Natural language multichannel control method and system based on instruction classification

Abstract

The invention discloses a natural language multichannel control method and system based on instruction classification, and belongs to the technical field of artificial intelligence. The multi-channel control method comprises the following steps: S1, inputting a control instruction; s2, performing instruction classification in combination with a large language model, and identifying a control instruction channel; s3, analyzing and identifying the control instruction according to the type of the control instruction channel; s4, outputting a control instruction command corresponding to each control channel, and executing an operation instruction; and S5, if the control instruction is continuously input, repeatedly executing the steps S2 to S4. According to the method, the operation instructions are classified by utilizing a natural language processing technology, so that the cooperative control of multi-channel equipment is realized, the problems of low operation efficiency, poor environmental adaptability and the like of a traditional control mode are solved, and the multi-task processing capability of the system is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of large language model prompting engineering, and particularly relates to a natural language multi-channel control method and system based on instruction classification. BACKGROUND

[0002] Natural language processing is an important branch of artificial intelligence, aiming to enable computers to understand, interpret and generate human language. With the improvement of computing power and the breakthrough of deep learning technology, natural language processing technology has made significant progress in recent years. Based on the semantic understanding and reasoning ability of natural language, machines can identify potential instructions, tasks and intentions from sentences. Today, natural language processing is not limited to simple instruction execution, but can also deeply understand user context information and complex context to provide more intelligent and personalized services. These technological advances provide strong support for the development of various intelligent devices and platforms, enabling human-computer interaction to gradually develop in a more natural and convenient direction.

[0003] With the widespread application of Internet of Things, smart home, intelligent transportation, robots and other technologies, multi-channel control platforms have become a core component of modern intelligent devices and systems, and users need to input instructions to control different devices or perform different tasks. In order to adapt to complex interaction requirements, modern multi-channel control platforms can provide more flexible and efficient control experience. However, how to intelligently identify user intentions and perform reasonable instruction classification and scheduling, and efficiently integrate multiple input and output channels, remains a difficult point in technology research and development. Therefore, how to optimize the interaction process of multi-channel control platforms through intelligent algorithms to achieve seamless switching and efficient instruction execution has become an urgent need in the development of intelligent systems. SUMMARY

[0004] The main content of the application is a natural language multi-channel control method and system based on instruction classification. The application aims to accurately identify and reasonably distribute different types of control tasks through intelligent instruction classification technology, ensuring coordination and seamless operation between multiple channels, thereby significantly improving the response speed and operation accuracy of the system to meet the control requirements in complex environments.

[0005] A natural language multi-channel control method based on instruction classification, comprising the following steps: S1, obtaining original control instruction text information input by a user; S2, combining a large language model to perform instruction classification and identify control instruction channels; S3, performing control instruction analysis and identification according to the type of control instruction channel; S4, outputting control instruction commands corresponding to each channel and executing operation instructions; S5, if the input control instruction is continued, repeating S2-S4.

[0006] Optionally, the original control instruction input by the user in step S1 includes multiple modes, including voice input and text input.

[0007] Optionally, the control instruction channel in step S2 includes multiple preset types, including some or all of the unmanned aerial vehicle control channel, the interface control channel, the intelligent question and answer channel, and other channels.

[0008] Optionally, step S2 specifically includes the following sub-steps: S21, receiving original control instruction text information; S22, generating prompt words of the large language model according to the preset control instruction channel type and the corresponding control instruction semantic features; the prompt words include the roles and task settings of the interactive parties, the classification range definition of the control instruction channel, the preset control instruction semantic description under each control instruction channel type, and part or all of the control logic, exception handling method, and structured output requirement; S23, concatenating the obtained prompt words and question text into a piece of text and inputting it to the large language model; S24, obtaining the text output by the large language model, and determining the control instruction channel type according to the output text.

[0009] Optionally, the prompt words of the large language model in sub-step S22 can be supplemented or deleted according to the control instruction channel category and the instruction category.

[0010] Optionally, step S3 specifically includes the following sub-steps: S31, obtaining the control instruction channel type and the original control instruction text; S32, selecting the corresponding large language model prompt words according to the control instruction channel type, the prompt words including the control intent analysis logic and the structured output format definition of the corresponding control instruction channel; S33, concatenating the obtained prompt words and question text into a piece of text and continuing to input it to the large language model; S34, obtaining the structured text output by the large language model, the text including the corresponding control instruction parameters under different channels in the predetermined format.

[0011] Optionally, when selecting the corresponding large language model prompt words according to the instruction channel category in sub-step S32, the prompt words under different operation instruction channels are as follows: the prompt words under the unmanned aerial vehicle operation channel include semantic descriptions and output format definitions of unmanned aerial vehicle motion control parameters; the prompt words under the interface operation channel include semantic descriptions and output format definitions of interface operation input methods and corresponding instructions; and the prompt words under the intelligent question and answer channel include intelligent assistant question and answer interaction logic and response format definitions.

[0012] In addition, the present application also provides a natural language multi-channel operation system based on instruction classification, which comprises a microprocessor and a memory connected to each other, and the microprocessor is programmed or configured to execute the steps of the natural language multi-channel operation method based on instruction classification.

[0013] In addition, the present application also provides a computer readable storage medium, which stores a computer program / instruction programmed or configured to execute the steps of the natural language multi-channel operation method based on instruction classification by a processor.

[0014] In addition, the present application also provides a computer program product comprising a computer program / instruction programmed or configured to execute the steps of the natural language multi-channel operation method based on instruction classification by a processor.

[0015] Compared with the prior art, the beneficial effects of the technical scheme of the present application are as follows: I. The existing multi-channel operation system usually requires the user to manually switch different output operation channels, which is relatively cumbersome and not intuitive. The present application can automatically classify the input operation instructions through natural language understanding, allowing the user to automatically select the operation method according to the input instructions. The intelligent instruction classification and channel switching improve the operation flexibility of the system and optimize the user's interaction experience.

[0016] II. The scheme of the present application has stronger universality. In different application fields, the multi-channel operation platform system can flexibly select the type of output channel and operation instruction to meet the operation requirements in different scenarios. Whether it is device control in a home environment or fine operation in a complex industrial environment, the system can provide more accurate and efficient operation experience. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of the drawings shown.

[0018] Figure 1 This is a basic flowchart of the natural language multi-channel control method based on instruction classification described in this invention; Figure 2 This is a flowchart of the voice control command processing in this embodiment; Figure 3 This is the channel output for inputting voice commands to control the drone in this embodiment; Figure 4 This is the channel output for inputting question-and-answer voice commands in this embodiment. Detailed Implementation

[0019] The technical solutions in this embodiment will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are not all embodiments of the present invention. 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.

[0020] This invention designs a natural language multi-channel control method and system based on instruction classification, such as... Figure 1 As shown, the steps and technical principles are as follows: Step S1: Obtain the original control command text information input by the user; The input method for the control commands described in step S1 can include multiple channels, including but not limited to voice input and text input. In this embodiment, the user commands are input via voice, and then the voice control commands are converted into text information using a speech-to-text module. The voice control command processing flow is as follows: Figure 2 As shown.

[0021] Step S2: Combine the large language model to classify instructions and identify the control instruction channels.

[0022] In this embodiment, the selected output channel includes multiple preset types, including drone control channel, interface control channel, intelligent question and answer channel and some or all of other channels.

[0023] Step S2 includes the following sub-steps: S21, Receive the original control command text information; S22, Based on the preset control command channel type and its corresponding control command semantic features, generate prompt words for the large language model; the prompt words include the roles and tasks of the two parties in the interaction, the classification range definition of the control command channel, the preset control command semantic description under each control command channel type, and part or all of the control logic, exception handling method, and structured output requirements. Among them, control logic refers to the rule system that governs how a system converts natural language commands into executable operations within a specific control channel (such as user interface control or drone control). It includes: ① Semantic mapping rules, which map user commands (such as "scroll wheel up") to predefined operation identifiers (such as "roller_up"); ② Operation execution rules, which define the specific device behavior corresponding to the identifier (such as triggering the mouse wheel to scroll up); and ③ Parameter constraint rules, which specify the parameter formats required for the operation (such as speed values ​​and direction enumeration values).

[0024] In this embodiment, the prompt words of the generated large language model are: "I am the commander, and you are my assistant. What you need to do now is to determine, based on what I say, whether I want to control a drone, control the interface, or have a conversation with you. The operations I may need to perform include drone control, interface control, intelligent question answering, etc. Typically, when I want to control the drone, the language may include stop flying, take off, fly up, fly down, fly faster, fly slower, etc.; when I want to control the interface, the language may include left mouse click, right mouse click, scroll wheel up, scroll wheel down, etc.; otherwise, I want to have a conversation with you, in which case intelligent response will be provided; of course, there may also be an input error, resulting in incomprehensible words. When encountering this situation where it is impossible to determine, reply with unknown or other; please analyze what I want to do, and then based on your analysis, you need to select one of the following items to return: drone control, interface control, intelligent question answering, unknown or other. Here is what I say:" In the above prompts, "I am the commander, and you are my assistant" defines the roles of the two interacting parties; "What you need to do now is to determine, based on what I say, whether I want to control a drone, control the interface, or have a conversation with you" defines the task; "The controls I might need to perform include drone control, interface control, and intelligent question answering" defines the scope of the control command channels; "Typically, when I want to control the drone, the language might include stop, take off, fly up, fly down, fly faster, fly slower, etc.; when I want to control the interface, the language might include..." This includes left-clicking, right-clicking, scrolling up, scrolling down, etc.; otherwise, I just want to have a conversation with you, in which case I will provide an intelligent response. This is the preset semantic description and control logic of the control commands for each type of control command channel. "Of course, it is also possible that there is an input error, resulting in an incomprehensible message. When encountering this situation and being unable to determine the cause, reply with 'unknown' or 'other'." This is the exception handling method. "Please analyze what I want to do, and then based on your analysis, you need to choose one of the following options to return: drone control, interface control, intelligent question and answer, unknown, or other." This is the structured output requirement.

[0025] S23, the obtained prompt words and question text are concatenated into a single text and input into the large language model; S24: Obtain the text output by the large language model, and determine the instruction channel category based on the output text.

[0026] Step S3 involves parsing and recognizing control commands based on the control command channel type, including the following sub-steps: S31, Obtain the instruction channel type and original control instruction text; S32, Select the corresponding large language model prompt word according to the control command channel type. The prompt word includes the control intention parsing logic and structured output format definition of the corresponding control command channel. In sub-step S32, when selecting the corresponding large language model prompt words according to the command channel category, the prompt words under different control command channels are as follows: the prompt words under the UAV control channel include the semantic description and output format definition of the UAV motion control parameters; the prompt words under the interface control channel include the semantic description and output format definition of the interface operation input method and the corresponding command; and the prompt words under the intelligent question-and-answer channel include the question-and-answer interaction logic and response format definition of the intelligent assistant.

[0027] In this embodiment, the corresponding large language model prompt word selected according to the instruction channel category in sub-step S32 is: (1) The prompt for the UAV control command is: "I am controlling a UAV to fly. Please analyze my intention based on what I say and then get the command output I want. The following is the output format of the UAV control command: {"on_moving" : false, "moving_direction" : "ahead", "moving_speed" : 0, "on_lifting" : false, "lifting_direction" : "up", "lifting_speed" : 0}. "on_moving" indicates whether the aircraft is moving; "moving_direction" can be selected from "ahead", "back", "left", "right", and "keep", which respectively represent "forward", "backward", "left", "right", and "stationary"; "on_lifting" indicates whether the aircraft is ascending or descending; "lifting_direction" can be selected from "up", "down", and "keep", which respectively represent "ascending", "descending", and "stationary". The following is what I say"; (2) The interface control instruction prompt is: "I am using a mouse or keyboard to operate an interface. Please analyze my interface operation intention based on what I say, and then get the instruction output I want. The following is the interface control instruction output format: {"operation_type" : mouse, "input_command" : "left_press"}. "operation_type" is used to represent the interface control input method, which can be selected from "mouse" and "keyboard", representing mouse input and keyboard input respectively; "input_command" represents the input instruction. When it is recognized as mouse input, "input_command" can be selected from "left_press", "right_press", "roller_up", and "roller_down", representing left click, right click, scroll wheel up, and scroll wheel down respectively. When it is recognized as keyboard input, "input_command" can be selected from numbers and letters. The following is what I say"; (3) The prompt for the intelligent question-and-answer instruction is "You are my intelligent assistant. Please answer the questions I ask. Here are the questions I ask."

[0028] S33, the obtained prompt words and question text are concatenated into a single text, which is then input into the large language model; S34, obtain the structured text output by the large language model, which contains control command parameters corresponding to different channels and conforming to a predetermined format.

[0029] Step S4: Output the control commands corresponding to each channel and execute the operation commands; Step S5: If you continue to input control commands, repeat steps S2 to S4.

[0030] In this embodiment, the voice control command "drone fly upward" is input. The natural language processing model classifies and recognizes this voice command as "drone control," and outputs the control command as {"on_moving" : false, "moving_direction" : "keep", "on_lifting" : true, "lifting_direction" : "up"}. A schematic diagram of the terminal console is shown below. Figure 3 As shown.

[0031] In this embodiment, the user continues to input the voice command "How to make scrambled eggs with tomatoes". The natural language processing model classifies and recognizes this voice command as "intelligent question and answer", and the response information is as follows: Figure 4 As shown.

[0032] The present invention further provides a natural language multi-channel control system based on instruction classification, including a microprocessor and a memory interconnected thereto, wherein the microprocessor is programmed or configured to execute the steps of the natural language multi-channel control method based on instruction classification.

[0033] The present invention further provides a computer-readable storage medium storing a computer program / instructions that are programmed or configured to execute steps of a natural language multichannel manipulation method based on instruction classification by a processor.

[0034] The present invention further provides a computer program product, including a computer program / instructions that are programmed or configured to perform steps of a natural language multichannel manipulation method based on instruction classification via a processor.

[0035] The system and medium of the present invention, corresponding to the methods described above, also have the advantages described above.

[0036] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. All equivalent designs made under the inventive concept of the present invention using the contents of the present invention specification and drawings, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A natural language multi-channel control method based on instruction classification, characterized in that... Includes the following steps: S1, Obtain the original control command text information input by the user; S2, combined with a large language model, performs instruction classification and identifies the control instruction channel; S3, based on the control command channel type, performs control command parsing and recognition; S4 outputs the corresponding control commands for each channel and executes the operation commands. S5. If you continue to input control commands, S2 to S4 will be executed repeatedly.

2. The natural language multi-channel control method based on instruction classification according to claim 1, characterized in that, The original control commands input by the user in step S1 include multiple methods, including voice input and text input.

3. The natural language multi-channel control method based on instruction classification according to claim 1, characterized in that, The control command channel in step S2 includes multiple preset types, including drone control channel, interface control channel, intelligent question and answer channel and some or all of other channels.

4. The natural language multi-channel control method based on instruction classification according to claim 1, characterized in that, Step S2 specifically includes the following sub-steps: S21, Receive the original control command text information; S22, Based on the preset control command channel type and its corresponding control command semantic features, generate prompt words for the large language model; the prompt words include the roles and tasks of the two parties in the interaction, the classification range definition of the control command channel, the preset control command semantic description under each control command channel type, and part or all of the control logic, exception handling method, and structured output requirements. S23, the obtained prompt words and question text are concatenated into a single text and input into the large language model; S24: Obtain the text output by the large language model, and determine the type of control command channel based on the output text.

5. The natural language multi-channel control method based on instruction classification according to claim 4, characterized in that, The prompt words of the large language model mentioned in sub-step S22 can be supplemented or deleted according to the control command channel category and command type.

6. The natural language multi-channel control method based on instruction classification according to claim 3, characterized in that, Step S3 specifically includes the following sub-steps: S31, Obtain the control command channel type and the original control command text; S32, Select the corresponding large language model prompt word according to the control command channel type. The prompt word includes the control intention parsing logic and structured output format definition of the corresponding control command channel. S33, the obtained prompt words and question text are concatenated into a single text, which is then input into the large language model; S34, obtain the structured text output by the large language model, which contains control command parameters corresponding to different channels and conforming to a predetermined format.

7. A natural language multi-channel control method based on instruction classification according to claim 6, characterized in that, In sub-step S32, when selecting the corresponding large language model prompt word according to the command channel category, the prompt words under different control command channels are as follows: the prompt words under the UAV control channel include the semantic description of the UAV motion control parameters and the definition of the output format; The prompts under the interface control channel include semantic descriptions and output format definitions of the interface operation input methods and corresponding commands; the prompts under the intelligent question-and-answer channel include the intelligent assistant's question-and-answer interaction logic and response format definitions.

8. A natural language multi-channel control system based on instruction classification, comprising interconnected microprocessors and memory, characterized in that, The microprocessor is programmed or configured to perform the steps of the instruction classification-based natural language multichannel manipulation method according to any one of claims 1 to 7.

9. A computer-readable storage medium storing a computer program / instructions, characterized in that, The computer program / instructions are programmed or configured to execute the steps of the instruction-based natural language multichannel manipulation method according to any one of claims 1 to 7 via a processor.

10. A computer program product comprising a computer program / instructions, characterized in that, The computer program / instructions are programmed or configured to execute the steps of the instruction-based natural language multichannel manipulation method according to any one of claims 1 to 7 via a processor.

Images