Program, information processing device, and information processing method

Abstract

To improve the accuracy of task planning using language models. [Solution] This program causes a computer to execute a request acquisition step, a confirmation item generation step, an identification step, an information acquisition step, and a procedure generation step. The request acquisition step acquires a request to generate an execution procedure written in natural language. The confirmation item generation step uses the first confirmation item and the generation request to confirm reference information when generating the execution procedure to generate a second confirmation item other than the first confirmation item. The identification step identifies missing information from the reference information confirmed by the first and second confirmation items that cannot be obtained from the generation request. The information acquisition step acquires the missing information from information sources other than the generation request. The procedure generation step selects one or more elements from the elements of the execution procedure and generates the execution procedure using the selected elements.

Description

Technical Field

[0001] Embodiments of the present invention relate to a program, an information processing apparatus, and an information processing method.

Background Art

[0002] With the advancement of artificial intelligence (AI) technology, the application of language models such as large language models (LLMs) in the field of natural language processing (NLP) has been rapidly expanding. A language model is a model trained using an enormous dataset and has the ability to understand and generate human language. In particular, in task planning for generating an execution procedure of a process for realizing a request by combining pre-prepared processes (tasks) from a user's request, by utilizing a language model, it becomes possible to automate and streamline the generation of complex execution procedures.

[0003] Conventional task planning techniques that do not use a language model mainly rely on a rule-based approach and may lack flexibility and adaptability. In contrast, language models are attracting attention as a new means for realizing flexible task planning using natural language.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] An object of the present invention is to provide a program, an information processing apparatus, and an information processing method that can improve the accuracy of task planning using a language model.

Means for Solving the Problems

[0006] The program of this embodiment is a program that causes a computer to execute a request acquisition step, a confirmation item generation step, an identification step, an information acquisition step, and a procedure generation step. The request acquisition step acquires a request to generate an execution procedure described in natural language. The confirmation item generation step uses the first confirmation item and the generation request to confirm reference information when generating the execution procedure to generate a second confirmation item other than the first confirmation item. The identification step identifies missing information from the reference information confirmed by the first and second confirmation items that cannot be obtained from the generation request. The information acquisition step acquires the missing information from information sources other than the generation request. The procedure generation step selects one or more elements from the elements of the execution procedure and generates the execution procedure using the selected elements. [Brief explanation of the drawing]

[0007] [Figure 1] Block diagram of an information processing device according to an embodiment. [Figure 2] A diagram showing an example of elemental information stored in the memory unit. [Figure 3] A flowchart of the procedure generation process in the embodiment. [Figure 4] A flowchart of the process for generating confirmation items in the embodiment. [Figure 5] A flowchart of the process for generating confirmation items in the embodiment. [Figure 6] A flowchart of the process for generating confirmation items in the embodiment. [Figure 7] A diagram illustrating an example of a user interface. [Figure 8] A diagram illustrating an example of a user interface. [Figure 9] A diagram showing an example of a confirmation screen for verifying reference information. [Figure 10] Hardware configuration diagram of the information processing device according to the embodiment. [Modes for carrying out the invention]

[0008] A preferred embodiment of the information processing device according to this invention will be described in detail below with reference to the attached drawings.

[0009] In task planning, the generation of execution procedures may not be accurate if the user's requests, which are described in natural language, lack sufficient information necessary for generating those procedures. For example, if the user's input is ambiguous, it may not be possible to generate accurate execution procedures.

[0010] In configurations where a language model infers missing information to generate execution procedures, incorrect inferences can lead to situations where the desired execution procedure cannot be obtained. To address this problem, a technique called slot filing has been proposed. Slot filing is a technique that pre-sets specific slots (confirmation items), fills in the slots based on user requests, and collects the information necessary to realize the execution procedure in advance by making inquiries about any information that cannot be filled.

[0011] With techniques like slot filing, it is difficult to comprehensively set up verification items when the required execution procedures are diverse. For example, when generating new software (an example of an execution procedure) by combining components of previously created image recognition software, it is difficult to prepare all query slots in advance because user requirements are diverse.

[0012] The information processing device of this embodiment is a device that generates execution procedures for one or more processes to fulfill a user's request, and collects in advance any information that is missing for generating the execution procedures. The information processing device of this embodiment can generate a complete set of confirmation items (slots) even when the user's requests are diverse. As a result, there is no longer any missing information for generating the execution procedures, and the accuracy of the task plan is improved.

[0013] FIG. 1 is a block diagram showing an example of the configuration of the information processing apparatus 100 according to the embodiment. As shown in FIG. 1, the information processing apparatus 100 includes a request acquisition unit 101, a search unit 102, a confirmation item generation unit 103, a specification unit 104, an information acquisition unit 105, a procedure generation unit 106, an output control unit 111, a storage unit 121, and a display unit 122.

[0014] The storage unit 121 stores various kinds of information used in the information processing apparatus 100. For example, the storage unit 121 stores in advance one or more pieces of element information (operation information) representing elements of the execution procedure. The element information may be information in any format, but for example, it is information as follows. · API (Application Programming Interface) information such as software components or hardware components: The API information includes, for example, descriptions (explanation texts) corresponding to arguments (inputs), functions, and return values (outputs). · Procedure information representing each procedure in manuals such as operation manuals and troubleshooting manuals: The procedure information includes, for example, descriptions of specific procedures.

[0015] FIG. 2 is a diagram showing an example of the element information stored in the storage unit 121. The element information includes an element name and an explanation. The element name is information for identifying the element. The explanation represents an explanation for the corresponding element.

[0016] Elements whose element names include "API" are examples of the above API information. For example, "API-01" represents an API that inputs an image and identifies an object in the image. "API-02" represents an API that inputs point cloud data and identifies an object in the point cloud data. "API-03" represents an API that inputs an image and an object identified from the image and displays an image of the object. "API-04" represents an API that inputs the number of detections for each time and displays a graph representing the change in the number of detections.

[0017] Elements whose names contain "procedure" are examples of the above procedure information. For example, for the element "Procedure-01", the event that occurred and the corresponding method when the event occurred are associated as an explanation.

[0018] Note that the storage unit 121 can be composed of any commonly used storage medium such as a flash memory, a memory card, a RAM (Random Access Memory), a HDD (Hard Disk Drive), and an optical disk.

[0019] Return to the description of FIG. 1. The display unit 122 is a display device such as a liquid crystal display for displaying various information.

[0020] The request acquisition unit 101 acquires a generation request described in natural language. The generation request is a request for generating the execution procedures of one or more processes. The generation request may be in any format, but for example, it is input in text data or voice data.

[0021] The method for the request acquisition unit 101 to acquire information may be any method, but for example, methods such as acquiring information input via an interactive user interface, receiving from an external device via a network, and reading information from a storage medium can be applied.

[0022] The generation request assumes that a request that can be realized by combining one or more pieces of element information stored in the storage unit 121 is input. When the element information is API information, the generation request is, for example, a request for a new system that can be realized by combining software components or hardware components. When the element information is procedure information, the generation request is a request for a response procedure at the time of an incident and a request for creating a new procedure manual, etc.

[0023] The search unit 102 searches for one or more elements from the element information stored in the memory unit 121 to fulfill the generation request. The search unit 102 can use any method to search for elements, but for example, one of the following methods can be used. (M1) The language model is asked to search for candidate elements necessary to realize the execution procedure requested by the generation request. (M2) The memory unit 121 searches for element information that has a high similarity to the keyword included in the generation request from among the element information stored therein. (M3) The memory unit 121 searches for element information among the element information stored therein that has a high similarity to the feature quantities of the element information with respect to the feature quantities of the generation request (such as feature vectors). The feature quantities can be calculated, for example, using a pre-trained model (such as a neural network model).

[0024] The search unit 102 is not required. For example, if all element information stored in the memory unit 121 is to be considered as candidates, the search unit 102 is not necessary.

[0025] The confirmation item generation unit 103 generates one or more confirmation items CB (second confirmation items) other than confirmation items CA, using one or more predetermined confirmation items CA (first confirmation items) and a generation request. Confirmation items represent items used to verify reference information that is referenced when generating the execution procedure.

[0026] Confirmation Items CA include, for example, items that are expected to require common confirmation across various execution procedures, and items that are the minimum necessary for generating execution procedures. When building (generating) a new system by combining API information, for example, the inputs, functions, and outputs of the system to be generated would be defined as Confirmation Items CA. When generating procedure information for a manual, the model numbers of the products covered by the manual would be defined as Confirmation Items CA.

[0027] The confirmation item generation unit 103 may generate confirmation item CB using elements (element information) in addition to confirmation item CA and generation request. The element information may be element information retrieved by the search unit 102. Details of the method for generating confirmation items by the confirmation item generation unit 103 will be described later.

[0028] The identification unit 104 identifies one or more missing pieces of information that represent reference information that cannot be obtained from the generation request, among the reference information confirmed by confirmation item CA and confirmation item CB. Techniques such as filling slots in slot filing can be used to identify missing pieces of information. For example, the identification unit 104 provides the language model with a generation request and confirmation items (slots), and if the generation request contains reference information corresponding to a confirmation item, it instructs the language model to fill the confirmation item with that reference information. The identification unit 104 can identify reference information confirmed by unfilled confirmation items as missing pieces of information.

[0029] The information acquisition unit 105 acquires the identified missing information from sources other than the generation request. For example, the information acquisition unit 105 acquires the missing information by at least one of the following acquisition methods. (A1) A method for obtaining input missing information in response to a request for input of identified missing information. (A2) Method for obtaining missing information from an external device of the information processing device 100

[0030] The procedure generation unit 106 generates the execution procedure requested by the generation request. For example, the procedure generation unit 106 uses the generation request, the missing information obtained by the information acquisition unit 105, and the element information to generate an execution procedure for carrying out the generation request by combining one or more element information.

[0031] The procedure generation unit 106 first selects one or more elements from the elements represented by the element information using the generation request, the reference information obtained from the generation request, and the acquired missing information. The procedure generation unit 106 generates an execution procedure using the one or more selected elements. The procedure generation unit 106 generates an execution procedure, for example, by requesting the language model to generate an execution procedure using the selected elements.

[0032] The procedure generation unit 106 may classify (organize) the reference information (including the acquired missing information) by type and generate an execution procedure using the classified reference information and missing information. The type indicates that the reference information and missing information are at least one of the following: input to the execution procedure, output to the execution procedure, or function to the execution procedure. Details of the classification method for reference information will be described later.

[0033] The output control unit 111 controls the output of various types of information used by the information processing device 100. For example, the output control unit 111 outputs reference information and missing information classified by type. The method of outputting the information can be any method, but for example, it can be displayed on a display device such as the display unit 122, or transmitted to an external device via a network.

[0034] At least a portion of each of the above parts (request acquisition unit 101, search unit 102, confirmation item generation unit 103, identification unit 104, information acquisition unit 105, procedure generation unit 106, and output control unit 111) may be implemented by one or more processing units. Each of the above parts may be implemented by, for example, one or more processors. For example, each of the above parts may be implemented by having a processor such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit) execute a program, i.e., by software. Each of the above parts may be implemented by a processor such as a dedicated IC (Integrated Circuit), i.e., by hardware. Each of the above parts may be implemented by using both software and hardware. When multiple processors are used, each processor may implement one of the above parts, or two or more of the above parts.

[0035] The information processing device 100 may be composed of one physical device or multiple physical devices. For example, the information processing device 100 may be built on a cloud environment. Furthermore, each part of the information processing device 100 may be distributed and provided on multiple devices.

[0036] Next, the procedure generation process by the information processing device 100 of the embodiment will be described. Figure 3 is a flowchart showing an example of the procedure generation process in the embodiment.

[0037] The request acquisition unit 101 acquires a generation request, for example, input by the user (step S101). The confirmation item generation unit 103 uses the generation request to generate confirmation items CB other than the default confirmation items CA (step S102). As described above, the confirmation item generation unit 103 may also generate confirmation items CB using element information retrieved by the search unit 102, for example.

[0038] The identification unit 104 identifies missing information that cannot be obtained from the generation request among the reference information to be confirmed in the confirmation items (confirmation items CA, CB) (step S103). The information acquisition unit 105 acquires the identified missing information (step S104).

[0039] The procedure generation unit 106 generates an execution procedure to fulfill the generation request using the acquired information (step S105). For example, the procedure generation unit 106 generates an execution procedure by combining one or more elemental information using the generation request, the missing information acquired in step S104, and the elemental information.

[0040] Next, the details of the method for generating confirmation items CB by the confirmation item generation unit 103 (step S102) will be described. The confirmation item generation unit 103 generates, for example, some or all of the following three confirmation items CB1 to CB3. (CB1) Confirmation items for input when the execution procedure (elements of the execution procedure) is incomplete. (CB2) Confirmation items necessary to identify the elements (element information) required to generate the execution procedure (CB3) Confirmation items for concretizing the generation request

[0041] Confirmation item CB1 corresponds to a confirmation item that verifies the input of an element using reference information. For example, the confirmation item generation unit 103 generates confirmation item CB to verify one or more inputs of an element that are not verified by confirmation item CA.

[0042] Confirmation item CB2 corresponds to confirmation items that verify the elements necessary for generating the execution procedure as reference information. For example, the confirmation item generation unit 103 generates confirmation item CB to verify reference information that represents one or more elements used to generate the execution procedure.

[0043] Confirmation item CB3 corresponds to a confirmation item that uses reference information to confirm the phrases that specify the phrases that need to be specified among the phrases included in the generation request. For example, the confirmation item generation unit 103 generates a confirmation item CB to confirm reference information that specifies the phrases that specify the phrases that need to be specified among one or more phrases included in the generation request.

[0044] The following describes how to generate each of the three confirmation items CB1 to CB3. First, we will explain how to generate confirmation item CB1. Figure 4 is a flowchart showing an example of the confirmation item generation process in the embodiment. Figure 4 corresponds to an example of how to generate confirmation item CB1.

[0045] For example, the search unit 102 searches for candidate elements necessary to realize the generation request (step S201). Assuming that the elements are APIs for software components, the confirmation item generation unit 103 identifies the missing inputs (inputs not included in the generation request) from among the API inputs (arguments) (step S202). The confirmation item generation unit 103 generates the identified inputs as confirmation items CB (step S203).

[0046] For example, the following methods can be used to obtain API inputs that cannot be obtained from the generation request. • A method that utilizes function calling capabilities using language models, etc. This method involves having a language model generate execution steps and then analyzing the arguments of those steps to determine any missing inputs.

[0047] When using a language model, the confirmation item generation unit 103 may request the language model to perform the search using the above method (M1) along with the generation of confirmation items. In other words, the confirmation item generation unit 103 may request the language model to perform both the search for candidate elements and the processing necessary for generating confirmation items.

[0048] In the case of a generation request realized by combining multiple elements (APIs), the confirmation item generation unit 103 may, for example, generate confirmation item CBs by performing function calling for each of the multiple elements. The confirmation item generation unit 103 may also determine missing arguments by having the language model generate an execution procedure using multiple APIs.

[0049] Next, the method for generating the confirmation items CB2 will be described. Figure 5 is a flowchart showing another example of the confirmation item generation process in the embodiment. Figure 5 corresponds to an example of the method for generating the confirmation items CB2.

[0050] For example, the search unit 102 searches for candidate elements necessary to fulfill the generation request (step S301). Assuming that the elements are APIs for software components, the confirmation item generation unit 103 generates confirmation item CB to determine which of several elements (APIs) with similar functions to adopt. For example, the confirmation item generation unit 103 classifies several elements (APIs) with similar functions into one group (step S302). The confirmation item generation unit 103 generates confirmation item CB to indicate the differences between several elements included in the group (step S303). For example, the confirmation item generation unit 103 generates confirmation item CB using a language model to indicate questions to confirm the differences between several elements included in the group.

[0051] Similar functions can be determined in any way, but for example, a method that determines them based on the similarity of inputs, outputs, and functions, or a method that determines them by referring to information on similar functions stored in advance in the memory unit 121 or the like, can be applied.

[0052] Next, the method for generating the confirmation items CB3 will be described. Figure 6 is a flowchart showing another example of the confirmation item generation process in the embodiment. Figure 6 corresponds to an example of the method for generating the confirmation items CB3.

[0053] For example, the confirmation item generation unit 103 extracts ambiguous phrases (words, clauses, etc.) from among the phrases (words, phrases, etc.) included in the generation request (step S401). The ambiguity of a phrase can be determined by any method. For example, the confirmation item generation unit 103 may calculate the frequency of each phrase included in the generation request and the phrases included in the element descriptions stored in the memory unit 121, and determine the ambiguity based on the frequency. For example, the confirmation item generation unit 103 extracts a phrase included in the generation request as an ambiguous phrase if the frequency of the phrase included in the generation request is lower than that of the phrases included in the element descriptions. The confirmation item generation unit 103 may also use a language model to determine the ambiguity of the phrases included in the generation request.

[0054] The confirmation item generation unit 103 identifies ambiguous phrases as phrases that need to be specified and generates confirmation item CBs to confirm the phrases that specify those ambiguous phrases (step S402).

[0055] Next, we will explain in detail the methods for obtaining missing information (A1) and (A2). (A1) corresponds to a method that requests input of reference information to be confirmed in an interactive format based on the items to be confirmed, or a method that requests input of reference information by presenting a list of items to be confirmed. (A2) corresponds to a method that obtains missing information using an external device (system) such as a search engine.

[0056] Figure 7 shows an example of a user interface that prompts for reference information input in an interactive format. The input screen 700 shown in Figure 7 is an example of a screen that appears when a generation request is received stating, "I want to identify recyclable items on the conveyor belt in order to automate the sorting process."

[0057] For such a generation request, for example, "API-01" in Figure 2, which is an API that takes an image as input and identifies objects within that image, and "API-02" in Figure 2, which is an API that takes point cloud data as input and identifies objects within that point cloud data, are searched as candidate elements. Assume that the input, function, and output of the API are defined as verification items CA. The generation request in Figure 7 includes the function of identifying recyclable objects (object identification) and the output of the identification result (identified object), but does not include information that specifies whether the input is an image or point cloud data. For this reason, among the reference information verified by verification items CA, the reference information representing the input is identified as missing information.

[0058] The information acquisition unit 105 acquires the identified missing information from information sources other than the generation request, namely, from information entered into the interactive input screen 700. The question "What is the input?" in Figure 7 corresponds to the question used to acquire this missing information (input). The user enters information in response to this question, for example, "It's an image."

[0059] In order for "API-01" and "API-02" to function properly, it is necessary to further specify the type of object to be identified. Since the type of object is not included in the generation request, the confirmation item generation unit 103 generates the type of object as a confirmation item CB that is not included in the generation request. For example, the confirmation item generation unit 103 assumes that "recyclable object" included in the generation request is an ambiguous phrase and generates a confirmation item CB to confirm a phrase that clarifies that phrase. "What is a recyclable object?" in Figure 7 corresponds to a question that indicates the reference information (missing information) that this confirmation item CB clarifies. In response to this question, the user inputs reference information (type of object), for example, "bottles, cans, and plastic bottles."

[0060] Thus, in this embodiment, not only are predetermined verification items CA (input, function, output) generated, but verification items CB other than CA can also be generated and used to verify reference information for generating execution procedures. Therefore, it is possible to suppress the occurrence of insufficient information for generating execution procedures and improve the accuracy of task planning using language models.

[0061] Figure 8 shows an example of a user interface that presents a list of items to be checked and prompts the user to input reference information. The input screen 800 shown in Figure 8 is an example of a screen that represents missing information with "???" and prompts the user to input that missing information.

[0062] The information acquisition unit 105 may determine the acquisition order (priority, rank) for one or more identified missing pieces of information and acquire the missing pieces of information according to the determined acquisition order. For example, the information acquisition unit 105 may acquire the missing pieces of information according to the acquisition order using an external device or an interactive input screen.

[0063] The acquisition order can be determined, for example, by using a machine learning model (such as a neural network model) trained on training data that defines the order of missing information. However, the method for determining the acquisition order is not limited to this. For example, a method may be used in which missing information is input to a language model and the acquisition order is output.

[0064] The information acquisition unit 105 may be configured not to acquire any missing information that is in a later acquisition order than the missing information that could not be acquired if it could not be acquired.

[0065] For example, if the system attempts to acquire missing information interactively according to the acquisition order, and the user inputs something like "I don't know" or "Nothing in particular," and therefore the missing information cannot be obtained (the slot cannot be filled), the information acquisition unit 105 stops acquiring further information. This reduces inquiries about non-essential confirmation items.

[0066] Next, we will explain the classification method for reference information and the details of generating execution procedures using the classified reference information.

[0067] For example, the procedure generation unit 106 classifies the reference information (including the acquired missing information) by type. The types are, for example, the input of the execution procedure, the output of the execution procedure, and the function of the execution procedure. The types may also include the types of reference information acquired by the confirmation items CB. The procedure generation unit 106 may classify the reference information by type using a language model.

[0068] The procedure generation unit 106 uses the generation request, classified reference information (including missing information), and element information to generate an execution procedure for carrying out the generation request by combining one or more pieces of element information. For example, the procedure generation unit 106 inputs a request to the language model that includes the generation request, classified reference information (including missing information), and element information, and requests the generation of an execution procedure that uses the elements represented by the element information, thereby causing the language model to generate an execution procedure.

[0069] The procedure generation unit 106 can generate execution procedures for the language model by, for example, using functions such as function calling and ReAct (Reasoning and ACTing) of the language model.

[0070] As described above, the language model receives information that classifies (organizes) the reference information used when generating execution procedures. This clarifies the inputs, outputs, and functions of the execution procedures, thereby improving the accuracy of procedure generation.

[0071] The output control unit 111 may output information indicating the classified reference information. Figure 9 shows an example of a confirmation screen 900 for checking the classified reference information. On the confirmation screen 900, along with the generation request entered by the user, etc., reference information classified into four types: "input", "function", "type of object", and "output" is output. The user can use the confirmation screen 900 to check, for example, the information provided to the language model.

[0072] The procedure generation unit 106 checks the validity of the execution procedure generated by the language model, and if it is not valid, it may request the language model to regenerate the execution procedure. For example, the procedure generation unit 106 checks whether the inputs and outputs of the generated execution procedure match the requested inputs and outputs. If either or both of the inputs and outputs do not match, the procedure generation unit 106 feeds back information indicating the mismatch to the language model and causes the language model to regenerate the execution procedure. This makes it possible to generate an execution procedure that is closer to the generation request.

[0073] Thus, the information processing device of this embodiment can improve the accuracy of task planning using a language model.

[0074] Next, the hardware configuration of the information processing device of the embodiment will be described using Figure 10. Figure 10 is an explanatory diagram showing an example of the hardware configuration of the information processing device of the embodiment.

[0075] The information processing device of this embodiment includes a control device such as a CPU (Central Processing Unit) 51, a storage device such as a ROM (Read Only Memory) 52 and a RAM (Random Access Memory) 53, a communication interface 54 that connects to a network for communication, and a bus 61 that connects each part.

[0076] The program to be executed in the information processing device of this embodiment is provided pre-installed in a ROM 52 or the like.

[0077] The program executed by the information processing device of this embodiment may be configured to be provided as a computer program product by recording it in an installable or executable file format onto a computer-readable recording medium such as a CD-ROM (Compact Disk Read Only Memory), a flexible disk (FD), a CD-R (Compact Disk Recordable), or a DVD (Digital Versatile Disk).

[0078] Furthermore, the program executed by the information processing device of the embodiment may be stored on a computer connected to a network such as the Internet and provided by downloading it via the network. Alternatively, the program executed by the information processing device of the embodiment may be provided or distributed via a network such as the Internet.

[0079] The program executed in the information processing device of this embodiment can cause the computer to function as a component of the information processing device described above. This computer can read the program from a computer-readable storage medium onto the main memory and execute it using the CPU 51.

[0080] While several embodiments of the present invention have been described, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments can be carried out in a variety of other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents. [Explanation of Symbols]

[0081] 100 Information Processing Devices 101 Request acquisition part 102 Search Section 103 Confirmation item generation section 104 Specific part 105 Information acquisition department 106 Procedure generation unit 111 Output Control Unit 121 Storage section 122 Display section

Claims

1. On the computer, A request acquisition step that acquires a generation request written in natural language that requests the generation of one or more execution steps, A confirmation item generation step that generates one or more second confirmation items other than the first confirmation items, using one or more predetermined first confirmation items, which are confirmation items for confirming reference information referenced when generating the execution procedure, and the generation request, A selection step to identify one or more missing pieces of information that represent the reference information that cannot be obtained from the generation request, among the reference information confirmed by the first and second confirmation items, An information acquisition step of obtaining the identified missing information from a source other than the generation request, A procedure generation step that uses the generation request, the reference information obtained from the generation request, and the acquired missing information to select one or more elements from one or more predetermined elements of an execution procedure, and generates the execution procedure using the selected one or more elements; A program to execute.

2. The verification item generation step generates one or more second verification items using one or more first verification items, the generation request, and the elements. The program according to claim 1.

3. The method further includes a search step of searching for one or more of the elements to realize the execution procedure, The verification item generation step generates one or more second verification items using one or more first verification items, the generation request, and the retrieved elements. The program according to claim 2.

4. The aforementioned reference information includes one or more inputs for each of the one or more elements, The aforementioned step of generating confirmation items is: A second confirmation item is generated to confirm one or more of the inputs other than the input confirmed by the first confirmation item. The program according to claim 1.

5. The confirmation item generation step generates a second confirmation item for confirming the reference information representing one or more of the elements used to generate the execution procedure. The program according to claim 1.

6. The aforementioned step of generating confirmation items is: The second confirmation item is generated for confirming the reference information that specifies one or more words or phrases included in the generation request that need to be specified. The program according to claim 1.

7. The information acquisition step involves acquiring the missing information by at least one of the following methods: a method for acquiring the missing information that has been input in response to an input request for the identified missing information, and a method for acquiring the missing information from a device outside the computer. The program according to claim 1.

8. The information acquisition step involves determining the acquisition order of one or more of the identified missing pieces of information, and acquiring the missing pieces of information according to the determined acquisition order. The program according to claim 1.

9. If the missing information cannot be obtained, the information acquisition step will not acquire the missing information that is acquired later in the acquisition order than the missing information that could not be acquired. The program according to claim 8.

10. The procedure generation step involves classifying the reference information and the missing information by type, and generating the execution procedure using the classified reference information and the missing information. The program according to claim 1.

11. The aforementioned type is such that the reference information and the missing information indicate at least one of the following: the input of the execution procedure, the output of the execution procedure, and the function of the execution procedure. The program according to claim 10.

12. The system further includes an output control step that outputs the reference information and information indicating the missing information, classified according to the aforementioned types. The program according to claim 10.

13. A request acquisition unit that acquires generation requests written in natural language that request the generation of one or more execution procedures, A confirmation item generation unit generates one or more second confirmation items other than the first confirmation items, using one or more predetermined first confirmation items, which are confirmation items for confirming reference information referenced when generating the execution procedure, and the generation request. A specification unit identifies one or more missing pieces of information that represent the reference information that cannot be obtained from the generation request, among the reference information confirmed by the first and second confirmation items, An information acquisition unit that acquires the identified missing information from an information source other than the generation request, A procedure generation unit that, using the generation request, the reference information obtained from the generation request, and the acquired missing information, selects one or more elements from one or more predetermined elements of an execution procedure and generates the execution procedure using the selected one or more elements, An information processing device equipped with the following features.

14. An information processing method performed by an information processing device, A request acquisition step that acquires a generation request written in natural language that requests the generation of one or more execution steps, A confirmation item generation step that generates one or more second confirmation items other than the first confirmation items, using one or more predetermined first confirmation items, which are confirmation items for confirming reference information referenced when generating the execution procedure, and the generation request, A selection step to identify one or more missing pieces of information that represent the reference information that cannot be obtained from the generation request, among the reference information confirmed by the first and second confirmation items, An information acquisition step of obtaining the identified missing information from a source other than the generation request, A procedure generation step that uses the generation request, the reference information obtained from the generation request, and the acquired missing information to select one or more elements from one or more predetermined elements of an execution procedure, and generates the execution procedure using the selected one or more elements; Information processing methods including

Images