A Dual-Channel Intent Parsing and Function Direct Access System and Method Based on a Large Language Model

By using a dual-channel intent parsing system based on a large language model, the problem of the disconnect between question-and-answer results and system navigation in intelligent question-answering systems has been solved, enabling direct page jumps and optimized interaction paths for natural language requests.

CN123068547APending Publication Date: 2026-11-10SUNSHINE PROPERTY & CASUALTY INSURANCE CO
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611232435.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-14
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

In existing intelligent question-answering systems, the question-answering results and system navigation are disconnected, and functional requests cannot be directly triggered to jump to the next page. Users need to manually find the function entry point, resulting in a long interaction chain.

Method used

The system employs a dual-channel intent parsing and function access system based on a large language model. Through a user request input module, a dual-channel intent parsing module, a knowledge base question answering module, a function menu mapping module, and a dialogue state management module, it processes information inquiry-type and function execution-type requests respectively, thereby achieving the classification and recognition of natural language requests and direct page jumps.

Benefits of technology

It enables differentiated responses to two types of requests: information consultation and function execution. It improves the processing speed and user experience of function execution requests and shortens the interaction path.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123068547A_ABST
    Figure CN123068547A_ABST
Patent Text Reader

Abstract

This invention relates to the fields of natural language processing and artificial intelligence, specifically disclosing a dual-channel intent parsing and function direct access system and method based on a large language model. The technical solution of this invention constructs a dual-channel intent parsing architecture, uses a large language model to classify and identify natural language requests as information-seeking or function-execution-oriented, and triggers knowledge base question answering or function menu mapping and page navigation for different intent types. Simultaneously, it maintains the multi-turn dialogue context through dialogue state management, solving the problems in existing intelligent question-answering systems where question-answer results and system navigation are disconnected, function-execution requests cannot directly trigger page navigation, and users need to manually search for function entry points, resulting in a long interaction path. This achieves differentiated responses for information-seeking and function-execution requests, improves the processing speed and user experience of function-execution requests, and shortens the interaction path from request expression to business operation completion.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of natural language processing and artificial intelligence, and in particular to a dual-channel intent parsing and function access system and method based on a large language model. Background Technology

[0002] Two main technological approaches have emerged in the field of intelligent question answering. One approach constructs a domain knowledge graph and uses semantic retrieval to extract entities and relationships from the knowledge base to provide question-and-answer services. The other approach leverages large language models for semantic understanding and generates natural language responses through prompt word engineering. Additionally, some systems trigger function navigation by setting fixed instruction templates or preset shortcuts, but this method requires users to memorize and input specific instructions, failing to adapt to the flexibility of natural language expression.

[0003] However, the above solution has significant shortcomings. The question-and-answer results are disconnected from the system navigation; the system can only return textual answers and cannot directly jump to function pages, requiring users to manually search for entry points, resulting in a lengthy interaction path. Furthermore, intent recognition is only used for question classification and fails to establish a mapping relationship with function pages, failing to trigger navigation actions for function-driven requests, thus disconnecting the question-and-answer service from business operations.

[0004] Therefore, there is an urgent need to provide a technical solution to address the above problems. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a dual-channel intent parsing and function access system and method based on a large language model.

[0006] In a first aspect, the present invention provides a dual-channel intent parsing and function access system based on a large language model, the technical solution of which is as follows: The user request input module is used to receive natural language requests input by users and preprocess the natural language requests to obtain preprocessed request text. The dual-channel intent parsing module is used to guide the large language model to perform intent parsing on the preprocessed request text through prompt word templates, and output the intent type corresponding to the natural language request, wherein the intent type is information consultation type or function execution type. The knowledge base question answering module is used to retrieve matching content from a preset knowledge base based on the preprocessed request text when the intent type is information consultation, input the matching content into the large language model, and obtain the response text generated by the large language model. The function menu mapping module is used to extract function action keywords and business object keywords from the preprocessed request text when the intent type is function execution type, and match the corresponding function menu node in the preset function menu mapping relationship table according to the function action keywords and the business object keywords. The function recommendation and redirection module is used to generate function entry recommendation information based on the function menu node, and in response to the user's trigger operation on the function entry recommendation information, redirect to the function page corresponding to the function menu node; The dialogue state management module is used to maintain the context state of multi-turn dialogues, record the intent type in the context state, and input the context state into the dual-channel intent parsing module at the start of the next round of dialogue.

[0007] The beneficial effects of the dual-channel intent parsing and function direct access system based on a large language model of the present invention are as follows: The system of this invention constructs a dual-channel intent parsing architecture, uses a large language model to classify and identify natural language requests as information-seeking or function-execution-oriented, and triggers knowledge base question-and-answer or function menu mapping and page jumps for different intent types. At the same time, it maintains the multi-turn dialogue context by using dialogue state management. This solves the problems in existing intelligent question-answering systems, such as the disconnect between question-and-answer results and system navigation, the inability to directly trigger page jumps for function-execution requests, and the long interaction links caused by users having to manually search for function entry points. It realizes differentiated responses for information-seeking and function-execution requests, improves the processing speed and user experience of function-execution requests, and shortens the interaction path from request expression to business operation completion.

[0008] Based on the above scheme, the dual-channel intent parsing and function access system based on a large language model of the present invention can be further improved as follows.

[0009] In one alternative approach, the user request input module is specifically used for: The system receives the natural language request in speech form and converts the speech-to-text request into the natural language request in text form using a built-in speech-to-text unit.

[0010] The advantages of adopting the above optional methods are: further support for natural language input in the form of voice, converting speech into text through the speech-to-text unit, expanding the user input methods, making it more convenient to express needs in mobile scenarios, and improving the flexibility of interaction.

[0011] In one alternative approach, the user request input module is specifically used to: perform word segmentation, entity recognition, and intent classification on the natural language request in text form to obtain the preprocessed request text.

[0012] The advantages of adopting the above-mentioned optional methods are: further performing word segmentation, entity recognition and intent classification on the natural language request in text form, extracting key information from the lexical and semantic levels, making the structure of the preprocessed request text clearer, and providing a standardized input basis for subsequent intent parsing.

[0013] In one alternative approach, the dual-channel intent parsing module incorporates the prompt word template, which includes intent category definition information and parsing rule information. The intent category definition information is used to define two intent categories: information consultation and function execution. The parsing rule information is used to guide the large language model to output the corresponding intent type based on the preprocessed request text.

[0014] The beneficial effects of adopting the above optional approach are as follows: by further embedding prompt word templates in the dual-channel intent parsing module, the large language model is guided to classify intents through intent category definitions and parsing rules, making the distinction between information consultation and function execution more explicit and improving the accuracy of intent parsing.

[0015] In one alternative approach, the preset knowledge base includes company introduction knowledge data, product consultation knowledge data, and claims consultation knowledge data, and the matching content originates from at least one of the company introduction knowledge data, the product consultation knowledge data, and the claims consultation knowledge data.

[0016] The beneficial effects of adopting the above optional methods are: further limiting the data scope of the preset knowledge base to cover three types of knowledge data: company introduction, product consultation and claims consultation, making the source of the response content for information consultation requests more focused, and the knowledge retrieval scope highly matched with the insurance business scenario.

[0017] In an alternative approach, the function menu mapping module is specifically used for: The functional action keywords and the business object keywords are compared with the keyword fields in the preset functional menu mapping table, and the functional menu nodes corresponding to the matching keyword fields are output as the matching results.

[0018] The beneficial effects of adopting the above optional method are as follows: further comparing the functional action keywords and business object keywords with the keyword fields in the preset functional menu mapping table, and locating the functional menu node through dual keyword matching, so as to make the mapping relationship between functional execution requests and functional entry more accurate.

[0019] In one alternative approach, the function recommendation and redirection module is specifically used for: Generate function entry recommendation information in the form of function cards or function buttons, wherein the function entry recommendation information includes the function name and jump link address corresponding to the function menu node; In response to a user's click on the function card or the function button, the user is redirected to the function page corresponding to the function menu node based on the redirect link address.

[0020] The advantages of adopting the above optional methods are: further displaying function entry recommendation information in the form of function cards or function buttons, responding to user click operations to complete page jump, directly converting natural language requests into visual interactive components, and shortening the path from request recognition to business page loading.

[0021] Secondly, this invention provides a dual-channel intent parsing and function access method based on a large language model, employing the dual-channel intent parsing and function access system based on a large language model provided by this invention. The technical solution of this method is as follows: Receive a natural language request input by the user, and preprocess the natural language request to obtain a preprocessed request text; The prompt word template guides the large language model to perform intent parsing on the preprocessed request text and outputs the intent type corresponding to the natural language request, which is either information consultation or function execution. When the intent type is information consultation, matching content is retrieved from the preset knowledge base based on the preprocessed request text, the matching content is input into the large language model, and the response text generated by the large language model is obtained. When the intent type is function execution type, extract function action keywords and business object keywords from the preprocessed request text, and match the corresponding function menu node in the preset function menu mapping relationship table according to the function action keywords and the business object keywords. Based on the function menu node, function entry recommendation information is generated, and in response to the user's trigger operation on the function entry recommendation information, the user is redirected to the function page corresponding to the function menu node.

[0022] The beneficial effects of the dual-channel intent parsing and function access method based on a large language model of the present invention are as follows: The method of this invention constructs a dual-channel intent parsing architecture, uses a large language model to classify and identify natural language requests as information-seeking or function-execution-oriented, and triggers knowledge base question-and-answer or function menu mapping and page jumps for different intent types. At the same time, it uses dialogue state management to maintain the multi-turn dialogue context, which solves the problems of existing intelligent question-answering systems where question-and-answer results and system navigation are disconnected, function-execution requests cannot directly trigger page jumps, and users need to manually search for function entry points, resulting in a long interaction chain. It realizes differentiated responses for information-seeking and function-execution requests, improves the processing speed and user experience of function-execution requests, and shortens the interaction path from request expression to business operation completion.

[0023] Thirdly, the technical solution of an electronic device according to the present invention is as follows: It includes a memory, a processor, and a program stored in the memory and running on the processor, wherein the processor executes the program to implement the steps of the dual-channel intent parsing and function access method based on a large language model as described in this invention.

[0024] Fourthly, the technical solution of a computer-readable storage medium provided by the present invention is as follows: The computer-readable storage medium stores instructions that, when read, cause the computer-readable storage medium to perform the steps of the dual-channel intent parsing and function access method based on a large language model as described in this invention.

[0025] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0026] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a schematic diagram of an embodiment of a dual-channel intent parsing and function direct access system based on a large language model according to the present invention. Figure 2 This is a schematic diagram of the specific processing flow; Figure 3 This is a flowchart illustrating an embodiment of a dual-channel intent parsing and function access method based on a large language model according to the present invention. Figure 4 This is a schematic diagram of an embodiment of an electronic device according to the present invention. Detailed Implementation

[0027] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0028] Figure 1 This diagram illustrates the structure of an embodiment of a dual-channel intent parsing and function access system 100 based on a large language model provided by the present invention. Figure 1 As shown, the dual-channel intent parsing and function access system 100 based on a large language model includes: User request input module 101 is used to receive natural language requests input by users and preprocess the natural language requests to obtain preprocessed request text.

[0029] Natural language requests refer to user requests to the system, expressed through natural language, for information or to perform functions. For example, an insurance salesperson inputting the sentence "Help me record a form" is a natural language request. Preprocessed request text refers to the text data obtained after processing the user's natural language request through word segmentation, entity recognition, and intent classification. For example, preprocessing "Help me record a form" yields a preprocessed request text containing "record" as an action word and "form" as an object word.

[0030] Specifically, users input their requests in natural language, either as text or voice. In the case of voice input, the natural language request is converted from spoken to written form using speech-to-text conversion. Preprocessing is then performed on the natural language request, including word segmentation, entity recognition, and intent classification, resulting in a preprocessed request text.

[0031] The dual-channel intent parsing module 102 is used to guide the large language model to perform intent parsing on the preprocessed request text through prompt word templates, and output the intent type corresponding to the natural language request, wherein the intent type is information consultation type or function execution type.

[0032] The prompt word template refers to a pre-written text template used to guide the large language model to reason and output according to a specified format and logic. For example, a prompt word template containing guiding statements such as "Please classify the user's request as information consultation or function execution" is pre-written and input into the large language model along with the pre-processed request text when it is invoked. The intent type refers to the classification label output by the large language model after parsing the pre-processed request text, used to identify the category to which the user's request belongs. For example, if "Help me record a form" is parsed as function execution, the classification label is the intent type.

[0033] Among them, "information consultation type" refers to the user's intention to obtain knowledge-based answers through the system; for example, when a user asks "What materials are needed for claims?", the intention type of this request is determined to be information consultation type. "Function execution type" refers to the user's intention to directly complete a certain business operation through the system; for example, when a user expresses "Help me enter a form", the intention type of this request is determined to be function execution type.

[0034] Specifically, the preprocessed request text and prompt word template are input into the large language model. The prompt word template contains intent category definition information and parsing rule information. The intent category definition information defines two intent categories: information consultation and function execution. The parsing rule information guides the large language model to perform intent parsing on the preprocessed request text according to the specified judgment logic. The large language model outputs the corresponding intent type based on the content of the preprocessed request text, which is either information consultation or function execution.

[0035] The knowledge base question answering module 103 is used to retrieve matching content from a preset knowledge base based on the preprocessed request text when the intent type is information consultation, input the matching content into the large language model, and obtain the response text generated by the large language model.

[0036] The pre-built knowledge base refers to a set of knowledge pre-constructed and stored for providing information consultation and Q&A services. For example, the pre-built knowledge base includes information such as the coverage of insurance products and claims procedures; when a user inquires about product information, relevant content is retrieved from this knowledge base. Matching content refers to knowledge entries related to the user's request retrieved from the pre-built knowledge base based on the pre-processed request text. For example, retrieving a list of claim materials from the pre-built knowledge base based on the question "What materials are needed for a claim?" constitutes matching content.

[0037] The response text refers to the natural language response text generated by the large language model after the matched content is input into it. For example, after inputting the retrieved list of claim materials into the large language model, the large language model generates a natural language response stating "The following materials are required for claiming: claim application form, identity certificate, copy of the policy, and original medical expense invoices." This response is the response text.

[0038] Specifically, when the intent type is information consultation, the system searches a pre-processed knowledge base based on the pre-processed request text. The pre-processed knowledge base contains company introduction knowledge data, product consultation knowledge data, and claims consultation knowledge data. The system retrieves knowledge entries from the pre-processed knowledge base that match the pre-processed request text as matching content. The matching content comes from at least one of the company introduction knowledge data, product consultation knowledge data, and claims consultation knowledge data. The matching content is then input into the large language model, which generates and outputs a response text in natural language based on the matching content.

[0039] The function menu mapping module 104 is used to extract function action keywords and business object keywords from the preprocessed request text when the intent type is function execution type, and match the corresponding function menu node in the preset function menu mapping relationship table according to the function action keywords and the business object keywords.

[0040] Among them, functional action keywords refer to words extracted from the preprocessed request text that characterize the type of action the user wishes to perform; for example, "record" is extracted as a functional action keyword from the preprocessed result of "help me record a form". Business object keywords refer to words extracted from the preprocessed request text that characterize the specific business object or business data the user wishes to operate on; for example, "form" is extracted as a business object keyword from the preprocessed result of "help me record a form".

[0041] The preset function menu mapping table refers to a pre-established data table used to record the mapping relationships between function action keywords, business object keywords, and function menu nodes. For example, the preset function menu mapping table records the mapping relationship between the keyword combinations "record" and "order" and the "policy entry" function node. A function menu node refers to the specific function entry location in the function menu tree corresponding to the user's function execution request. For example, the "policy entry" function entry obtained by matching "record" and "order" in the preset function menu mapping table is the function menu node.

[0042] Specifically, when the intent type is function execution type, function action keywords and business object keywords are extracted from the preprocessed request text. Function action keywords represent the type of action the user wants to perform, and business object keywords represent the specific business object the user wants to operate. The function action keywords and business object keywords are compared with the keyword fields in the preset function menu mapping table, and the function menu nodes corresponding to the matching keyword fields are output as matching results.

[0043] The function recommendation and redirection module 105 is used to generate function entry recommendation information based on the function menu node, and in response to the user's trigger operation on the function entry recommendation information, redirect to the function page corresponding to the function menu node.

[0044] Among them, the function entry recommendation information refers to the visual information generated based on the matched function menu node to display the function entry to the user; for example, a function card containing the name "policy entry" and the corresponding jump link address is generated based on the "policy entry" function menu node, and this card is the function entry recommendation information.

[0045] Here, "triggering operation" refers to the user's interactive action on the recommended information for the function entry point; for example, when a user clicks the button on the displayed "Policy Entry" function card, this click action is a triggering operation. "Function page" refers to the operation interface that the user is redirected to after triggering the recommended information for the function entry point, corresponding to the function menu node; for example, when a user clicks the "Policy Entry" function card, they are redirected to the operation interface for filling in and submitting policy information, and this interface is the function page.

[0046] Specifically, based on the matched function menu nodes, function entry recommendation information is generated. The function entry recommendation information includes the function name and jump link address corresponding to the function menu node. The function entry recommendation information is displayed to the user in the form of function cards or function buttons. In response to the user's click operation on the function card or function button, the user is redirected to the function page corresponding to the function menu node according to the jump link address.

[0047] The dialogue state management module 106 is used to maintain the context state of multi-turn dialogues, record the intent type in the context state, and input the context state into the dual-channel intent parsing module 102 at the start of the next round of dialogue.

[0048] The context state refers to the data set maintained during a multi-turn dialogue to record historical information and intent types for the current dialogue turn; for example, in the first round of dialogue, the user intent type is recorded as function execution type, and this recorded information is a component of the context state.

[0049] Specifically, the context state is maintained during the multi-turn dialogue process. The context state records historical dialogue information and the intent type obtained from each parsing. At the beginning of the next round of dialogue, the context state along with the new natural language request is input into the dual-channel intent parsing module 102. The dual-channel intent parsing module 102 combines the intent type recorded in the context state to perform intent parsing on the new natural language request and outputs the updated intent type.

[0050] It should be noted that when the intent type recorded in the context state by the dialogue state management module 106 is inconsistent with the updated intent type, the intent type recorded in the context state will be updated to the updated intent type, and the dual-channel intent parsing module 102 will be triggered to respond according to the processing path corresponding to the updated intent type.

[0051] The technical solution of this embodiment constructs a dual-channel intent parsing architecture, uses a large language model to classify and identify natural language requests as information-seeking or function-execution-oriented, and triggers knowledge base question-and-answer or function menu mapping and page jumps for different intent types. At the same time, it uses dialogue state management to maintain the multi-turn dialogue context, which solves the problems of existing intelligent question-and-answer systems where question-and-answer results and system navigation are disconnected, function-execution requests cannot directly trigger page jumps, and users need to manually search for function entry points, resulting in a long interaction chain. It realizes differentiated responses for information-seeking and function-execution requests, improves the processing speed and user experience of function-execution requests, and shortens the interaction path from request expression to business operation completion.

[0052] In an alternative embodiment, the user request input module 101 is specifically used for: The system receives the natural language request in speech form and converts the speech-to-text request into the natural language request in text form using a built-in speech-to-text unit.

[0053] The speech-to-text unit refers to a built-in functional component that converts a user's speech-based natural language request into a text-based natural language request. For example, if a user says "Help me record a form" by voice, the speech-to-text unit will convert the speech into the text "Help me record a form".

[0054] Specifically, when a user inputs a natural language request in the form of speech, the speech-to-text unit receives the speech signal, performs recognition and conversion processing on the speech signal, and outputs the converted natural language request in the form of text. This natural language request in the form of text is used in subsequent preprocessing steps.

[0055] It should be noted that the speech-to-text unit adopts automatic speech recognition technology, performs acoustic feature extraction and language model decoding on the received speech signal, converts the speech signal into a corresponding text sequence, and the text sequence retains the complete semantic content of the natural language demand expressed by the original speech, so as to support the subsequent preprocessing steps to perform word segmentation, entity recognition and intent classification processing on the natural language demand in text form.

[0056] In the above optional manner, the input of natural language demands in speech form is further supported, and the speech is converted into text through the speech-to-text unit, which expands the user input mode, makes the demand expression in mobile scenarios more convenient, and improves the interaction flexibility.

[0057] In an optional manner, the user demand input module 101 is specifically configured to: Perform word segmentation processing, entity recognition processing and intent classification processing on the natural language demand in the text form to obtain the preprocessed demand text.

[0058] The word segmentation processing refers to a processing operation of splitting a continuous natural language demand in text form into a word sequence according to semantic units; for example, performing word segmentation processing on "help me create an order" to obtain a word sequence of "help", "me", "create", "an", "order". The entity recognition processing refers to a processing operation of identifying and extracting entity words with specific semantic categories from the natural language demand in text form; for example, performing entity recognition processing on "help me create an order" to identify the category label of "order" as a business object entity. The intent classification processing refers to a processing operation of determining the category to which the user intent belongs according to the content of the natural language demand in text form; for example, performing intent classification processing on "help me create an order" to determine that the intent category of the demand is a function execution type.

[0059] Specifically, word segmentation processing is performed on the natural language demand in text form to split the continuous text into a word sequence; on the basis of the word segmentation processing, business object entities with specific semantic categories are identified and extracted from the word sequence, and the category labels of the business object entities are marked; according to the content of the natural language demand in text form and the identified business object entities, it is determined whether the user intent category belongs to an information consultation type or a function execution type; after the word segmentation processing, entity recognition processing and intent classification processing, the preprocessed demand text is obtained.

[0060] In the above optional manner, word segmentation, entity recognition and intent classification processing are further performed on the natural language demand in text form to extract key information from the lexical and semantic levels, so that the structure of the preprocessed demand text is clearer, and a standardized input basis is provided for subsequent intent parsing.

[0061] In an optional embodiment, the dual-channel intent parsing module 102 incorporates the prompt word template, which includes intent category definition information and parsing rule information. The intent category definition information is used to define two intent categories: information consultation and function execution. The parsing rule information is used to guide the large language model to output the corresponding intent type based on the preprocessed request text.

[0062] The intent category definition information refers to the descriptive information contained in the prompt word template that defines the names and meanings of each category in the intent type system. For example, the intent category definition information in the prompt word template includes descriptions such as "Information Consultation Type: The user wants to obtain a knowledge-based answer" and "Function Execution Type: The user wants to perform a specific business operation." The parsing rule information refers to the rule description information contained in the prompt word template that guides the large language model to parse intents according to specified judgment logic and output format. For example, the parsing rule information in the prompt word template includes judgment rule descriptions such as "If the user's request includes a specific action, it is determined to be Function Execution Type; if the user's request includes a question word or inquiry content, it is determined to be Information Consultation Type."

[0063] In the above optional methods, a prompt word template is further built into the dual-channel intent parsing module 102. The intent category definition and parsing rules guide the large language model to classify intents, making the distinction between information consultation and function execution more explicit and improving the accuracy of intent parsing.

[0064] In one alternative approach, the preset knowledge base includes company introduction knowledge data, product consultation knowledge data, and claims consultation knowledge data, and the matching content originates from at least one of the company introduction knowledge data, the product consultation knowledge data, and the claims consultation knowledge data.

[0065] The company introduction knowledge data refers to the content data stored in the preset knowledge base related to the company's basic information, development history, organizational structure, and business scope. For example, the company introduction knowledge data includes information such as the year the insurance company was established, its registered capital, and the number of branches nationwide. The product consultation knowledge data refers to the content data stored in the preset knowledge base related to product types, product functions, applicable scenarios, and product advantages. For example, the product consultation knowledge data includes information such as the coverage scope of insurance products, premium calculation methods, and deductible settings. The claims consultation knowledge data refers to the content data stored in the preset knowledge base related to the claims process, claims material lists, claims conditions, and claims timelines. For example, the claims consultation knowledge data includes information such as the claims application form completion guidelines, lists of materials required for different accident scenarios, and the time it takes for claims payments to be credited to the account.

[0066] Among the above optional methods, the data scope of the preset knowledge base is further limited to cover three types of knowledge data: company introduction, product consultation and claims consultation. This makes the source of the response content for information consultation requests more focused, and the scope of knowledge retrieval is highly matched with the insurance business scenario.

[0067] In an alternative embodiment, the function menu mapping module 104 is specifically used for: The functional action keywords and the business object keywords are compared with the keyword fields in the preset functional menu mapping table, and the functional menu nodes corresponding to the matching keyword fields are output as the matching results.

[0068] The keyword field refers to the preset word data column recorded in the preset function menu mapping relationship table, which is used to compare with the function action keywords and business object keywords. For example, there is a column of preset word data in the preset function menu mapping relationship table, which records words such as "record" and "order". This column is the keyword field.

[0069] Specifically, after extracting functional action keywords and business object keywords from the preprocessed request text, the functional action keywords are compared with the keyword fields in the preset function menu mapping table, and the business object keywords are also compared with the keyword fields in the preset function menu mapping table. Based on the comparison results of the functional action keywords and business object keywords respectively, the keyword field that matches both the functional action keywords and business object keywords is searched in the preset function menu mapping table, and the function menu node corresponding to the keyword field is output as the matching result.

[0070] In the above optional methods, the functional action keywords and business object keywords are further compared with the keyword fields in the preset functional menu mapping table. The functional menu nodes are located through dual keyword matching, making the mapping relationship between functional execution requests and functional entry points more accurate.

[0071] In an alternative embodiment, the function recommendation and redirection module 105 is specifically used for: Generate function entry recommendation information in the form of function cards or function buttons. The function entry recommendation information includes the function name and jump link address corresponding to the function menu node.

[0072] The "function card format" refers to a visual display method that presents recommended information about function entry points in a card style. For example, the "Policy Entry" function entry point is displayed in the dialog interface as a rectangular card containing an icon, function name, and operation button. This display method is the function card format. The "function button format" refers to a visual display method that presents recommended information about function entry points in a button style. For example, the "Policy Entry" function entry point is displayed below the response text in the dialog interface as a clickable rectangular control containing the function name. This display method is the function button format.

[0073] Here, "function name" refers to the standard name displayed on the interface for the function page corresponding to the function menu node; for example, the function name corresponding to the "policy entry" function menu node is "policy entry". "Jump link address" refers to the internal access path information used to locate and jump to a specific function page; for example, the jump link address assigned to the "policy entry" function page is the internal routing path " / pages / policy / create".

[0074] Specifically, the function name and jump link address are determined based on the matched function menu node. Function entry recommendation information is generated based on the function name and jump link address. When the function entry recommendation information is displayed in the form of a function card, the function name and jump link address are carried in a card-style visual component. When the function entry recommendation information is displayed in the form of a function button, the function name and jump link address are carried in a button-style visual component.

[0075] In response to a user's click on the function card or the function button, the user is redirected to the function page corresponding to the function menu node based on the redirect link address.

[0076] The click operation refers to the interactive action of selecting and triggering a function card or function button by the user through mouse click, finger touch click, or keyboard confirmation; for example, the user clicks the operation button on the "Policy Entry" function card, and this mouse click action is a click operation.

[0077] Specifically, when a user performs a click operation on a function card or function button, the click operation includes any one of mouse click, finger touch click, or keyboard confirmation. In response to the click operation, the jump link address carried by the function card or function button is obtained, and the function page corresponding to the function menu node is located according to the jump link address, thus completing the jump from the current interface to the function page.

[0078] Among the above optional methods, the recommended information for function entry is further displayed in the form of function cards or function buttons, and the page jump is completed in response to the user's click operation. Natural language requests are directly transformed into visual interactive components, which shortens the path from request recognition to business page loading.

[0079] Figure 2 A schematic diagram illustrating the specific processing flow of this embodiment is shown. For example... Figure 2 As shown, the user inputs a request, which can be either text or voice input. When the user uses voice input, the natural language request in speech-to-text format is converted into a text-based natural language request. After receiving the user's natural language request, request preprocessing is performed, including word segmentation, entity recognition, and intent classification. After preprocessing, the preprocessed request text is obtained.

[0080] After preprocessing, a dual-channel intent parsing process using a large language model is executed. The model is guided by a prompt word template to parse the intent of the preprocessed request text, outputting the intent type corresponding to the natural language request. The intent type is either information inquiry or function execution. When the large language model parses the preprocessed request text, the prompt word template contains intent category definition information and parsing rule information. The intent category definition information defines the two intent categories: information inquiry and function execution. The parsing rule information guides the large language model to output the corresponding intent type based on the preprocessed request text.

[0081] After determining the intent type, different processing paths are initiated based on the intent type. When the intent type is information inquiry, a knowledge base retrieval is performed. Matching content is retrieved from the pre-processed request text using a pre-defined knowledge base. This matching content is then input into a large language model, which generates a response text and returns the answer to the user. The pre-defined knowledge base contains company introduction knowledge data, product consultation knowledge data, and claims consultation knowledge data. The matching content originates from at least one of these three categories.

[0082] When the intent type is function execution, extract function action keywords and business object keywords. Perform function menu mapping matching based on these keywords, matching the corresponding function menu node in a pre-defined function menu mapping table. During the function menu mapping matching process, compare the function action keywords and business object keywords with the keyword fields in the pre-defined function menu mapping table. Output the function menu node corresponding to the matching keyword field as the matching result. After matching, recommend a function entry and execute a one-click redirect. Generate function entry recommendation information based on the function menu node. This information includes the function name and redirect link address corresponding to the function menu node. The recommendation information is displayed to the user in the form of a function card or function button. In response to the user's click on the function card or button, redirect to the function page corresponding to the function menu node based on the redirect link address.

[0083] During the process of handling user requests, multiple rounds of dialogue state updates are performed to maintain the context state of the multi-round dialogue. The intent type is recorded in the context state, and the context state is input into the dual-channel intent parsing module 102 at the start of the next round of dialogue. When the user switches between information inquiry and function operation, the intent change is dynamically identified and the intent type recorded in the context state is updated, triggering the dual-channel intent parsing module 102 to respond according to the processing path corresponding to the updated intent type.

[0084] After responding to the user's request through the above processing path, wait for the next round of user input.

[0085] The technical solution of this embodiment uses a large language model to perform dual-channel intent parsing of users' natural language requests, distinguishing between information inquiry and function execution requests. Information inquiry requests enter the knowledge base question and answer path to obtain the answer text, while function execution requests enter the function menu mapping navigation path to obtain function entry recommendation information and realize one-click jump. At the same time, in multi-turn dialogue, the intent switching is dynamically perceived and the response mode is adjusted through dialogue state management, realizing a complete processing flow from the user's natural language request input to information response or direct function access.

[0086] In another embodiment of the dual-channel intent parsing and function access system 100 based on a large language model of the present invention, it includes: The user request input module 101 is used to receive natural language requests input by insurance business personnel, perform preprocessing on the natural language requests, including word segmentation, entity recognition and intent classification, and obtain the preprocessed request text after preprocessing. The preprocessed request text contains insurance business action words and insurance business object words.

[0087] The dual-channel intent parsing module 102 is used to input the preprocessed request text and prompt word template into the large language model. The prompt word template contains intent category definition information and parsing rule information. The intent category definition information defines two intent categories: information consultation type and function execution type. The parsing rule information guides the large language model to perform intent parsing on the preprocessed request text according to the specified judgment logic. The large language model outputs the corresponding intent type based on the content of the preprocessed request text.

[0088] The knowledge base question-answering module 103 is used to search a preset knowledge base based on the pre-processed request text when the intent type is information consultation. The preset knowledge base contains company introduction knowledge data, product consultation knowledge data, and claims consultation knowledge data. It retrieves knowledge entries from the preset knowledge base that match the pre-processed request text as matching content. The matching content comes from at least one of the company introduction knowledge data, product consultation knowledge data, and claims consultation knowledge data. The matching content is input into the big language model, and the big language model generates and outputs a response text in natural language form based on the matching content.

[0089] The function menu mapping module 104 is used to extract function action keywords and business object keywords from the preprocessed request text when the intent type is function execution type. Function action keywords represent the type of action that the insurance agent wants to perform, and business object keywords represent the specific business object that the insurance agent wants to operate. The function action keywords and business object keywords are compared with the keyword fields in the preset function menu mapping relationship table, and the function menu node corresponding to the matching keyword field is output as the matching result.

[0090] In the process of extracting functional action keywords and business object keywords, the following intent-function matching confidence function is used to calculate the matching confidence between the preprocessed request text and each mapping record in the preset function menu mapping relationship table: In the formula, The table showing the mapping relationship between preprocessed request text and preset function menus is as follows: Match confidence between mapping records; This represents the prediction confidence score when the output intent type of the large language model is functional executive type. These represent functional action keywords extracted from the preprocessed request text. This refers to the keywords related to the business objects extracted from the preprocessed request text. This indicates the first item in the preset function menu mapping table. The mapping records correspond to the standard function action keywords; This indicates the first item in the preset function menu mapping table. The mapping record corresponds to the standard business object keywords; This indicates the concatenation operation of two keyword sequences; This function represents the semantic similarity calculation between two concatenated keyword sequences. The function represents the edit distance between two concatenated keyword sequences; iterates through all mapping records in the preset function menu mapping table and matches the confidence level. The first value corresponding to the maximum value The function menu nodes in the mapping record are output as matching results.

[0091] It should be noted that, in the processing path of functional execution intent, the above formula is based on the intent prediction confidence output by the large language model, and combines the functional action keywords and business object keywords extracted from the preprocessed appeal text. The concatenation of these two keywords is then compared with the standard keyword concatenation sequence of each mapping record in the preset function menu mapping relationship table to calculate semantic similarity and edit distance. Semantic similarity is used to measure the degree of matching between the appeal text and the mapping record at the semantic level, while edit distance is used to measure the degree of difference between the two at the character sequence level. The product of intent prediction confidence and semantic similarity is used as the numerator, and the logarithmically scaled value of the edit distance is used as part of the denominator to form the comprehensive matching confidence.

[0092] The above formula introduces intent prediction confidence to weight and guide the keyword matching process, so that the matching confidence reflects the credibility of the large language model's judgment on functional execution intent as well as the semantic similarity and character difference between the appeal text and the mapping record. This avoids the bias caused by relying on a single matching dimension, thereby improving the accuracy of matching functional menu nodes.

[0093] The function recommendation and redirection module 105 is used to determine the function name and redirection link address corresponding to the insurance business function page based on the matched function menu node. Based on the function name and redirection link address, it generates insurance business function entry recommendation information, which is displayed to insurance business personnel in the form of function cards or function buttons. In response to the insurance business personnel's click operation on the function card or function button, the system redirects to the insurance business function page corresponding to the insurance business function menu node according to the redirection link address.

[0094] The dialogue state management module 106 is used to maintain the context state during multi-turn dialogue. The context state records historical dialogue information and the intent type obtained from each parsing. At the beginning of the next round of dialogue, the context state along with the new natural language request is input into the dual-channel intent parsing module 102. The dual-channel intent parsing module 102 combines the intent type recorded in the context state to perform intent parsing on the new natural language request and outputs the updated intent type.

[0095] The technical solution of this embodiment calculates the comprehensive matching confidence of functional execution requests by introducing a matching function that combines intent prediction confidence with keyword semantic similarity and edit distance. Based on the maximum matching confidence, the function menu node is located and the page jump is completed. At the same time, the context state is maintained and historical intent types are recorded during multi-turn dialogues, so that the recorded intent types can be used to assist in judging the current request in subsequent interactions. This solves the matching deviation problem caused by single-dimensional judgment in the matching process and the problem of delayed response path switching when intent changes in multi-turn interactions. It improves the accuracy of functional menu node matching, reduces the probability of mismatch and missed match caused by the diversification of keyword expressions in functional execution requests by insurance business personnel, shortens the interaction path from request input to function page loading, and improves the response accuracy in intent switching scenarios.

[0096] Figure 3 This diagram illustrates an embodiment of a dual-channel intent parsing and function access method based on a large language model provided by the present invention. This dual-channel intent parsing and function access method based on a large language model employs the dual-channel intent parsing and function access system 100 provided by the present invention. Figure 3 As shown, the method includes the following steps: S1. Receive the natural language request input by the user, and preprocess the natural language request to obtain the preprocessed request text. S2. Guide the large language model to perform intent parsing on the preprocessed request text through prompt word templates, and output the intent type corresponding to the natural language request, wherein the intent type is information consultation type or function execution type. S3. When the intent type is information consultation, retrieve matching content from the preset knowledge base according to the preprocessed request text, input the matching content into the big language model, and obtain the response text generated by the big language model. S4. When the intent type is function execution type, extract function action keywords and business object keywords from the preprocessed request text, and match the corresponding function menu node in the preset function menu mapping relationship table according to the function action keywords and the business object keywords. S5. Generate function entry recommendation information based on the function menu node, and in response to the user's trigger operation on the function entry recommendation information, jump to the function page corresponding to the function menu node.

[0097] It should be noted that the beneficial effects of the dual-channel intent parsing and function access method based on a large language model provided in the above embodiments are the same as those of the dual-channel intent parsing and function access system based on a large language model, and will not be repeated here. Furthermore, the methods and system embodiments provided in the above embodiments belong to the same concept, and their specific implementation processes are detailed in the system embodiments, and will not be repeated here.

[0098] The dual-channel intent parsing and function access system 100 based on a large language model of the present invention can be a computer program (including program code) running on a computer device. For example, the dual-channel intent parsing and function access system 100 based on a large language model of the present invention is an application software that can be used to execute the corresponding steps in the dual-channel intent parsing and function access method based on a large language model of the present invention.

[0099] In some embodiments, the dual-channel intent parsing and function access system 100 based on a large language model of the present invention can be implemented in a combination of hardware and software. As an example, the dual-channel intent parsing and function access system 100 based on a large language model of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the dual-channel intent parsing and function access method based on a large language model of the present invention. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0100] The modules described in the embodiments of this invention can be implemented in software or hardware. The names of the modules are not, in some cases, limiting the scope of the module itself.

[0101] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned dual-channel intent parsing and function access methods based on a large language model. That is, an electronic device according to an embodiment of the present invention may include, but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute the dual-channel intent parsing and function access method based on a large language model shown in any embodiment of the present invention by calling the computer program.

[0102] In one alternative embodiment, an electronic device is provided, such as Figure 4 As shown, Figure 4 The illustrated electronic device 4000 includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0103] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0104] Bus 4002 may include a path for transmitting information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 4002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4The bus 4002 is represented by only one thick line, but this does not mean that there is only one bus or one type of bus.

[0105] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0106] The memory 4003 stores application code (computer program) for executing the present invention, and its execution is controlled by the processor 4001. The processor 4001 executes the application code stored in the memory 4003 to implement the content shown in the foregoing method embodiments.

[0107] Among them, electronic devices can also be terminal devices. A terminal device can be any terminal device that can install applications and access web pages through applications, including at least one of smartphones, tablets, laptops, desktop computers, smart speakers, smartwatches, smart TVs, and smart in-vehicle devices.

[0108] It should be noted that, Figure 4 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0109] An embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the above-mentioned dual-channel intent parsing and function access methods based on a large language model.

[0110] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, and an optical data storage device, etc.

[0111] In an exemplary embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the aforementioned dual-channel intent parsing and function access method based on a large language model.

[0112] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0113] It should be understood that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0114] The computer-readable storage medium provided in this invention can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0115] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the method shown in the above embodiments.

[0116] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0117] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

[0118] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this invention can be specifically implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, this invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0119] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A dual-channel intent parsing and function access system based on a large language model, characterized in that, include: The user request input module is used to receive natural language requests input by users and preprocess the natural language requests to obtain preprocessed request text. The dual-channel intent parsing module is used to guide the large language model to perform intent parsing on the preprocessed request text through prompt word templates, and output the intent type corresponding to the natural language request, wherein the intent type is information consultation type or function execution type. The knowledge base question answering module is used to retrieve matching content from a preset knowledge base based on the preprocessed request text when the intent type is information consultation, input the matching content into the large language model, and obtain the response text generated by the large language model. The function menu mapping module is used to extract function action keywords and business object keywords from the preprocessed request text when the intent type is function execution type, and match the corresponding function menu node in the preset function menu mapping relationship table according to the function action keywords and the business object keywords. The function recommendation and redirection module is used to generate function entry recommendation information based on the function menu node, and in response to the user's trigger operation on the function entry recommendation information, redirect to the function page corresponding to the function menu node; The dialogue state management module is used to maintain the context state of multi-turn dialogues, record the intent type in the context state, and input the context state into the dual-channel intent parsing module at the start of the next round of dialogue.

2. The dual-channel intent parsing and function access system based on a large language model according to claim 1, characterized in that, The user request input module is specifically used for: The system receives the natural language request in speech form and converts the speech-to-text request into the natural language request in text form using a built-in speech-to-text unit.

3. The dual-channel intent parsing and function access system based on a large language model according to claim 2, characterized in that, The user request input module is specifically used to: perform word segmentation, entity recognition, and intent classification on the natural language request in text form to obtain the preprocessed request text.

4. The dual-channel intent parsing and function access system based on a large language model according to claim 1, characterized in that, The dual-channel intent parsing module has the built-in prompt word template. The prompt word template contains intent category definition information and parsing rule information. The intent category definition information is used to define two intent categories: information consultation and function execution. The parsing rule information is used to guide the large language model to output the corresponding intent type based on the preprocessed request text.

5. The dual-channel intent parsing and function access system based on a large language model according to claim 4, characterized in that, The preset knowledge base includes company introduction knowledge data, product consultation knowledge data, and claims consultation knowledge data. The matching content comes from at least one of the company introduction knowledge data, product consultation knowledge data, and claims consultation knowledge data.

6. The dual-channel intent parsing and function access system based on a large language model according to claim 4, characterized in that, The function menu mapping module is specifically used for: The functional action keywords and the business object keywords are compared with the keyword fields in the preset functional menu mapping table, and the functional menu nodes corresponding to the matching keyword fields are output as the matching results.

7. The dual-channel intent parsing and function access system based on a large language model according to claim 6, characterized in that, The function recommendation and redirection module is specifically used for: Generate function entry recommendation information in the form of function cards or function buttons, wherein the function entry recommendation information includes the function name and jump link address corresponding to the function menu node; In response to a user's click on the function card or the function button, the user is redirected to the function page corresponding to the function menu node based on the redirect link address.

8. A dual-channel intent parsing and function access method based on a large language model, characterized in that, The method employs the dual-channel intent parsing and function access system based on a large language model as described in any one of claims 1 to 7, comprising: Receive a natural language request input by the user, and preprocess the natural language request to obtain a preprocessed request text; The prompt word template guides the large language model to perform intent parsing on the preprocessed request text and outputs the intent type corresponding to the natural language request, which is either information consultation or function execution. When the intent type is information consultation, matching content is retrieved from the preset knowledge base based on the preprocessed request text, the matching content is input into the large language model, and the response text generated by the large language model is obtained. When the intent type is function execution type, extract function action keywords and business object keywords from the preprocessed request text, and match the corresponding function menu node in the preset function menu mapping relationship table according to the function action keywords and the business object keywords. Based on the function menu node, function entry recommendation information is generated, and in response to the user's trigger operation on the function entry recommendation information, the user is redirected to the function page corresponding to the function menu node.

9. An electronic device, characterized in that, The electronic device includes a processor coupled to a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor to enable the electronic device to implement the dual-channel intent parsing and function access method based on a large language model as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one computer program, which, when executed by a processor, implements the dual-channel intent parsing and function access method based on a large language model as described in any one of claims 1 to 7.