Internet of Things control method and device based on natural language, equipment and medium

Abstract

The invention discloses an Internet of Things control method, device and equipment based on a natural language and a medium, and the method comprises the steps: receiving a natural language text of a user, and carrying out the intention recognition of the natural language text through a preset large language model, and obtaining intention analysis data; judging whether a tool needs to be called or not according to the intention analysis data, and if the tool needs to be called, selecting a target tool from a preset tool list according to the intention analysis data to obtain a target tool set; and sorting the target tools in the target tool set according to a preset business logic chain and the intention analysis data, generating a tool calling sequence, and calling the target tools in the target tool set in sequence according to the tool calling sequence to realize Internet of Things control. According to the invention, through intelligent analysis of the control task of the Internet of Things, the intelligent degree of control of the Internet of Things and the adaptability of complex scenes can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent Internet of Things, and in particular to an Internet of Things control method and device based on natural language, equipment and medium. BACKGROUND

[0002] Internet of Things control is the core technology to realize smart home, industrial automation, smart city and other application scenarios. Its purpose is to efficiently and accurately manage and operate various networked devices through remote or automatic instructions, thereby improving system response speed, reducing labor costs, and enhancing the intelligence and collaboration of the overall system. Traditional control methods usually rely on special software, fixed instruction sets or manual intervention, which are difficult to adapt to complex and changing user needs, especially in high-concurrency, multi-device collaboration and non-technical user scenarios, there are obvious efficiency and usability bottlenecks.

[0003] In the prior art, Internet of Things control, especially in the field of industrial automation, mainly relies on two ways: one is to implement control logic by manually writing code, which requires users to have professional programming skills and in-depth understanding of platform interfaces; the second is to configure through a visual interface, which reduces the technical threshold to some extent, but still requires users to be familiar with platform operations and professional terms. Both of these two methods have significant defects: one is that the threshold is high and non-professionals cannot operate; two is that the flexibility and scalability are insufficient, which cannot adapt to dynamically changing control requirements; three is the lack of understanding of natural language, which cannot realize "people-oriented" intelligent interaction.

[0004] Although many fields have gradually begun to develop Internet of Things control technology based on natural language, most of them can only parse simple instructions through large language models to control pre-recorded target Internet of Things devices. This method lacks effective support for complex, multi-step tasks, cannot dynamically schedule, sort and combine multiple functional tools (such as API calls, database queries and computing services) according to business logic, and cannot interactively complete parameters when missing, resulting in that it can only handle simple device control scenarios, but cannot handle complex Internet of Things control tasks that require multiple services to collaborate. SUMMARY

[0005] The present application provides an Internet of Things control method, device, equipment and medium based on natural language, which can improve the intelligence and complexity of Internet of Things control by intelligently analyzing Internet of Things control tasks.

[0006] In a first aspect, the present application provides an Internet of Things control method based on natural language, comprising:

[0007] receive a natural language text of a user, and perform intent recognition on the natural language text through a preset large language model to obtain intent analysis data; wherein the intent analysis data includes an intent classification result and an entity parameter set;

[0008] According to the intent analysis data, it is judged whether a tool needs to be called, if the tool needs to be called, a target tool set is selected from a preset tool list according to the intent analysis data;

[0009] According to the preset business logic chain and the intent analysis data, the target tools in the target tool set are sorted to generate a tool calling sequence, and the target tools in the target tool set are called in sequence according to the tool calling sequence to realize Internet of Things control.

[0010] The embodiment of the application realizes deep understanding of user instructions based on the natural language text of the user through the large language model, wherein the intent classification result clearly indicates the core needs of the user, and the entity parameter set extracts key information, laying a foundation for subsequent tool selection and device control; through the target tool set obtained according to the intent analysis data, the precise mapping of demand-tool is realized, and the target tool is selected from the tool list to ensure that the called tool matches the user's intention, so that the user does not need to perform parameter configuration or step-by-step guidance and other redundant operations, and the internal workflow required to achieve the purpose is directly analyzed according to the user's intention, simplifying the user's use process and operation threshold, improving the use convenience and adaptation ability to complex requirements; through the preset business logic chain and the intent analysis data, the tool calling sequence is generated to ensure that the tool calling conforms to the actual business process, avoiding control failure due to chaotic order. In summary, the embodiment of the application realizes the intelligent and process-oriented conversion from natural language to Internet of Things control through the implementation logic chain of "accurate intent analysis-tool dynamic selection-tool sequential calling", without the need for the user to master professional knowledge, while solving the problems of complicated operation, disconnection between demand and execution in traditional control, and compared with the single control logic of directly controlling the device according to the user's intention in the prior art, the application disassembles the control instruction at the tool level, thereby realizing the adaptation of different complex scenes through the combination of multiple tools. Compared with the prior art, the application can improve the intelligent degree and complex scene adaptability of Internet of Things control through intelligent analysis of Internet of Things control tasks.

[0011] Further, the natural language text is subjected to intent recognition through a preset large language model to obtain intent analysis data, specifically:

[0012] The interaction context information transmitted by the BI-IOT interface is obtained, and the natural language text is subjected to information supplementing according to the interaction context information and a preset instruction template to obtain an intermediate natural language text;

[0013] mapping the intermediate natural language text through a pre-stored professional term library to obtain a final natural language text;

[0014] performing intent classification on the final natural language text through a pre-set large language model to obtain an intent classification result, and performing parameter extraction on the final natural language text to obtain an entity parameter set.

[0015] The embodiment of the application supplements incomplete instruction information by using interactive context and instruction templates, avoids parsing failure caused by ambiguous instructions, improves the integrity and accuracy of instruction parsing, maps user colloquial expressions into professional terms recognizable by the system to eliminate semantic ambiguity, and reduces the operation threshold of users and improves the accuracy of semantic understanding without the need for users to master professional terms, further optimizes the results of intent classification and parameter extraction based on the completed and standardized text, ensures more accurate parsed data, and provides more reliable basis for subsequent tool selection and device control, so as to improve the parsing accuracy through multi-step preprocessing.

[0016] Further, the intent classification is performed on the final natural language text through a pre-set large language model to obtain an intent classification result, specifically:

[0017] The semantic features of the final natural language text are extracted through a pre-set natural language processing model to obtain text semantic features;

[0018] The text semantic features are fused through the interactive context information to obtain fused multi-dimensional features;

[0019] The fused multi-dimensional features are input into a pre-set classifier, so that the classifier outputs an intent classification result according to the fused multi-dimensional features.

[0020] The embodiment of the application extracts deep semantic information of the text to provide a basis for subsequent feature fusion and intent classification, improves the basis granularity of intent classification, realizes multi-dimensional fusion of text semantics to interactive scenarios in combination with context and text semantic features, and improves the scene adaptability and accuracy of intent classification, and the classifier accurately classifies multi-dimensional features to ensure the accuracy of the intent classification result.

[0021] Further, the parameter extraction is performed on the final natural language text to obtain an entity parameter set, specifically:

[0022] The parameter extraction is performed on the final natural language text according to a pre-set parameter structured template through a pre-set named entity recognition technology to obtain an entity parameter set.

[0023] This invention, through standardized parameter extraction techniques and templates, ensures the uniformity of entity parameter set format and the integrity of information, providing standardized data support for subsequent tool selection, sorting, and invocation, and ultimately improving the smoothness and reliability of data flow in all aspects of the system.

[0024] Furthermore, based on the intent parsing data, a target tool is selected from a preset tool list to obtain a target toolset, specifically:

[0025] Based on a preset intent-tool mapping table, the intent classification results are associated with tools to select target tools from a preset tool list, thereby obtaining a candidate tool set;

[0026] Based on the entity parameter set, the candidate toolset is filtered to obtain the target toolset.

[0027] This invention establishes a fixed association between intent and tool through a mapping table, quickly filtering out candidate tools that match the intent and reducing the time cost of tool selection; it further filters candidate tools by combining entity parameters to ensure that the target tool matches the specific parameters and avoids incompatibility between tools and parameters that leads to call failure.

[0028] Furthermore, based on the preset business logic chain and the intent parsing data, the target tools in the target toolset are sorted to generate a tool invocation sequence, specifically as follows:

[0029] Based on the preset business logic chain and the intent parsing data, the target tools in the target toolset are sorted to obtain the tool invocation order, and the inter-tool transmission parameters are generated based on the tool invocation order.

[0030] The tool invocation order and the parameters passed between the tools are integrated to obtain the tool invocation sequence.

[0031] This invention, through sorting tools according to business logic, ensures that the calling order conforms to the actual operation process, while generating parameters for inter-tool transfer to realize data flow between tools, and ultimately achieves data interoperability and process standardization for multi-tool collaboration, thereby improving the executability of multi-tool collaboration.

[0032] Furthermore, before sequentially invoking the target tools in the target toolset according to the tool invocation sequence, the method further includes:

[0033] According to the preset tool call sequence template, the tool call sequence is subjected to an integrity check to obtain an integrity check result. If the integrity check result indicates that a parameter is missing, the corresponding information request data is generated and sent to the user.

[0034] This invention uses a template to check whether key parameters are missing in the sequence, thus avoiding tool call failures due to missing parameters; if missing parameters are found, the system proactively requests the user to supplement them, improving system interactivity and fault tolerance.

[0035] Secondly, embodiments of the present invention provide an Internet of Things (IoT) control device based on natural language, including an intent parsing module, a toolset acquisition module, and a device control module, wherein...

[0036] The intent parsing module is used to receive natural language text from users and perform intent recognition on the natural language text using a preset large language model to obtain intent parsing data; wherein, the intent parsing data includes intent classification results and entity parameter sets;

[0037] The toolset acquisition module is used to determine whether a tool needs to be invoked based on the intent parsing data. If a tool needs to be invoked, the module selects a target tool from a preset tool list based on the intent parsing data to obtain the target toolset.

[0038] The device control module is used to sort the target tools in the target toolset according to the preset business logic chain and the intent parsing data, generate a tool call sequence, and call the target tools in the target toolset in sequence according to the tool call sequence, so as to realize the control of the target Internet of Things.

[0039] This invention employs an intent parsing module that, based on the user's natural language text, deeply understands user commands using a large language model. The intent classification results clearly identify the user's core needs, and the entity parameter set extracts key information, laying the foundation for subsequent tool selection and device control. A toolset acquisition module, based on the intent parsing data, obtains the target toolset, achieving a precise mapping between needs and tools. It also filters target tools from a tool list to ensure that the invoked tools match the user's intent, eliminating redundant operations such as parameter configuration or step-by-step guidance. The internal workflow required to achieve the stated purpose is directly parsed based on the user's intent, simplifying the user experience and reducing operational barriers, improving ease of use and adaptability to complex requirements. Finally, a device control module generates a tool invocation sequence based on a preset business logic chain and the intent parsing data, ensuring that tool invocation conforms to the actual business process and avoiding control failures due to disordered sequence.

[0040] Thirdly, embodiments of the present invention provide a terminal device, including: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus;

[0041] The memory is used to store at least one executable instruction that causes the processor to perform operations as described in any of the above-described natural language-based Internet of Things control methods.

[0042] Fourthly, embodiments of the present invention provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device or apparatus where the computer-readable storage medium is located to perform the Internet of Things control method based on natural language as described in any of the above.

[0043] The above description is merely an overview of the technical solutions of the embodiments of the present invention. In order to better understand the technical means of the embodiments of the present invention and to implement them in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0044] Figure 1 A schematic diagram of an Internet of Things (IoT) control method based on natural language provided in an embodiment of the present invention;

[0045] Figure 2 A data interaction diagram of an Internet of Things (IoT) control system based on natural language is provided for an embodiment of the present invention;

[0046] Figure 3 This is a structural diagram of an Internet of Things (IoT) control device based on natural language, provided as an embodiment of the present invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Example 1:

[0049] like Figure 1 As shown, an embodiment of the present invention provides an Internet of Things (IoT) control method based on natural language, comprising the following steps:

[0050] S101, Receive the user's natural language text, and perform intent recognition on the natural language text through a preset large language model to obtain intent parsing data; wherein, the intent parsing data includes intent classification results and entity parameter sets;

[0051] In this embodiment, the step of performing intent recognition on the natural language text using a preset large language model to obtain intent parsing data specifically involves: acquiring interaction context information transmitted via the B I-IoT interface, and supplementing the natural language text with information based on the interaction context information and a preset instruction template to obtain intermediate natural language text; mapping the intermediate natural language text using a pre-stored terminology database to obtain final natural language text; classifying the intent of the final natural language text using the preset large language model to obtain intent classification results, and extracting parameters from the final natural language text to obtain an entity parameter set.

[0052] In one specific embodiment, the user inputs natural language commands through the B I-IoT interface, serving as the initial interaction entry point for the system. The B I-IoT module can also obtain context, feedback results, and other information from the user's end.

[0053] In one specific embodiment, the MCP client module acts as an intermediate communication bridge, receiving natural language text information from B I-IoT and passing it to the LLM module, while simultaneously forwarding the results back to the IoT platform. The client is also responsible for coordinating the acquisition and updating of the tool list.

[0054] In this embodiment, the step of classifying the final natural language text using a preset large language model to obtain an intent classification result specifically involves: extracting semantic features from the final natural language text using a preset natural language processing model to obtain text semantic features; fusing the text semantic features using the interaction context information to obtain fused multi-dimensional features; and inputting the fused multi-dimensional features into a preset classifier so that the classifier outputs an intent classification result based on the fused multi-dimensional features.

[0055] In one specific embodiment, an NLP model is used as the natural language processing model.

[0056] In one specific embodiment, the LLM semantic parsing module, based on large language model technology, understands natural language instructions and determines whether a tool (such as a query service or device control) needs to be invoked. Its main outputs are the tool invocation determination result, the invocation parameters for selecting the appropriate tool, and the parsed structure of the user intent.

[0057] In one specific embodiment, the LLM module processes natural language instructions as follows:

[0058] (1) Intent Classification: The classifier identifies which category a user instruction belongs to, such as "device control", "tool call", "data query", or "fault diagnosis".

[0059] For example, "Create device" is categorized as a device configuration operation, while "View temperature" is categorized as a data acquisition operation.

[0060] (2) Parameter extraction (Slot Filling): Extract entity information such as device name, tool name, and parameter items.

[0061] (3) Determine whether tools are needed.

[0062] Finally, the LLM generates structured instructions and sends them to the "decision calling tool" module.

[0063] In this embodiment, the final natural language text is subjected to parameter extraction to obtain an entity parameter set. Specifically, the final natural language text is subjected to parameter extraction using a preset named entity recognition technology and a preset parameter structure template to obtain an entity parameter set.

[0064] In one specific embodiment, Named Entity Recognition (NER) technology is used for named entity recognition.

[0065] S102, based on the intent parsing data, determine whether it is necessary to call a tool. If it is necessary to call a tool, select a target tool from a preset tool list based on the intent parsing data to obtain a target tool set.

[0066] In this embodiment, a target tool is selected from a preset tool list based on the intent parsing data to obtain a target tool set. Specifically, the intent classification result is associated with a tool according to a preset intent-tool mapping table to select a target tool from the preset tool list to obtain a candidate tool set; the candidate tool set is then filtered according to the entity parameter set to obtain the target tool set.

[0067] In one specific embodiment, after the LLM output, the system completes tool invocation through the MCP service module based on the inference results, including but not limited to: Python function execution, computational logic calculation, HTTP interface request, database query, and device resource invocation.

[0068] In one specific embodiment, the MCP service module is used to uniformly handle invocation requests from backend tools. The MCP service supports multiple tool protocols and encapsulation logic, and can be extended to access various IoT control platforms and data sources.

[0069] In one specific embodiment, the tool resource library module stores a list of all tools supported by the system. The system dynamically determines and recommends suitable tools for use. Tool types cover device control, timing services, data acquisition, etc.

[0070] In this embodiment, the step of sorting the target tools in the target tool set according to the preset business logic chain and the intent parsing data to generate a tool call sequence specifically involves: sorting the target tools in the target tool set according to the preset business logic chain and the intent parsing data to obtain the tool call order, and generating inter-tool transmission parameters according to the tool call order; and integrating the tool call order and the inter-tool transmission parameters to obtain the tool call sequence.

[0071] S103, according to the preset business logic chain and the intent parsing data, sort the target tools in the target tool set, generate a tool call sequence, and call the target tools in the target tool set in sequence according to the tool call sequence to realize IoT control.

[0072] In this embodiment, before sequentially calling the target tools in the target toolset according to the tool call sequence, the method further includes: performing an integrity check on the tool call sequence according to a preset tool call sequence template to obtain an integrity check result; if the integrity check result indicates that a parameter is missing, generating corresponding information request data and sending the information request data to the user.

[0073] In one specific embodiment, a data query and time series service module is included. When a user requests historical data or device status information, the system accesses this module through the MCP service module and returns the results for the language model to further generate the final response.

[0074] In one specific embodiment, the tool invocation and execution process is as follows: If a tool needs to be invoked, the tool scheduling module searches for a matching tool in the tool registration information, returns a structure of type calltool, and encapsulates the structured parameters into a request body to initiate a request to the MCP client.

[0075] If the requested parameters are incorrect or the parameter type is wrong, the server will return the corresponding prompt. After receiving the user's response, the large model will re-extract the parameters to call the tool again. If the tool is not called again, the result will be returned directly.

[0076] Preferably, the embodiments of the present invention are applicable to industrial automation scenarios, such as robotic arm control and factory machinery control.

[0077] In one specific embodiment, such as Figure 2 As shown, Figure 2The present invention provides a data interaction diagram for an Internet of Things (IoT) control system based on natural language. The core interaction process of the IoT control system includes: (1) the user inputs a natural language command at the BI-IOT terminal; (2) after receiving the input, the MCP client transmits the command to the LLM module; (3) the LLM module parses the semantics of the command and determines whether to call a tool; (4) if a tool needs to be called, the system selects a suitable tool from the tool list and calls it through the MCP service; (5) the MCP service accesses the corresponding backend service or resource (such as database, time series service and device resources); (6) the call result is returned to the LLM module, and the LLM module generates a natural language response; (7) the final result is returned to the BI-IOT terminal, completing one intelligent interaction.

[0078] This invention, through its embodiments, deeply understands user commands using a large language model based on the user's natural language text. The intent classification results clearly identify the user's core needs, and the entity parameter set extracts key information, laying the foundation for subsequent tool selection and device control. By analyzing the intent parsing data, a target toolset is obtained, achieving a precise mapping between needs and tools. Target tools are then selected from the tool list to ensure that the invoked tools match the user's intent, eliminating redundant operations such as parameter configuration or step-by-step guidance. The internal workflow required to achieve the stated purpose is directly parsed based on the user's intent, simplifying the user experience and reducing operational barriers, improving ease of use and adaptability to complex needs. Furthermore, by generating a tool invocation sequence based on a preset business logic chain and the intent parsing data, the tool invocation conforms to the actual business process, avoiding control failures due to disordered sequence. In summary, this invention, through its implementation logic chain of "precise intent parsing - dynamic tool selection - orderly tool invocation," achieves intelligent and streamlined transformation from natural language to IoT control. This eliminates the need for users to possess specialized knowledge and solves the problems of cumbersome operation and disconnect between requirements and execution in traditional control. Furthermore, compared to the single control logic of direct device control based on user intent common in existing technologies, this invention decomposes control commands at the tool level, thereby enabling adaptation to different complex scenarios through the combination of various tools. Compared to existing technologies, this invention improves the intelligence level and adaptability to complex scenarios of IoT control through intelligent parsing of IoT control tasks.

[0079] Example 2:

[0080] like Figure 3 As shown, this embodiment provides an IoT control device based on natural language, including an intent parsing module 201, a toolset acquisition module 202, and a device control module 203, wherein...

[0081] The intent parsing module 201 is used to receive natural language text from a user and perform intent recognition on the natural language text using a preset large language model to obtain intent parsing data; wherein, the intent parsing data includes intent classification results and entity parameter sets;

[0082] In this embodiment, the intent parsing module 201 performs intent recognition on the natural language text using a preset large language model to obtain intent parsing data. Specifically, the intent parsing module 201 acquires the interaction context information transmitted by the BI-IOT interface, and supplements the natural language text with information based on the interaction context information and a preset instruction template to obtain intermediate natural language text; it maps the intermediate natural language text using a pre-stored terminology library to obtain final natural language text; it performs intent classification on the final natural language text using the preset large language model to obtain intent classification results, and extracts parameters from the final natural language text to obtain an entity parameter set.

[0083] In this embodiment, the intent parsing module 201 performs intent classification on the final natural language text using a preset large language model to obtain intent classification results. Specifically, the intent parsing module 201 extracts semantic features from the final natural language text using a preset natural language processing model to obtain text semantic features; it then fuses the text semantic features using the interaction context information to obtain fused multi-dimensional features; and finally, it inputs the fused multi-dimensional features into a preset classifier so that the classifier outputs intent classification results based on the fused multi-dimensional features.

[0084] In this embodiment, the intent parsing module 201 extracts parameters from the final natural language text to obtain an entity parameter set. Specifically, the intent parsing module 201 extracts parameters from the final natural language text using a preset named entity recognition technology and a preset parameter structure template to obtain an entity parameter set.

[0085] The toolset acquisition module 202 is used to determine whether a tool needs to be invoked based on the intent parsing data. If a tool needs to be invoked, the module selects a target tool from a preset tool list based on the intent parsing data to obtain a target toolset.

[0086] In this embodiment, the toolset acquisition module 202 selects target tools from a preset tool list based on the intent parsing data to obtain a target toolset. Specifically, the toolset acquisition module 202 associates the intent classification results with tools according to a preset intent-tool mapping table to select target tools from the preset tool list and obtain a candidate toolset; and filters the candidate toolset based on the entity parameter set to obtain the target toolset.

[0087] The device control module 203 is used to sort the target tools in the target tool set according to the preset business logic chain and the intent parsing data, generate a tool call sequence, and call the target tools in the target tool set in sequence according to the tool call sequence to realize Internet of Things control.

[0088] In this embodiment, the device control module 203 sorts the target tools in the target tool set according to the preset business logic chain and the intent parsing data, and generates a tool call sequence. Specifically, the device control module 203 sorts the target tools in the target tool set according to the preset business logic chain and the intent parsing data to obtain the tool call order, and generates inter-tool transmission parameters according to the tool call order; the tool call order and the inter-tool transmission parameters are integrated to obtain the tool call sequence.

[0089] For a more detailed explanation of the working principle and procedures of this embodiment, please refer to the relevant description in Embodiment 1.

[0090] This invention, through an intent parsing module 201, deeply understands user commands based on the user's natural language text using a large language model. The intent classification result clarifies the user's core needs, and the entity parameter set extracts key information, laying the foundation for subsequent tool selection and device control. The toolset acquisition module 202 obtains the target toolset based on the intent parsing data, achieving precise mapping between needs and tools. It also filters target tools from the tool list to ensure that the invoked tools match the user's intent, eliminating redundant operations such as parameter configuration or step-by-step guidance. The internal workflow required to achieve the stated purpose is directly parsed based on the user's intent, simplifying the user experience and reducing operational barriers, improving ease of use and adaptability to complex needs. The device control module 203 generates a tool invocation sequence based on a preset business logic chain and the intent parsing data, ensuring that tool invocation conforms to the actual business process and avoiding control failures due to disordered sequence.

[0091] Example 3:

[0092] This embodiment provides a terminal device, including: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus;

[0093] The memory is used to store at least one executable instruction that causes the processor to perform operations as described in any of the above-described natural language-based Internet of Things control methods.

[0094] Example 4:

[0095] This invention provides a computer-readable storage medium including a stored computer program, wherein the computer program, when running, controls the device or apparatus containing the computer-readable storage medium to perform the natural language-based Internet of Things control method as described above.

[0096] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0097] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A natural language-based Internet of Things (IoT) control method, characterized in that, include: The system receives natural language text from users and performs intent recognition on the natural language text using a pre-defined large language model to obtain intent parsing data; wherein, the intent parsing data includes intent classification results and entity parameter sets; Based on the intent parsing data, determine whether it is necessary to invoke a tool. If it is necessary to invoke a tool, select the target tool from the preset tool list based on the intent parsing data to obtain the target toolset. Based on the preset business logic chain and the intent parsing data, the target tools in the target toolset are sorted to generate a tool call sequence, and the target tools in the target toolset are called sequentially according to the tool call sequence to achieve IoT control.

2. The Internet of Things control method based on natural language as described in claim 1, characterized in that, The process of performing intent recognition on the natural language text using a pre-defined large language model to obtain intent parsing data specifically involves: Obtain the interaction context information transmitted by the BI-IOT interface, and supplement the natural language text with information based on the interaction context information and the preset instruction template to obtain intermediate natural language text; The intermediate natural language text is mapped using a pre-stored terminology database to obtain the final natural language text. The final natural language text is classified by intent using a pre-defined large language model to obtain intent classification results. Parameters are then extracted from the final natural language text to obtain an entity parameter set.

3. The Internet of Things control method based on natural language as described in claim 2, characterized in that, The intention classification of the final natural language text is performed using a pre-defined large language model to obtain the intention classification result, specifically as follows: The semantic features of the final natural language text are extracted using a pre-defined natural language processing model. By using the interactive context information, the semantic features of the text are fused to obtain fused multi-dimensional features; The fused multi-dimensional features are input into a preset classifier so that the classifier outputs the intended classification result based on the fused multi-dimensional features.

4. The Internet of Things control method based on natural language as described in claim 2, characterized in that, The step of extracting parameters from the final natural language text to obtain an entity parameter set is specifically as follows: Using a pre-defined named entity recognition technology and based on a pre-defined parameter structure template, parameters are extracted from the final natural language text to obtain an entity parameter set.

5. The Internet of Things control method based on natural language as described in claim 1, characterized in that, Based on the intent parsing data, target tools are selected from a preset tool list to obtain a target toolset, specifically: Based on a preset intent-tool mapping table, the intent classification results are associated with tools to select target tools from a preset tool list, thereby obtaining a candidate tool set; Based on the entity parameter set, the candidate toolset is filtered to obtain the target toolset.

6. The Internet of Things control method based on natural language as described in claim 1, characterized in that, The step of sorting the target tools in the target toolset based on the preset business logic chain and the intent parsing data to generate a tool call sequence is as follows: Based on the preset business logic chain and the intent parsing data, the target tools in the target toolset are sorted to obtain the tool invocation order, and the inter-tool transmission parameters are generated based on the tool invocation order. The tool invocation order and the parameters passed between the tools are integrated to obtain the tool invocation sequence.

7. The Internet of Things control method based on natural language as described in claim 1, characterized in that, Before sequentially invoking the target tools in the target toolset according to the tool invocation sequence, the process further includes: According to the preset tool call sequence template, the tool call sequence is subjected to an integrity check to obtain an integrity check result. If the integrity check result indicates that a parameter is missing, the corresponding information request data is generated and sent to the user.

8. A natural language-based Internet of Things (IoT) control device, characterized in that, It includes an intent parsing module, a toolset acquisition module, and a device control module, among which, The intent parsing module is used to receive natural language text from users and perform intent recognition on the natural language text using a preset large language model to obtain intent parsing data; wherein, the intent parsing data includes intent classification results and entity parameter sets; The toolset acquisition module is used to determine whether a tool needs to be invoked based on the intent parsing data. If a tool needs to be invoked, the module selects a target tool from a preset tool list based on the intent parsing data to obtain the target toolset. The device control module is used to sort the target tools in the target toolset according to the preset business logic chain and the intent parsing data, generate a tool call sequence, and call the target tools in the target toolset in the target toolset in sequence according to the tool call sequence to realize IoT control.

9. A terminal device, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction that causes the processor to perform the operation of the natural language-based Internet of Things control method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device or apparatus containing the computer-readable storage medium to perform the natural language-based Internet of Things control method as described in any one of claims 1 to 7.

Images