Intent setting system, method, and program

Abstract

This framework provides a way to input natural language intents into an LLM (Language Language Machine) and configure network intents based on the inference results output by the LLM. If there are any deficiencies in the requirements of the natural language intents, the framework interactively obtains the necessary information from the user to resolve these deficiencies, thereby providing the desired network construction service without requiring the user to have specialized knowledge or an understanding of the system specifications. [Solution] The system includes an LLM control unit 20 that inputs natural language intents and each requirement definition item to the LLM 30, commands it to infer the setting value of each requirement definition item, and registers the setting value for each requirement definition item based on the inference results output by the LLM 30. The LLM control unit 20 requests information from the service user regarding the requirement definition items for which the LLM 30 could not infer the setting value, and uses the information obtained to command the LLM 30 to infer the setting value that could not be inferred.

Description

[Technical Field]

[0001] The present invention relates to an intent setting system, method, and program, and particularly to an intent setting system, method, and program for setting a network intent using an inference result obtained by inputting a natural language intent into a Large Language Model (LLM). [Background Art]

[0002] In mobile networks, factors such as higher speed, larger capacity, lower latency, mass simultaneous connections, autonomy, and ultra-high safety and reliability are expected. In realizing these networks, it is assumed that the functions of Network Functions (NF) will diversify, infrastructure will be hierarchized due to virtualization technology, and observation will become more complicated.

[0003] On the other hand, with the declining population, it has become difficult to secure human resources with specialized knowledge in network operation. As a solution to this problem, it is expected that networks will be constructed and operated autonomously.

[0004] For the realization of autonomous networks, the introduction of technology that defines the desired state of a network using intents (requirements that users demand of the network) and allows the network to autonomously construct and make adjustments to achieve that state is under consideration. However, setting the network intent that is input to construct a desired network requires specialized technical knowledge.

[0005] Against the background of such technical issues, intent control via natural language input has attracted attention. In an autonomous network (NW) based on intent control, using natural language as the intent input method makes it easier for users to use. In addition, due to the need to process natural language input without a defined format, LLMs have attracted attention as the processing system for this approach.

[0006] Non-patent document 1 discloses a proposed architecture for setting the correct intent by classifying intents entered by the user in natural language using natural language processing, then inferring the requirements necessary for the intent using an LLM, checking for any missing or conflicting information, and prompting the user for correct input.

[0007] Non-patent document 2 discloses a proposed architecture that classifies intents by inputting the user's natural language into an LLM, and then generates policies from the intents and performs inspections. [Prior art documents] [Non-patent literature]

[0008] [Non-Patent Document 1] Intent-Based Management of Next-Generation Networks: an LLM-Centric Approach [ https: / / ieeexplore.ieee.org / document / 10574890 ] [Non-Patent Document 2] LLM-Based Policy Generation for Intent-Based Management of Applications [ https: / / ieeexplore.ieee.org / document / 10327837 ] [Overview of the project] [Problems that the invention aims to solve]

[0009] Intents entered in natural language may contain ambiguity, such as including multiple keywords, which can make it difficult to determine which requirements to include in the network intent. For example, if the natural language intent input is "I want to build a network that streams video through a corporate VPN. We might also have occasional meetings using voice calls," the presence of multiple keywords ("corporate VPN," "video streaming," and "voice calls") can make it difficult to determine the necessary requirements for selecting a network intent.

[0010] Furthermore, because there is no defined format for the natural language used for intent input, it may lack the necessary information (requirements) for the system. For example, if a customer request is "We want to build a network for 4K video distribution. We want the distributed video to be uninterrupted. The studio for distribution is in Tokyo, and the viewing venue is in Osaka. Distribution will be stopped in case of network failure," but the natural language intent input is "We want to build a network for 4K video distribution," then it becomes difficult to determine the requirements necessary for selecting a network intent, and the inability to uniquely grasp the user's intent becomes a challenge.

[0011] The object of the present invention is to solve the above technical problems and to provide a framework that can provide the desired network construction service without requiring the user to have specialized knowledge or an understanding of the system specifications, by inputting natural language intents into an LLM and setting network intents based on the inference results output by the LLM, and by obtaining additional information necessary to compensate for any deficiencies in the requirements of the natural language intent through an interactive process with the user. [Means for solving the problem]

[0012] To achieve the above objective, the present invention is characterized by having the following configuration in an intent setting system that sets a network intent based on the inference result output by an LLM that receives a natural language intent as input.

[0013] (1) The system comprises an input unit that receives natural language intents from service users, a requirements definition control unit that controls the registration of setting values ​​to requirements definition items that define the requirements necessary for network construction, an LLM control unit that inputs natural language intents and each requirements definition item to the LLM and commands the LLM to infer the setting values ​​of each requirements definition item, and registers the setting values ​​of each requirements definition item based on the inference results output by the LLM, and an intent setting unit that sets network intents based on natural language intents, each requirements definition item and its setting values, wherein the LLM control unit further comprises means for requesting and obtaining additional information from service users regarding requirements definition items for which the LLM could not infer the setting values, and means for using the obtained additional information to command the LLM to infer the setting values ​​that could not be inferred.

[0014] (2) The LLM control unit repeatedly requests and obtains the additional information from the service user and instructs the LLM to infer the setting value using the obtained additional information until the registration of setting values ​​for all requirement definition items is completed.

[0015] (3) The requirements definition unit includes means for controlling the registration of setting values ​​for common requirements definition items common to network construction, and means for controlling the registration of setting values ​​for detailed requirements definition items that depend on the setting values ​​of the common requirements definition items. The LLM control unit selects detailed requirements definition items based on the setting values ​​of the common requirements definition items, inputs natural language intents and detailed requirements definition items to the LLM and commands inference regarding the setting values ​​of each detailed requirements definition item, registers setting values ​​for each detailed requirements definition item based on the inference results output by the LLM, and the intent setting unit sets network intents based on the setting values ​​of the common requirements definition items and detailed requirements definition items. [Effects of the Invention]

[0016] According to the present invention, the following effects are achieved.

[0017] (1) If the natural language intent input by the service user has insufficient requirements, additional information required to resolve such insufficient requirements can be acquired from the user in an interactive manner, so that the network intent can be accurately set without omission. Therefore, service users can enjoy their desired services without being required to have expertise or understanding of system specifications, which can be expected to improve the user's experience value and expand the application scope of the system.

[0018] (2) Acquiring additional information from the user in an interactive manner can be repeated until all setting values of the requirement definition items are registered, so that even an unskilled user without expertise can accurately set the network intent without omission.

[0019] (3) Requirement definition items are classified into common requirement definition items and detailed requirement definition items, and the detailed requirement definition items are selected based on the setting values of the common requirement definition items, so that only requirement definition items suitable for the service user's demand can be selectively set from among numerous requirement definition items. Brief Description of the Drawings

[0020] [Figure 1] It is a functional block diagram showing the configuration of main parts of the intent setting system according to the first embodiment of the present invention. [Figure 2] It is a sequence diagram (Part 1) showing the operation of the first embodiment of the present invention. [Figure 3] It is a sequence diagram (Part 2) showing the operation of the first embodiment of the present invention. [Figure 4] It is a sequence diagram (Part 3) showing the operation of the first embodiment of the present invention. [Figure 5] It is a diagram showing examples of common requirement definition items and their setting values. [Figure 6] It is a diagram showing an example of selecting detailed requirement definition items based on setting values of common requirement definition items. [Figure 7] It is a diagram showing an example of requirement definition items used for setting network intent. [Figure 8] It is a functional block diagram showing the configuration of the main parts of an intent setting system according to a second embodiment of the present invention. [Figure 9] It is a sequence diagram (Part 1) showing the operation of the second embodiment of the present invention. [Figure 10] It is a sequence diagram (Part 2) showing the operation of the second embodiment of the present invention.

Mode for Carrying Out the Invention

[0021] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a functional block diagram showing the configuration of the main parts of an intent setting system 1 according to a first embodiment of the present invention. By applying a request (natural language intent) input in natural language by a service user regarding network construction to an LLM, a network intent in which requirements necessary for network construction are described in detail is set. The network intent is then provided to an intent-based network construction system (not shown) and used for constructing a network to be provided to the service user.

[0022] In the present invention, as service users, it is assumed that there are general customers who desire to receive a network construction service but do not have specialized knowledge, and operators who perform network operation management. If the input natural language intent is deficient in that information necessary for LLM and natural language processing is insufficient, the process of repeatedly prompting the service user to input additional information is performed.

[0023] The intent setting system 1 of the present invention mainly includes: an input unit 10 that accepts a natural language intent input from a service user; an LLM control unit 20 that controls input / output of an LLM 30 based on the input natural language intent; a requirement definition unit 40 that controls requirement definition items that define requirements essential for setting a network intent; and an intent setting unit 50 that sets a network intent based on setting values of the requirement definition items.

[0024] Such an intent setting system 1 can be configured by implementing an application (program) that realizes each of the functions detailed below on a general-purpose computer or server equipped with a CPU, ROM, RAM, bus, storage, interface, etc., or on a portable smartphone or tablet device. Alternatively, it can be configured as a dedicated machine or single-function machine in which part of the application is implemented in hardware or software.

[0025] The input unit 10 receives requests related to network construction in the form of natural language intents, such as "We want to build a mobile network for video distribution" or "We want to expand the network for operational purposes."

[0026] The requirements definition unit 40 includes a requirements definition control unit 401 and a requirements definition DB 402, in which "common requirements definition items" and "detailed requirements definition items" are pre-stored.

[0027] As shown in Figure 5 as an example, the common requirements definition items consist of a combination of at least one item and its setting value that defines a common or essential requirement that is independent of the network in intent settings related to network construction. As shown in Figure 6 as an example, the detailed requirements definition items consist of a combination of at least one item and its setting value that is additionally selected based on the natural language intent and the setting value of the common requirements definition items.

[0028] The LLM30 outputs inference results corresponding to the input inference command. The inference results are generated based on training data. In addition to publicly available language models, LLM30 can use a dedicated language model tuned for the service. In this invention, LLM30 can be used without specifying its performance or characteristics.

[0029] The LLM control unit 20 inputs the natural language intent and each requirement definition item to the LLM 30, commands it to infer the setting value for each requirement definition item, and registers the setting value for each requirement definition item based on the inference result output by the LLM.

[0030] The intent setting unit 50 includes a network intent control unit 501 and a network intent database 502. The network intent database 502 has multiple network intent items stored in advance. Network intent items can use formats defined by standardization organizations (e.g., TM Forum, 3GPP) or proprietary formats. The network intent control unit 501 selects and retrieves items from the network intent database 502 regarding network intents and performs control to ensure that specific values ​​are set for those items without omission.

[0031] Next, the operation of the first embodiment of the present invention will be described in detail with reference to the sequence diagrams in Figures 2 to 4. Here, we will explain a method for configuring network intents that can improve the inference accuracy and speed of LLM30 by compensating for deficiencies in natural language intents entered by service users, and that can fully reflect the requests of service users.

[0032] In step S1, the service user inputs a natural language intent into the input unit 10. The input natural language intent is transmitted to the LLM control unit 20. In this embodiment, there are no restrictions on the input format of the natural language intent, and users can freely express their desired actions in natural language, just as in human conversation. Here, we will continue the explanation assuming that the request "I want to create a network for 4K video distribution" is input as a natural language intent.

[0033] In step S2, the LLM control unit 20 responds to the input of the natural language intent and requests the requirements definition control unit 401 of the requirements definition unit 40 to send common requirements definition items. In step S3, the requirements definition control unit 401 responds to the request, refers to the requirements definition DB 402, and sends the common requirements definition items along with the registered setting values ​​to the LLM control unit 20.

[0034] In this embodiment, as shown in Figure 5(a) as an example, three items, "Service Type," "Network Type," and "Customer Type," are pre-stored in the requirements definition DB402 as common requirements definition items. Initially, none of the items have registered settings, so only the common requirements definition items are sent. Each common requirements definition item is not limited to just one setting value, but may be combined with multiple setting values.

[0035] In step S4, the LLM control unit 20 sends an inference command to the LLM 30 to infer the setting values ​​for each item based on the content of the natural language intent input, with respect to the acquired common requirements definition items. The inference command includes, for example, the following information.

[0036] (1) Task Description: Infer the setting value of the common requirements definition item from the natural language intent input. (2) Input: "Build a mobile network capable of 4K video streaming" (3) Constraint: Output must be sent in JSON format. (4) Common requirement definition items: (4-1) "Service Type" (4-2) "Network type" (4-3) "Customer type"

[0037] In step S5, the LLM 30 performs inference in response to the inference command and sends the inference results for each setting value back to the LLM control unit 20. In step S6, the LLM control unit 20 registers the setting values ​​for each common requirement definition item based on the inference results and sends them to the requirement definition control unit 401.

[0038] In step S7, the requirements definition control unit 401 stores the combination of the common requirements definition items and their setting values ​​in the requirements definition DB 402 in a state that can be referenced later. The requirements definition control unit 401 further determines whether setting values ​​have been registered for all of the common requirements definition items. If there are any unregistered setting values ​​because the LLM 30 could not infer them, the process proceeds to step S8, where the LLM control unit 20 requests the service user to input additional information so that the LLM 30 can infer the unregistered setting values.

[0039] In step S9, the LLM control unit 20 requests the service user to input the additional information. For example, as shown in Figure 5(b), if the settings for "video distribution" and "mobile network" have been registered as inference results for the common requirements definition items "service type" and "network type," but the setting for "customer type" has not been registered, a request such as "The content of "customer type" could not be estimated from the input. Please enter additional customer information about the service user" is sent.

[0040] After that, the process returns to step S1, waits for additional information from the service user, and then proceeds to step S2. From this point onward, each of the above steps is repeated until the inference and registration of setting values ​​are completed for all common requirement definition items. In step S3 during the repetition, for common requirement definition items for which setting values ​​have already been registered, those registered setting values ​​are also sent.

[0041] Subsequently, as shown in Figure 5(c), once the registration of setting values ​​for all common requirement definition items is complete and storage in the requirement definition DB 402 is complete, the process proceeds to step S10. In step S10, the LLM control unit 20 requests the requirement definition control unit 401 to send detailed requirement definition information. In step S11, the requirement definition control unit 401 selects detailed requirement definition items from the requirement definition DB 402 based on the setting values ​​of the common requirement definition items and sends them to the LLM control unit 20.

[0042] Figure 6 shows an example of detailed requirements definition items and their selection method. Detailed requirements definition items consist of combinations with their setting values, and at least one is selected from a predetermined group of items based on the setting values ​​of the common requirements definition items.

[0043] For example, if the setting for the common requirement definition item "Service Type" is "Video Distribution" as shown in Figure (a), then "Network Quality," "Number of Simultaneous Connections," "Redundancy," and "Location of Use" will be selected as detailed requirement definition items, as shown in Figure (b).

[0044] Similarly, if the setting for the common requirement definition item "Service Type" is "Corporate VPN" as shown in Figure (d), then "Network Quality," "Number of Accounts," "VPN Information," and "Connection Location" will be selected as detailed requirement definition items, as shown in Figure (e).

[0045] In step S12, the LLM control unit 20 sends an inference command to the LLM 30 that causes the LLM 30 to infer setting values ​​for each of the selected detailed requirement definition items from the content of the natural language intent input by the service user. An example of an inference command is shown below.

[0046] (1) Task Description: For each item listed in the detailed requirements definition section, extract and set the parameters that can be inferred from the input. (2) Input: "Build a mobile network capable of 4K video streaming" (3) Constraint: Output must be sent in JSON format. (4) Detailed requirement definition items: (4-1) Network Quality (4-2) Number of simultaneous connections (4-3) Redundancy (4-4) Place of use

[0047] In step S13, the LLM 30 outputs the inference result for the inference command to the LLM control unit 20. In step S14, the LLM control unit 20 extracts the setting values ​​for each detailed requirement definition item from the outputted inference result and sends them to the requirement definition control unit 401.

[0048] For example, as shown in Figure 6(c), the "Network Quality" setting is set to "High," the "Number of Simultaneous Connections" setting is set to "10,000," the "Redundancy" setting is set to "None," and the "Location" setting is set to "Tokyo and Osaka." Alternatively, as shown in Figure 6(f), the "Network Quality" setting is set to "Low," the "Number of Accounts" setting is set to "100,000," the "VPN Information" setting is set to "Automatic," and the "Connection Location" setting is set to "Tokyo and Chiba."

[0049] Thus, in this embodiment, the setting values ​​for detailed requirements definition items can also be registered in a conversational format with the service user, so general terms that the service user can understand are used as item names. For example, instead of using items such as delay or jitter to express network quality, network quality can be categorized into stages such as "high," and instead of using business names (e.g., Osaka DC - Otemachi DC) for connection points, specific names such as regional names (Osaka and Tokyo) can be used.

[0050] In step S15, the detailed requirements definition items and their settings are stored in the requirements definition DB 402 in a way that makes them accessible. Then, it is determined whether or not settings have been registered for all of the detailed requirements definition items, and if there are any items for which settings have not been registered, the process proceeds to step S16. In step S16, in order to allow the LLM 30 to infer the unregistered settings, the LLM control unit 20 is requested to ask the service user to input additional information regarding the items for which settings have not been registered.

[0051] In step S17, the LLM control unit 20 requests additional information from the service user. Then, in step S18, when the service user enters a natural language intent, the process returns to step S10.

[0052] From this point onward, the above steps are repeated until the inference and registration of setting values ​​are completed for all detailed requirements definition items. In step S11 during the repetition, for detailed requirements definition items for which setting values ​​have already been registered, those registered setting values ​​are also sent to the LLM control unit 20.

[0053] Subsequently, once the registration of setting values ​​for all detailed requirements definition items is complete and they have been stored in the requirements definition DB402, the process proceeds to step S19, where all requirements definition items and their setting values ​​are sent to the LLM control unit 20, as shown in Figure 7 as an example.

[0054] In step S20, all requirement definition items and their settings are sent to the network intent control unit 501. In step S21, the network intent control unit 501 refers to the network intent DB 502 and selects network intent items based on each combination of requirement definition item and its setting. If it is rule-based, the selection is made based on, for example, the following rules.

[0055] • Rule: If the service type is video streaming, select the network intent item for video streaming.

[0056] In step S22, you configure the network intent items. Here, the configuration is done based on each requirement definition item, and an example of rule-based configuration is shown below.

[0057] Rule R1: If the service type is video streaming and the network quality is "high", the latency must be 10 msec or less.

[0058] Rule R2: If the location is Tokyo, select an available location from "Otemachi DC", "Shinjuku DC", or "Shibuya DC".

[0059] In step S23, the configured network intent is sent to the LLM control unit 20. The LLM control unit 20 provides the received network intent to the intent-based network construction system (not shown).

[0060] As a result, it becomes possible to create network intents related to the network construction requested by the customer from natural language intent input.

[0061] Figure 8 is a functional block diagram of the intent setting system 1 according to the second embodiment of the present invention. Reference numerals identical to those in the first embodiment represent the same or equivalent parts, so their explanation is omitted. This embodiment is characterized by the addition of a user information management DB 60 to improve the accuracy of detailed requirements definition and network intent creation by utilizing past performance information of existing users.

[0062] Figures 9 and 10 are sequence diagrams showing the operation of this embodiment. Here, the illustration and description of parts identical to those of the first embodiment are omitted, and the explanation will focus on the differences.

[0063] In step S14 of Figure 9, the LLM control unit 20 receives the setting values ​​of the detailed requirements definition items inferred by the LLM 30 and transmits them to the requirements definition control unit 401. In step S14-1, records in the user information management DB 60 that are similar to each setting value of the inferred detailed requirements definition items are searched for, for example, by calculating cosine similarity. If similar results exist, the procedure proceeds to step S14-2, where each setting value of the detailed requirements definition item currently being set is overwritten with the setting values ​​of the similar results, and thereafter the same procedure as in the first embodiment is executed.

[0064] For example, if LLM30 infers setting values ​​for m (m ≤ n) of the total n detailed requirement definition items in this case, and is unable to infer setting values ​​for the remaining nm items, and similar records to the m inferred setting values ​​are already registered as actual data in the user information management DB60, then the LLM30 will evaluate the setting values ​​that are identical to the setting values ​​for the same n detailed requirement definition items registered in the user information management DB60 as actual appropriate values, and will overwrite or newly register them as the setting values ​​for all n detailed requirement definition items in this case.

[0065] Similarly, when the network intent control unit 501 sets a network intent in step S22 of Figure 10, in step S22-1, it searches for a record in the user information management DB 60 that is similar to the current network intent setting result, for example by calculating cosine similarity. If a similar result exists, the process proceeds to step S22-2, where the current network intent being set is overwritten with the actual setting values ​​registered in the user information management DB 60, and thereafter the same procedure as in the first embodiment is executed.

[0066] Furthermore, according to each of the above embodiments, if there are any deficiencies in the natural language intent entered by the service user, the necessary information to resolve these deficiencies can be obtained from the user in a dialogue format, thus enabling the network intent to be set accurately and without omission. Therefore, service users can enjoy the desired service without being required to have specialized knowledge or an understanding of the system specifications, and an improvement in the user experience value and an expansion of the system's application area can be expected.

[0067] As a result, it will be possible to contribute to Goal 9, "Build resilient infrastructure and promote inclusive and sustainable industrialization," and Goal 11, "Make cities inclusive, safe, resilient and sustainable," both of which are led by the United Nations. [Explanation of Symbols]

[0068] 1…Intent setting system, 10…Input unit, 20…LLM control unit, 30…LLM, 40…Requirements definition unit, 50…Intent setting unit, 60…User information management DB, 401…Requirements definition control unit, 402…Requirements definition DB, 501…Network intent control unit, 502…Network intent DB

Claims

1. In an intent setting system that sets network intents based on the inference results output by an LLM that takes a natural language intent as input, An input unit that accepts natural language intent input from service users, A requirements definition unit that controls the registration of setting values ​​to requirements definition items that define the requirements necessary for network construction, An LLM control unit inputs the aforementioned natural language intent and each requirement definition item to the LLM, commands it to infer the setting value of each requirement definition item, and registers the setting value for each requirement definition item based on the inference result output by the LLM. The system comprises an intent setting unit that sets network intents based on the aforementioned natural language intents, each requirement definition item, and their setting values, The LLM control unit is A means of requesting and obtaining additional information from service users regarding requirement definition items for which LLM could not infer the setting value, An intent setting system characterized by comprising means for instructing the LLM to infer setting values ​​that could not be inferred using the additional information obtained.

2. The intent setting system according to claim 1, characterized in that the LLM control unit repeatedly requests and obtains the information from the service user and instructs the LLM to infer the setting value using the obtained information until the registration of setting values ​​for all requirement definition items is completed.

3. The requirements definition unit, A means for controlling the registration of setting values ​​to common requirements definition items common to network construction, The system includes means for controlling the registration of setting values ​​to detailed requirement definition items that depend on the setting values ​​of the common requirement definition items, The LLM control unit is Select the detailed requirements definition items based on the settings of the common requirements definition items mentioned above. The aforementioned natural language intent and detailed requirements definition items are input into the LLM, and an inference is instructed regarding the setting values ​​of each detailed requirements definition item. Based on the inference results output by LLM, set values ​​are registered for each detailed requirements definition item. The intent setting system according to claim 2, characterized in that the intent setting unit sets network intents based on the setting values ​​of the common requirements definition items and the detailed requirements definition items.

4. The intent setting system according to any one of claims 1 to 3, characterized in that, if the requirements definition unit has recorded a setting value similar to the inference result output by the LLM as past performance, it overwrites the inference result output by the LLM using the past setting value.

5. The intent setting system according to any one of claims 1 to 3, characterized in that if the intent setting unit has recorded a network intent similar to the one set based on the natural language intent, each requirement definition item, and its setting value as past performance, it overwrites the set network intent using the past network intent.

6. In an intent setting method in which a computer sets a network intent based on the inference result output by an LLM that takes a natural language intent as input, The service accepts natural language intent input from users. The requirements definition items that define the natural language intent and the requirements necessary for network construction are input into the LLM, and an inference command is issued regarding the setting values ​​of each requirements definition item. LLM requests and obtains information from service users regarding requirements definition items for which it could not infer the setting value. Using the information obtained above, the LLM is instructed to infer the setting value that could not be inferred above. Based on the inference results output by the LLM in response to each of the aforementioned commands, set values ​​are registered for each requirement definition item. An intent setting method characterized by setting a network intent based on the aforementioned natural language intent, each requirement definition item, and their setting values.

7. In an intent setting program that sets a network intent based on the inference result output by an LLM that takes a natural language intent as input, The procedure for receiving natural language intent input from service users, A procedure for inputting the requirements definition items that define the natural language intent and the requirements necessary for network construction into the LLM and instructing it to infer the setting values ​​for each of the requirements definition items, The procedure for requesting and obtaining information from service users regarding requirements definition items for which LLM could not infer the setting value, A procedure for instructing the LLM to infer the setting value that could not be inferred using the acquired information, A procedure for registering setting values ​​for each requirement definition item based on the inference results output by LLM in response to each of the aforementioned commands, An intent setting program characterized by causing a computer to execute a procedure for setting a network intent based on the aforementioned natural language intent, each requirement definition item, and its setting value.

Images