real-time information of the machine tool.

[0124] The present invention can be modified in the following ways.

[0125] (Variation 1) The information processing device may be an answer creation system 1, a terminal device 10 on which an answer creation device 20 is implemented, a console terminal on which an answer creation device 20 is implemented, an answer creation device 20, a machine tool 60 on which an answer creation device 20 is implemented, a cloud server, an edge server, or an edge computer.

[0126] (Modification 2) The selection unit 223 may calculate the degree of agreement using Bayesian estimation. See Figure 6. In this case, the keyword table 231 is assumed to have pre-stored the prior probability that each dataset 231a will be selected. Let P(H1), P(H2), and P(H3) be the prior probabilities for the first to third rows of dataset 231a. Assume that "machine" is extracted as the keyword. In this case, the selection unit 223 calculates the degree of agreement for the first row of dataset 231a as 0.91*P(H0), the degree of agreement for the second row of dataset 231a as 0.91*P(H2), and the degree of agreement for the third row of dataset 231a as 0.91*P(H3).

[0127] For example, if the extracted keywords include "machine" and "operation," the selection unit 223 should calculate the degree of matching for the first row of dataset 231a as 0.91*0.91*P(H1), the degree of matching for the second row of dataset 231a as 0.91*0.00*P(H2), and the degree of matching for the third row of dataset 231a as 0.91*0.00*P(H3).

[0128] (Variation 3) The selection unit 223 may calculate the degree of matching such that the value increases as the number of elements containing the extracted keyword increases. For example, if the first row of dataset 231a contains 2 extracted keywords, the second row of dataset 231a contains 0 extracted keywords, and the third row of dataset 231a contains 0 extracted keywords, the degrees of matching can be calculated as 2, 0, and 0, respectively. Then, if the threshold is 1, the first row of dataset 231a has a degree of matching equal to or greater than the threshold, so the first row of dataset is uniquely identified.

[0129] (Modification 4) Keyword table 231 may be updated as needed. For example, a survey may be conducted to inquire with operators about the accuracy of their response data, and keyword table 231 may be updated based on the survey results.

[0130] (Variation 5) In this embodiment, the relevant information source was uniquely identified, but multiple information sources may be identified as relevant information sources. Similarly, while the relevant sub-information source was uniquely identified, multiple sub-information sources may be identified as relevant sub-information sources.

[0131] The technical features of this invention can be summarized as follows:

[0132] (Technology 1) An information processing device in one aspect of the present invention includes: an acquisition unit that acquires question data from an operator indicating a question or instruction regarding a machine tool in natural language; an extraction unit that extracts keywords from the question data using a generation AI model; a selection unit that selects a related information source related to the question data from a plurality of information sources that store information about the machine tool by comparing the extracted keywords extracted by the extraction unit with a keyword table that stores related keywords related to the question data in association with the plurality of information sources; an answer data creation unit that creates answer data indicating a response to the question or instruction based on the related information source and the extracted keywords; and an output unit that outputs the answer data, wherein at least one of the plurality of information sources includes a memory that stores real-time information about the machine tool.

[0133] This configuration extracts keywords from question data, matches the extracted keywords with a keyword table, selects relevant information sources from multiple sources to match the question data, and creates answer data from the selected relevant information sources. Therefore, this configuration can provide appropriate responses to questions from operators regarding machine tools. Furthermore, this configuration stores real-time information about the machine tool from at least one of the multiple information sources. Therefore, this configuration can create appropriate responses to questions or instructions requesting real-time information about the machine tool. In addition, because this configuration can create answer data from the appropriate information source, it eliminates the need to access information sources repeatedly to obtain appropriate answer data, reducing the number of processing steps and improving processing capacity.

[0134] (Technology 2) The information processing device described in Technology 1 may use a generative AI model to create the aforementioned response data.

[0135] This configuration allows for easy creation of response data using an AI model.

[0136] (Technology 3) The information processing device according to Technology 1 or 2 further comprises a type determination unit that determines, based on the question data, which of a plurality of answer types the question data belongs to, the keyword table stores the related keywords classified according to the combination of the plurality of answer types and the plurality of information sources, and the selection unit determines the related information source by matching the related keywords related to the answer type determined by the type determination unit with the extracted keywords.

[0137] This configuration narrows down related keywords according to the answer type, and then matches the narrowed-down related keywords with the extracted keywords to select relevant information sources, thus enabling the selection of relevant information sources that are appropriate for the answer type.

[0138] (Technology 4) In the information processing device described in any one of technologies 1 to 3, the plurality of information sources include sub-information sources, the keyword table further stores the related keywords in association with the sub-information sources, the selection unit selects a related sub-information source from among the plurality of sub-information sources included in the related information source by comparing the extracted keywords with the keyword table, and the response data creation unit creates the response data based on the related sub-information source and the extracted keywords.

[0139] This configuration narrows down the relevant sub-information sources from the relevant information sources to those that are appropriate for the content of the question or instruction, and then creates response data from the narrowed-down relevant sub-information sources, thereby enabling the creation of more appropriate response data.

[0140] (Technology 5) In the information processing device described in any one of technologies 1 to 4, the keyword table includes multiple datasets containing multiple related keywords, and the selection unit may calculate the degree of agreement between the extracted keyword and each of the multiple datasets, and select the related information source by comparing the degree of agreement with a threshold.

[0141] This configuration allows for the precise selection of relevant information sources using the degree of similarity.

[0142] (Technology 6) In the information processing device described in Technical 5, if there is not one dataset with a degree of agreement equal to or greater than the threshold, the selection unit may create additional question data based on the related keywords, perform an interaction to obtain additional response data from the operator by presenting the additional question data to the operator, recalculate the degree of agreement based on the additional response data and the related keywords, and repeat the interaction until the related information source is identified.

[0143] This configuration allows for the selection of a more appropriate source of information in response to a question or instruction, as the interaction is repeated until a relevant source with a degree of relevance exceeding a threshold is identified.

[0144] (Technology 7) In the information processing device described in one of the technologies 1 to 6, the plurality of information sources may further include the machine tool and a server of the manufacturer of the machine tool or a server of the user of the machine tool.

[0145] This configuration can access the server of a machine tool, the machine tool manufacturer, or the machine tool user to generate appropriate answers to questions or instructions.

[0146] (Technology 8) In the information processing device described in one of the technologies 1 to 7, the response data creation unit may obtain parameter attributes corresponding to the related information source by referring to a parameter table that stores parameter attributes necessary to access the information source for each of the multiple information sources, extract parameters having the obtained parameter attributes from the question data, and use the extracted parameters to access the related information source.

[0147] This configuration allows access to relevant information sources by obtaining the necessary access data through questions or instructions, and by accessing the obtained access data.

[0148] This disclosure can also be implemented as an information processing program that causes a computer to execute each of the characteristic configurations included in such an information processing device, or as an information processing system that operates using this information processing program. Furthermore, it goes without saying that such an information processing program can be distributed via computer-readable non-temporary recording media such as CD-ROMs or via communication networks such as the Internet.

[0149] (Technology 9) An information processing method in another aspect of the present invention includes a computer acquiring question data from an operator indicating a question or instruction regarding a machine tool in natural language; extracting keywords from the question data using a generative AI model; selecting a related information source related to the question data from a plurality of information sources that store information about the machine tool by comparing the extracted keywords with a keyword table that stores related keywords related to the question data in association with the plurality of information sources; creating answer data indicating a response to the question or instruction based on the related information source and the extracted keywords; and outputting the answer data, wherein at least one of the plurality of information sources includes a memory that stores real-time information about the machine tool.

[0150] (Technology 10) An information processing program in yet another aspect of the present invention causes a computer to perform the following actions: acquire question data from an operator indicating a question or instruction regarding a machine tool in natural language; extract keywords from the question data using a generative AI model; select a related information source related to the question data from among a plurality of information sources that store information about the machine tool by comparing the extracted keywords with a keyword table that stores related keywords related to the question data in association with the plurality of information sources; create answer data indicating an answer to the question or instruction based on the related information source and the extracted keywords; and output the answer data, wherein at least one of the plurality of information sources includes a memory that stores real-time information about the machine tool. [Explanation of symbols]

[0151] 1: Answer generation system 10: Terminal device 11: Display 12: Processor 13:Operation section 14: Communications Department 15: Memory 20: Answer generation device 21: Communications Department 22: Processor 23: Memory 30: Cloud Server 40: On-premises servers 50: External Server 60: Machine tools 70: Network 220: Acquisition Department 221:Extraction part 222: Type determination unit 223: Selection section 224: Response Data Creation Department 225: Output section 231: Keyword Table 232: Parameter Table 232a: Parameter dataset

Claims

1. An acquisition unit that acquires question data from an operator indicating questions or instructions regarding machine tools in natural language, An extraction unit that extracts keywords from the aforementioned question data using a generation AI model, A selection unit selects a related information source related to the question data from among multiple information sources that store information about the machine tool by comparing the extracted keywords extracted by the extraction unit with a keyword table that stores related keywords related to the question data in association with the multiple information sources. A response data creation unit that creates response data indicating a response to the question or instruction based on the aforementioned related information source and the extracted keywords, The system includes an output unit that outputs the aforementioned response data, At least one of the aforementioned multiple information sources includes a memory for storing real-time information of the machine tool. Information processing device.

2. A generative AI model is used to create the aforementioned response data. The information processing apparatus according to claim 1.

3. The system further includes a type determination unit that determines, based on the question data, which of a plurality of answer types the question data corresponds to. The keyword table stores the relevant keywords classified according to the combination of the multiple response types and multiple information sources. The selection unit determines the related information source by matching the related keywords related to the response type determined by the type determination unit with the extracted keywords. The information processing apparatus according to claim 1 or 2.

4. The aforementioned multiple information sources include sub-information sources, The keyword table further stores the related keywords in association with the sub-information sources, The selection unit selects a relevant sub-information source from among the multiple sub-information sources included in the relevant information source by comparing the extracted keyword with the keyword table. The response data creation unit creates the response data based on the related sub-information sources and the extracted keywords. The information processing apparatus according to claim 1 or 2.

5. The keyword table includes multiple datasets containing multiple related keywords, The selection unit calculates the degree of agreement between the extracted keyword and each of the multiple datasets, and selects the relevant information source by comparing the degree of agreement with a threshold. The information processing apparatus according to claim 1 or 2.

6. The aforementioned selection unit is If there is not one dataset with a degree of agreement equal to or greater than the threshold, an interaction is performed in which additional question data is created based on the related keywords, and the additional question data is presented to the operator to obtain additional response data from the operator. The degree of match is recalculated based on the additional response data and the related keywords, and the interaction is repeated until the related information source is identified. The information processing apparatus according to claim 5.

7. The aforementioned multiple information sources further include the machine tool and the server of the manufacturer of the machine tool or the server of the user of the machine tool. The information processing apparatus according to claim 1 or 2.

8. The aforementioned response data creation unit, By referring to a parameter table that stores the parameter attributes necessary to access the information source for each of the aforementioned multiple information sources, the parameter attributes corresponding to the relevant information source are obtained. The parameters having the acquired parameter attributes are extracted from the question data. Using the extracted parameters, access the relevant information sources. The information processing apparatus according to claim 1 or 2.

9. Computers To obtain question data from operators indicating questions or instructions regarding machine tools in natural language, Extracting keywords from the aforementioned question data using a generative AI model, By comparing the extracted keywords with a keyword table that stores related keywords associated with the question data in relation to the multiple information sources, a related information source related to the question data is selected from among the multiple information sources that store information about the machine tool. To create response data showing a response to the question or instruction based on the aforementioned related information source and the extracted keywords, This includes outputting the aforementioned response data, At least one of the aforementioned multiple information sources includes a memory for storing real-time information of the machine tool. Information processing methods.

10. On the computer, To obtain question data from operators indicating questions or instructions regarding machine tools in natural language, Extracting keywords from the aforementioned question data using a generative AI model, By comparing the extracted keywords with a keyword table that stores related keywords associated with the question data in relation to the multiple information sources, a related information source related to the question data is selected from among the multiple information sources that store information about the machine tool. To create response data that shows the answer to the question or instruction based on the aforementioned related information source and the extracted keywords, To output the aforementioned response data and to execute the following: At least one of the aforementioned multiple information sources includes a memory for storing real-time information of the machine tool. Information processing program.

Images